Linux PCI-HOWTO
 by Michael Will, [email protected]
 v0.6g, 30 March 1997

 Information on what works with Linux and PCI-boards and what does not.
 Please get the latest version of this document at The Linux Documenta-
 tion Project <http://sunsite.unc.edu/LDP/linux.html>
 ______________________________________________________________________

 Table of Contents



 1. Introduction

 2. Why PCI?

    2.1 General overview
    2.2 Performance
    2.3 The onboard-SCSI-II-chip NCR53c810
    2.4 Drew Eckhardt on PCI-SCSI:
    2.5 New Alpha Version of the NCR driver
    2.6 The EATA-DMA driver and the PCI SCSI controllers from DPT
    2.7 BT-946C fully supported with kernel 1.3.x and newer
    2.8 Future Domain TMC-3260 PCI SCSI
    2.9 other thoughts on scsi

 3. ASUS-Boards

    3.1 ASUS and the NMI (Parity) -- impact on Gravis-Ultrasound
    3.2 Various types of ASUS Boards
       3.2.1 ASUS SP3 with saturn chipset I (rev. 2) for 486,
       3.2.2 ASUS SP3G with saturn chipset II (rev. 4) for 486,
       3.2.3 ASUS SP3-SiS chipset, for 486
       3.2.4 ASUS AP4, for 486, with PCI/ISA/VesaLocalbus
       3.2.5 ASUS SP4-SiS, for Pentium90, PCI/ISA
       3.2.6 ASUS TP4 with Triton chipset and EDO-Support
       3.2.7 ASUS TP4XE with Triton chipset and additional SRAM/EDORAM support
       3.2.8 ...and many others now.
    3.3 Benchmarks on ASUS Mainboards
       3.3.1 ASUS SP3 with amd486DX4-100
       3.3.2 ASUS SP3 with amd486DX4-120
       3.3.3 ASUS SP3 with intel486DX2-66
       3.3.4 ASUS TP4/XE with intel586-90
       3.3.5 ASUS TP4/XE with intel586-100
    3.4 Detailed information on the old ASUS PCI-I-SP3 with saturn chipset from [email protected]:
    3.5 Pat Dowler ([email protected]) with ASUS SP3G

 4. confusion about saturn chipsets

 5. Video-Cards

 6. Ethernet Cards

    6.1 3com-3c590-tpo
    6.2 DEC435 PCI NIC

 7. Motherboards

    7.1 ASUS
    7.2 Micronics P54i-90
    7.3 SA486P AIO-II
    7.4 Sirius SPACE
    7.5 Gateway-2000
    7.6 Intel-Premiere
    7.7 DELL Poweredge SP4100 [email protected] - successful
    7.8 DELL OptiPlex Gl+ 575 [email protected] - successful when turning off plug and play
    7.9 Comtrade Best Buy PCI / PCI48X MB Rev 1.0
    7.10 IDeal PCI / PCI48X MB Rev 1.0
    7.11 CMD Tech. PCI IDE / CSA-6400C
    7.12 GA-486iS (Gigabyte)
    7.13 GA-586-ID (Gigabyte) 90 Mhz Pentium PCI/EISA Board
    7.14 ESCOM 486dx2/66 - which board?
    7.15 J-Bond with i486dx2/66
    7.16 super micro 011895 03:50 SUPER P54CI-PCI rev 1.3 (Opti)

 8. reports on success

    8.1 GigaByte GA486-AM with AMD Am5x86-133-WB @ 160MHz (40MHz PCI)
    8.2 California Graphics - Sunray II Pro
    8.3 Micronics P54i-90 ([email protected])
    8.4 Angelo Haritsis ([email protected]) about SA486P AIO-II:
    8.5 [email protected] about his Micronics M5Pi
    8.6 Simon Karpen ([email protected]) with Micronics M54pi
    8.7 Goerg von Below ([email protected]) about DELL Poweredge
    8.8 [email protected] about Gateway2000 P-66
    8.9 James D. Levine ([email protected]) with Gateway2000
    8.10 [email protected] with SPACE
    8.11 [email protected] with INTEL
    8.12 Jermoe Meyers ([email protected]) with Intel Premiere
    8.13 Timothy Demarest ([email protected]) Intel Plato Premiere II
       8.13.1 Flash Bios upgrades
       8.13.2 NCR 53c810 BIOSless PCI SCSI
       8.13.3 apart from that - plug and play!
    8.14 [email protected] with ASUS
    8.15 [email protected] with ASUS
    8.16 Lars Heinemann ([email protected]) with ASUS
    8.17 [email protected] with ASUS
    8.18 robert logan ([email protected] with GW/2000)
    8.19 [email protected] and his friend use ASUS
    8.20 Michael Will with ASUS-SP3 486 (the old one)
    8.21 Mike Frisch ([email protected]) Giga-Byte 486IM
    8.22 Karl Keyte ([email protected]) Gigabyte GA586 Pentium
    8.23 [email protected] with G/W 2000
    8.24 Joerg Wedeck ([email protected]) / ESCOM
    8.25 Ulrich Teichert / ASUS

 9. Reports of problems

    9.1 Compaq PCI systems, especially Presarios
    9.2 VLSI Wildcat PCI chipset like in Zeos P120 box
    9.3 [email protected] G/W 2000
    9.4 [email protected] (Frank Hofmann) / ASUS
    9.5 [email protected] (Axel Mahler) / ASUS
    9.6 Frank Strauss ([email protected]) / ASUS
    9.7 [email protected] / ASUS
    9.8 [email protected] / GigaByte
    9.9 Steve Durst ([email protected]) with UMC 8500 mainboard
    9.10 Tom Drabenstott ([email protected]) with Comtrade / PCI48IX

 10. General tips for PCI-Motherboard + Linux NCR PCI SCSI

    10.1 DON'Ts:
    10.2 SIMM slots
    10.3 Praised PCI Pentium motherboard
    10.4 irq-lines
    10.5 Info about the different NCR 8xx family scsi chips:
       10.5.1 53C810
       10.5.2 53C815
       10.5.3 53C825
    10.6 future of 53c8xx
    10.7 Performance of the 53c810
    10.8 News about NCR53c825 support
    10.9 Frederic POTTER ([email protected]) about Pentium+NCR+Strap_bug
    10.10 PCIprobe in the latest Linux Kernels by Frederic Potter
    10.11 Other PCI Devices
       10.11.1 Cyclades: a 16-port PCI RISC-based multiport card.

 11. Conclusion

 12. Thanks

 13. copyright/legalese

 14. GPL - Gnu Public License
 ______________________________________________________________________

 1.  Introduction


 Many people, including me, would like to run Linux on a PCI-based
 machine.  Since it is not obvious which PCI motherboards and PCI cards
 will work with Linux and which do not, I conducted a survey and spent
 some hours to compile the information contained herein.

 If you have information to add, please mail me. If you have questions,
 feel free to ask.

 Help with my style/grammar/language is welcome as well. I am not a
 native- speaker of English and expect to make occasional mistakes.

 Note: "on-board chip" refers to a SCSI chip integrated onto the
 motherboard rather than on a PCI expansion card.

 Also, "quotes" herein may have slight context editing.


 2.  Why PCI?



 2.1.  General overview

 The PC-architecture has several BUS-Systems to choose from:

    ISA
       16 or 8bit, cheap, slow (usually 8Mhz), standard, many cards
       available>

    EISA
       32bit, expensive, fast, few cards available, fading>

    MCA
       32 or 16bit ex-IBM-proprietary, fast, becoming rare>

    VESA-Local-Bus
       32bit, based on 486 architecture, cheap, fast, many cards
       available>

    PCI-Local-Bus
       32bit (64 bit coming), cheap, fast, many cards available,
       nowadays standard>


 MCA worked fine, but never achieved much market, being used on only
 some early IBM PS/2 machines. There were very few cards.

 EISA was reliable, but rather expensive, and intended more for
 servers, than for the average user. It has the next fewest cards
 available.

 VESA-Local-Bus (VLB) had some problems with high bus-speeds, and was
 not very reliable, but mainly due to its low price and better-than-ISA
 performance, sold very well. Technically, it's almost a direct map of
 the 486 processor bus.  Most VESA boards should be stable by now.  At
 the beginning of 1996, many 486 motherboards still support VESA, but
 PCI is growing.  VESA busses are tied directly to the speed of the
 memory bus for 486's, or half the speed for Pentiums.

 PCI now has the advantage. Like EISA it is not proprietary. It is as
 faster than EISA or MCA, and cheaper.  Most current Pentium
 motherboards use the PCI bus; VESA is fading.  Virtualy all PCI
 motherboards and cards sold at the beginning of 1996 are 32 bit, and
 run at 0-33 MHz.

 Currently, most Pentium motherboards run the PCI bus at 1/2 the memory
 speed (ie: 33 MHz for the 66 MHz memory bus on the P66,P100,P133,P166;
 30 MHz for the 60 MHz memory bus on the P60,P90,P120,P150; and 25 Mhz
 on the 50 MHz memory bus of the P75).  This is probably true of Cyrix
 6x86 motherboards too.  NexGen 5x86 implemention isn't known.  The PCI
 spec does allow the PCI bus to be run asynchronously from the
 processor, (eg: 33 Mhz bus on P75), but this is not common yet.

 PCI 2.1 has been defined, allowing 64 bit PCI, and/or 0-66 MHz
 operations, but no x86 chipsets yet support these options.  64 bit PCI
 will probably appear first, in 32/64 bit dual compatible versions.
 That is, you will be able to mix 32 and 64 bit cards.  66 MHz PCI will
 take longer, as it's technically demanding, can only support one or
 maybe two slots per bridge, and may not work well with 33 MHz cards.

 PCI is not processor dependent like the VESA Local-Bus. This means you
 can use the winner-1000-PCI in an Alpha-driven-PCI computer as well as
 in a i486/Pentium-driven PCI computer, with the appropriate BIOS and
 software.  Beside Intel and DEC Alpha platforms, PCI is used on some
 PowerPC's.

 Some PCI variations to be aware of: some implementations support "Bus
 Master" cards in all PCI slots, some in only one slot, and some not at
 all; some implementations support "bridging" on cards and some do not.


 2.2.  Performance

 taken from Craig Sutphin's early Pro-PCI-Propaganda


      Unlike some local buses, which are aimed at speeding up
      graphics alone, the PCI Local Bus is a total system solu-
      tion, providing increased performance for networks, disk
      drives, full-motion video, graphics and the full range of
      high-speed peripherals. At 33 MHz, the synchronous PCI Local
      Bus transfers 32 bits of data at up to 132 Mbytes/sec. A
      transparent 64-bit extension of the 32-bit data and address
      buses can double the bus bandwidth (264 Mbytes/sec) and
      offer forward and backwards compatibility for 32 and 64-bit
      PCI Local Bus peripherals. Because it is processor-indepen-
      dent, the PCI Local Bus is optimized for I/O functions,
      enabling the local bus to operate concurrent with the pro-
      cessor/memory subsystem.  For users of high-end desktop
      PC's, PCI makes high reliability, high performance and ease
      of use more affordable than ever before; no trivial task at
      33 MHz bus-clock rates. Variable length linear or toggle
      mode bursting for both reads and writes improves write
      dependent graphics performance. By comprehending the loading
      and frequency requirements of the local bus at the component
      level, buffers and glue logic are eliminated.


 See the chapter about Benchmarks for some crude (and perhaps
 meaningless) benchmarks on ASUS PCI Boards with 486 and 586.


 2.3.  The onboard-SCSI-II-chip NCR53c810

 One very nice feature of some PCI mother boards is the NCR onboard-
 SCSI-II-chip, which is said to be as fast as the EISA-Adaptec-1742,
 but much cheaper. Drivers for DOS/OS2 are available. Drew Eckard has
 released his version of his NCR53c810-driver, which is in the standard
 kernel since v1.2.

 This works so well I sold my adaptec-1542B-ISA soon after I bought the
 ASUS SP3-saturn-chipset II PCI board, and found the onboard NCR-SCSI
 controller to be much faster.

 The NCR53c810-chip is onboard on some PCI-motherboards.  There are
 add-on-boards available too, for about US$ 70.00.

 There is only one thing I noticed did not work with the NCR-drivers
 when I tried them. Disconnect/Reconnect did not work, so using a SCSI-
 tape could be a pain, especially when using "mt erase" or the like
 blocks the whole SCSI-bus until it has finished. Since this was very
 unsatisfying for me, I bought one of these nice but expensive DPT PCI
 SCSI controller and had no such problems anymore.

 People have reported this problem has been solved by Drew by now.

 FreeBSD does support the NCR53c810 for quite a long time already,
 including Tagged Command Queues, FAST, WIDE and Disconnect for NCR
 53c810, 815, 825. Drew said, it would be possible to adapt the FreeBSD
 driver to Linux. I somewhere saw some patches to do exactly this, any
 pointer to the location?

 I personaly have the impression there are some important wheels
 invented more than once because of the differently evolving of FreeBSD
 and Linux. Some more cooperation could do both systems very well...


 2.4.  Drew Eckhardt on PCI-SCSI:


 Drew said on end of March 95 about the SCSI on PCI: (slightly edited
 for clarity in context)

 The Adaptec 2940, Buslogic BT946, BT946W, DPT PCI boards, Future
 Domain 3260, NCR53c810, NCR53c815, NCR53c820, and NCR53c825 all work
 for some definition of the word works.


 o  The Adaptec 2940 suffers from the same cabling sensitivity that
    plagues all recent boards, but otherwise works fine.

 o  The Future Domain boards are not busmasters, and the driver doesn't
    support multiple simultaenous commands.  If you don't (currently)
    need multiple simultaneous commands, get a NCR board, which will be
    cheaper and is busmastering.  If you need multiple simultaneous
    commands, get a Buslogic.

 o  The Buslogic BT956W will do WIDE SCSI with the Linux drivers
    (although you can't use targets 8-15), the Adaptec 2940W (with one
    line patch to the 2940 driver) won't, nor will the NCR53c820 and
    NCR53c825.

 o  The NCR boards are dirt cheap (< $ 70 US), are generally quite
    fast, but the driver currently doesn't support multiple
    simultaenous commands. Alpha which do neat things like
    disconnect/reconnect and synchronous transfer are now publicly
    available, see below.

 o  Emulux, Forex, and other unmentioned PCI SCSI controllers will not
    work.



 2.5.  New Alpha Version of the NCR driver


 Well, this is not exactly *that* new anymore, please try to he
 versions which are in the kernel by version 2.0.x before going for
 this entry.

 Alpha versions of the NCR driver which do neat things like
 disconnect/reconnect and synchronous transfers are now publically
 available.  Any one interested in playing with them should

 o  Join the NCR mailing list, by sending mail to
    [email protected] with subscribe ncr53c810 in the text.

 o  Get all of the readmes, and latest diffs file from
    ftp://tsx-11.mit.edu/pub/ALPHA/linux/SCSI/ncr53c810


 2.6.  The EATA-DMA driver and the PCI SCSI controllers from DPT


 The EATA-DMA scsi driver has undergone extensive changes and now also
 supports PCI SCSI controllers, multiple controllers and all SCSI
 channels on the multichannel SmartCache/Raid boards in all
 combinations of WIDE, FAST-20 (ULTRA) and DIFFERENTIAL.


 The driver supports all EATA-DMA Protocol (CAM document CAM/89-004
 rev. 2.0c) compliant SCSI controllers and has been tested with many of
 those controllers in mixed combinations.


 Those are:             (ISA)   (EISA) (PCI)
       DPT Smartcache: PM2011  PM2012B
       Smartcache III: PM2021  PM2022  PM2024
                               PM2122  PM2124
                               PM2322
       Smartcache IV:  PM2041  PM2042  PM2044
                               PM2142  PM2144
                               PM2322
       SmartRAID     : PM3021  PM3122
                               PM3222  PM3224
                                       PM3334
       and some controllers from NEC, AT&T, SNI, AST, Olivetti and Alphatronix.



 On a "base" DPT card (no caching or RAID module), a MC680x0 controls
 the bus-mastering DMA chip(s) and the SCSI controller chip.  The DPT
 SCSI card almost works like a SCSI coprocessor.

 The DPT card also will emulate an IDE controller/drive (ST506
 interface), which enables you to use it with all operating systems
 even if they don't have an EATA driver.

 On a card with the caching module, the 680x0 maintains and manages the
 on-board cacheing.  The DPT card supports up to 64 MB RAM for disk-
 cacheing.

 On a card with the RAID module, the 680x0 also performs the management
 of the RAID, doing the mirroring on RAID-1, doing the striping and ECC
 info generation on RAID-5, etc.

 The entry level boards utilize a Motorola 68000, the high-end, more
 raid specific DPT cards use a 68020, 68030 or 68040/40MHz processor.

 Official list prices range from $ 265 to $1.645 (January 18, 1996)


 Since I've been asked numerous times where you can buy those boards in
 Europe, I asked DPT to send me a list of their official European
 distributors. Here is a small excerpt:



 Austria: Macrotron GmbH            Tel:+43 1 408 15430   Fax:+43 1 408 1545
 Denmark: Tallgrass Technologies A/S Tel:+45 86 14 7000   Fax:+45 86 14 7333
 Finland: Computer 2000 Finnland OY Tel:+35 80 887 331    Fax:+35 80 887 333 43
 France : Chip Technologies         Tel:+33 1 49 60 1011  Fax:+33 1 49 599350
 Germany: Akro Datensysteme GmbH    Tel:+49 (0)89 3178701 Fax:+49 (0)89 31787299
 Russia : Soft-tronik               Tel:+7 812 315 92 76  Fax:+7 812 311 01 08
 U.K.   : Ambar Systems Ltd.        Tel:+44 1296 311 300  Fax:+44 296 479 461



 "IMHO, the DPT cards are the best-designed SCSI cards available for a
 PC.  And I've written code for just about every type of SCSI card for
 the PC.  (Although, in retrospect, I don't know why!) ;-)" Jon R.
 Taylor ([email protected]) President, Visionix, Inc.

 The latest version of the EATA-DMA driver and a Slackware bootdisk is
 available on: ftp.i-Connect.Net:/pub/Local/EATA


 Since patchlevel 1.1.81 the driver is included in the standard kernel
 distribution.

 The author can be reached under these addresses: [email protected]
 mainz.de or [email protected]


 2.7.  BT-946C fully supported with kernel 1.3.x and newer


 There is a driver in the 1.3.x kernels (available as a patch for the
 1.2.13 kernel) written by someone associated with buslogic that fully
 supports the 946C and ALL of it's features including strict round
 robin, tagged queueing, multiple scatter/gather, multiple mailboxes,
 IRQ sharing, and yes, 15 devices on Fast/Wide.  It is no longer
 necessary to use any ISA emulation with the driver (no DMA channel, no
 ISA address), and the driver is /fast/ and /stable/ (it's out of BETA
 and into full release).

 The driver is available on ftp.dandelion.com (the newest version can
 always be got by doing "get BusLogic*").  It supports ALL BusLogic
 controllers with the exception of the FlashPoint LT, which uses a
 different interface.  The driver is included in the 1.3.x kernels as
 standard for BusLogic devices.



 2.8.  Future Domain TMC-3260 PCI SCSI


 Rik Faith ([email protected]) informed me on Wed, 1 Feb 1995 about the
 Future Domain TMC-3260 PCI SCSI card being supported by the Future
 Domain 16x0 SCSI driver.  Newer information might be contained in the
 SCSI-HOWTO.


 o  Detection is not done well, and does not use standard PCI BIOS
    detection methods (someone who has a PCI board needs to send me
    patches to fix this problem).  So, you might have to fiddle with
    the detection routine in the kernel to get it detected.

 o  The driver still does not support multiple outstanding commands, so
    your system will hang while your tape rewinds.

 o  The driver does not support the enhanced pseudo-32bit transfer mode
    supported by recent Future Domain chips, so you will not get
    transfer rates as high as under DOS.

 o  The driver only supports the SCSI-I protocol, so your really fast
    hard disks will not get used at the highest possible throughput.
    (Again, fixes for all these problems are solicited -- no one is
    working on them at this time.)



 2.9.  other thoughts on scsi


 James Soutter ([email protected]) asked me to add the following
 information on Fast-Wide-SCSI-2:


      Fast Wide SCSI-2 is sometimes incorrectly called SCSI-3. It
      differs from the normal Fast SCSI-2 (like the Adapted
      1542B?) because it uses a 16 bit data bus rather than the
      more usual 8 bit bus. This improves the maximum transfer
      rate from 10 MB/s to 20 MB/s but requires the use of special
      Fast Wide SCSI-2 drives.

      The added performance of Fast Wide SCSI-2 will not
      necessarily improve the speed of your system.  Most hard
      disk drives have a maximum internal transfer rate of less
      than 10 MB/s and so one drive alone can not flood a FAST
      SCSI-2 bus.

      In Seagate's Oct 1993 product overview, only one Fast Wide
      SCSI-2 drive has an internal transfer rate of more than 10
      MB/s (the ST12450W).  Most of the drives have a maximum
      internal transfer rate of 6 MB/s or less, although the
      ST12450W is not the only exception to the rule.  In
      conclusion, Fast Wide SCSI is designed for the file server
      market and will not necessarily benefit a single user
      workstation style system.

      Rather than buying a PCI system with a SCSI interface on the
      motherboard, or rather than waiting for the NCR driver, you
      could purchase a separate PCI based SCSI card. According to
      Drew, the only PCI SCSI option that stands a chance of
      working is the Buslogic 946. It purports to be Adaptec 1540
      compatible, like the EISA/VESA/ISA boards in the series.

      Drew commented that other PCI based SCSI controllers are
      unlikely to be supported under Linux or the BSD's because
      the NCR based controllers are cheaper and more prevalent.


 I definitly recommend reading the SCSI HOWTO in regards to newer
 information about PCI SCSI drivers.

 Ernst Kloecker ([email protected]) wrote: (edited)

      Talus Corporation has finished a NS/FIP driver for PCI
      boards with NCR SCSI. It will be shipping very soon, might
      even be free because a third party might pay for the work
 and donate the driver to NeXT.


 Not every PCI-Board has got the chip. The old ASUS do, and one of the
 J-Bond boards does, too. (Most of the boards nowadays (6/95) do expect
 you to buy the NCR53c810 seperately.) Some vendors provide an
 alternative as you can read in Drew's text...

 The NCR-Chip is clever enough to work with drives formatted by other
 controllers, and should be no problem.


 3.  ASUS-Boards

 3.1.  ASUS and the NMI (Parity) -- impact on Gravis-Ultrasound

 The newer trition PCI-Mainboards in 1995 did not seem to support
 parity-SIMMS anymore. Since I usualy took the cheaper nonparity-SIMMS
 anyway, I did not consider this a problem until I put the Gravis-
 Ultrasound into my machine. Under DOS the SBOS-Driver and Setup/Test
 utility does complain about "nmi procedure disabled on this p.c.". The
 manual says I'd better get a better mainboard in that case, not very
 helpful.

 The gravis-ultrasound did work nice in the ASUS-SP3 and ASUS-SP4,
 inspite of this, but the gravis-ultrasound-max I have here got gmod to
 kernel panic on both boards, and sometimes when playing au-files via
 /dev/audio did strange things, like playing the rest of an older,
 previously played sound after the new one. The sounddriver does
 recommend a buffer of 65536 with the GUS Max instead of the small one
 like the GUS - why I do not know. I do not have such a problem with
 the newer ASUS TP4 XE boards, though.  Both are equipped with 1M DRAM
 onboard. These problems are probably not related to the NMI-problem,
 but because of the sounddriver?

 I heard not only ASUS but most of the newer PCI-Mainboards are lacking
 in parity/NMI-support.

 Strange enough - the ASUS-TP4 (Trition Chipset) does work with the GUS
 Max - it does load the SBOS-Driver. I have to admit, I am confused.


 3.2.  Various types of ASUS Boards



 3.2.1.  ASUS SP3 with saturn chipset I (rev. 2) for 486,


 o  2 x rs232 with 16550

 o  NCR53c810 onboard,

 o  slightly broken saturn-chipset I (rev. 2)


 3.2.2.  ASUS SP3G with saturn chipset II (rev. 4) for 486,

 like SP3, but less buggy saturn chipset


 3.2.3.  ASUS SP3-SiS chipset, for 486

 like AP4, but newer, SiS chipset, green functions and all the EIDE,
 rs232 with 2 16550 and centronics.  Only 2 SIMM Slots,  Does seem to
 work with AMD486DX4/120, but was not very reliably on NCR53c810 and
 various operating systems (Windows-NT, Windows95, OS2), after
 upgrading to a PentiumBoard ASUS SP4, all the problems vanished, so it
 must have been the board.  Still does seem to work nice for Linux,
 though.


 3.2.4.  ASUS AP4, for 486, with PCI/ISA/VesaLocalbus

 green functions, 1VL, 3 ISA, 4 PCI slots, only EIDE onboard, no fd-
 controller, no rs232/centronics. Very small size.

 does recognice AMD486DX2/66 as DX4/100 only. This can be corrected
 with soldering one pin (which?) to ground, but I would not recommend a
 board like this anyway.

 The one I tested was broken for OS2 and Linux, but people are said to
 use it for both.

 The VesaLocalbus-Slot is expected to be slower than the normal vesa-
 localbus boards because of the PCI2VL bridge, but without penalty to
 the PCI section.


 3.2.5.  ASUS SP4-SiS, for Pentium90, PCI/ISA

 like SP3-SiS, but for Pentium90/100.


 3.2.6.  ASUS TP4 with Triton chipset and EDO-Support

 has the Triton-Chipset for better performance and supports normal
 PS2-Simms as well as Fast-Page-Mode and EDO modules.


 3.2.7.  ASUS TP4XE with Triton chipset and additional SRAM/EDORAM sup-
 port

 supports the new EDORAM and upcoming SRAM standards. At least SRAM is
 said to considerabely increase performance. Did for some reason not
 accept the 8M PS2-SIMMS working ok in ASUS SP4, after changing them
 against others, bigger looking ones, (16 chips instead of 8 if I
 remember right) it worked ok. Has been tested with P90 and P100.


 3.2.8.  ...and many others now.

 if you have new information on problems with them, please report.


 3.3.  Benchmarks on ASUS Mainboards

 I tried to compare the speed of CPUs in two ASUS Mainboards: for 486 I
 tested the SP3 SiS (the one with one vesa-local-bus slot) and for 586
 I tested the ASUS TP4/XE, each with 16M RAM, always the same unloaded
 system with another CPU, with whetstone and dhrystone.

 I must admit, I have not read the benchmarks-faq yet, and will
 probably edit the section a loot soon. If you have any comments,
 please mail me.

 I am especially confused about the amd486DX4/100 being faster on
 dhrystones than the DX4/120 version? I did not see that kind of
 inconsistency on comparing the P90 and P100.


 Perhaps this was at fault: when I plugged in the amdDX4-100, I had the
 board jumpered for DX2-66. While the BIOS did report it as an DX4-100,
 the board might have used the wrong clockspeeds... but since DX2-66
 uses 33Mhz * 2 and DX4 uses 33Mhz * 3, this would have been correct?

 The board running with DX4-120 is jumpered to 40Mhz * 3 = 120 Mhz.

 Another thing I wonder about is why the whetstones-result does yield
 so even numbers on some machines?


 3.3.1.  ASUS SP3 with amd486DX4-100


 o  Dhrystone time for 500000 passes = 7 by 63559 dhrystones/second

 o  Whetstone time for 1000 passes = 5 by 200.0000 Whetstones/second


 3.3.2.  ASUS SP3 with amd486DX4-120


 o  Dhrystone time for 500000 passes = 8 by 56074 dhrystones/second

 o  Whetstone time for 1000 passes =  4 by 250.0000 Whetstones/second


 3.3.3.  ASUS SP3 with intel486DX2-66


 o  Dhrystone time for 500000 passes = 9 by 50761 dhrystones/second

 o  Whetstone time for 1000 passes = 7 by 142.8571 Whetstones/second


 3.3.4.  ASUS TP4/XE with intel586-90


 o  Dhrystone time for 500000 passes = 4 by 101010 dhrystones/second

 o  Whetstone time for 1000 passes = 3 by 333.3333 Whetstones/second


 3.3.5.  ASUS TP4/XE with intel586-100


 o  Dhrystone time for 500000 passes = 4 by 102040 dhrystones/second

 o  Whetstone time for 1000 passes = 2  by 500.0000 Whetstones/second


 3.4.  Detailed information on the old ASUS PCI-I-SP3 with saturn
 chipset from [email protected]:



 o  3 PCI, 4 ISA Slots (3x16, 1x8 Bit)

 o  ZIF Socket for the CPU

 o  room for 4 72pin-SIMMs (max. 128M)

 o  Award BIOS in Flash-Eprom

 o  Onboard: NCR-SCSI, 1par, 2ser (with FIFO), AT-Bus, Floppy

 The board does like most in that price class -- write-through cache,
 no write-back. This should not be significant, maybe 3% of
 performance.

 The BIOS supports scsi-drives under DOS/Windows without additional
 drivers, but with the board come additional drivers which are said to
 give better performance, for DOS/Windows(ASPI), OS2, Windows-NT, SCO-
 Unix, Netware (3.11 and 4, if interpreted correctly)

 Gert Doering ([email protected]) was saying the SCO-Unix-driver for
 the onboard-SCSI-Chip was not working properly. After two or three
 times doing: "time dd if=/dev/rhd20 of=/dev/null bs=100k count=500" it
 kernel-paniced...

 The trouble some people experienced with this board might be due to
 them using an outboard Adaptec-SCSI-Controller with "sync negotiation"
 turned on. (This predates the NCR driver release; hence the use of the
 Adaptec.) Please check that in the BIOS-Setup of the Adaptec-1542C if
 you use one and have problems with occasional hangups!

 There is a new version of the ASUS-Board which should have definitely
 less problems. It is called ASUS-PCI-I/SP3G, the G is important. It
 has the new Saturn-chipset rev. 4 and the bugs should be gone.  They
 use the Saturn-ZX-variant and the new SP3G has fully PCI conforming
 level-triggered (thus shareable), BIOS-configurable interrupts.  It
 has an on-board PS/2-mouseport, EPA-power-saving-modes and
 DX4-support, too. It performs excellently. If you can get the German
 computer magazine C't from July (?), you will find a test report where
 the ASUS-Board is the best around.

 Latest information about ASUS-SP3-G: You might experience crashes when
 using PCI-to-Memory-Posting. If you disable this, all works perfect.
 [email protected] said he believed it to be a
 problem of the current Linux-kernel  rather than the hardware, because
 part of the system still works when crashing, looking like a deadlock
 in the swapper, and OS2/DOS/WINDOZE don't crash at all.

 Someone else with a very old ASUS-SP3 (saturn-I chipset) reported
 crashes with using XFree86, which went away when he installed the very
 latest betaversion which seems to work around a bit of the problems.


 3.5.  Pat Dowler ([email protected]) with ASUS SP3G



 o  ASUS SP3G board (it is rev.4 == saturn II)

 o  AMD DX4-100 CPU (need to set jumper 36 to 1&2 rather than 2&3,
    otherwise it's set the same as other 486DXn chips)

 o  256K cache (comes with 15ns cache :-)

 o  16meg RAM (2x8meg)

 o  ET4000 ISA video card

 o  quantum IDE hard drive

 o  SMC Elitel16 combo ethernet card

 Unlike some other reports, I find the mouse pointer moves very smoothy
 under X (just like the ol' 386)  - it is jumpy under some, but not
 all, DOS games though...


 Performance is great!! I ran some large floating point tests and found
 the performance in 3x33 (100MHz) mode to be almost 1.5x that in 2x
 (66MHz) mode (large being 500x500 doubles - 4meg or so)... I was a
 little dubious about clock-tripling but I seem to be getting full
 benefit :-)

 The heavily configurable energy star stuff doesn't work with the
 current AMD DX4 chips - you need an SL chip

 I really need a SCSI disk and a PCI video card :-)

 (I had a phonecall by a person who had this problem with the buggy SMC
 FIFO chipset, after using X-window they hung.)


 4.  confusion about saturn chipsets


 Pat Duffy ([email protected]) said:


 Saturn I:  these are revisions 1 and 2 of the Saturn chipsets.
 Saturn II:  This is also called rev. 4 of the Saturn chipsets.

 As far as I know, rev. 3 never actually shipped, and (from a few people who
 have it) the SP3G now has rev. 4 (or Saturn II) in it.

 Confused?  Well, the only real definitive answer is to get ahold of the board
 and run the debug script in the PCI chipset list on it.  As far as I know,
 though, the SP3G board is indeed shipping with rev. 4 (Saturn II).



 5.  Video-Cards

 Linux people have successfully used # 9 XGE Level 12, ELSA Winner
 1000, and S3-928 video cards. The XFree86(tm)-3.1.1 does support
 boards with the tseng et4000/w32 in accelerated mode, as well as S3
 Vision 864 and 964 chipsets including boards like the ELSA Winner
 1000Pro and 2000Pro, Number Nine GXE64 and GXE64Pro, Miro Crystal
 20SV). Support in the S3 Server for the Chrontel8391 clock chip has
 been added.

 Trio32 and Trio64 S3 Boards like the SPEA V7 Mirage P64 PCI and  MIRO
 Crystal 40SV, are also supported, the Mach32 and Mach64 are supported
 in accelerated mode, too.

 The SVGA Driver

 16bpp mode (65K colors instead of the usual 256) support for Mach32
 boards as well as 32bpp for some S3 boards and the P9000 boards has
 been added.


 [email protected] reported:


 o  Diamond Stealth W32 (et4000/W32) -- Text mode works, X11 suffered
    from "pixel dust", unbearable never got it to work and returned it.

 o  # 9GXE L12 -- Works, virtual consoles corrupted when switched,
    fixed this with disabling the "fast dram mode" feature in his BIOS.
    Does not get a dot clock above 85, though.


 Genoa Phantom 8900PCI card seems to work well.  Genoa Phantom/W32 2MB
 does not work in an ASUS-Board.  Tseng 3000/W32i chipset seems to work
 well.  Spea-v7 mercury-lite works perfectly since XFree86(tm)-2.1.

 Spea V7 Mirage P64 PCI 2M with Trio64 works nice since
 XFree86(tm)-3.1.1


 ATI Graphics Ultra Pro for PCI with 2MB VRAM and an ATI68875C DAC run
 well as [email protected] tells us: "It's humming right along
 at 1280x1024 w/256 colors @74Hz non-interlaced. Looks great."

 Paradise WD90C33 PCI did lock up on screensaver/X - this has been
 solved in the newer versions of the kernel.  [email protected]
 (John Edward Bauer)

 miroChrystal 8S/PCI (1MB) S3 - no problem.

 Stephen Tweedie reported his Cirrus Logics 5434 PCI card works well.
 It is a 64bit with 2M and runs perfectly with the SVGA driver in 8, 16
 and 32 bit per pixel.



 6.  Ethernet Cards

 Of course the ISA-ethernet-cards still work, but people are asking for
 PCI-based ones. The author of many (if not most) ethernet- drivers
 said the following some time ago (unfortunately I have not managed to
 contact him about new information):


      From: Donald Becker ([email protected]) Subject:
      PCI ethernet cards supported?

      The LANCE code has been extended to handle the PCI version.
      I hope to get the PCI probe code (about a dozen extra lines
      in the LANCE driver) into the next kernel version.  I'm
      working on the 32 bit mode code.  I haven't yet started the
      21040 code.

      I'll write drivers for the PCnet32 mode and the DEC 21040.
      That will cover most of the PCI ethercard market.

      file://cesdis.gsfc.nasa.gov/pub/people/becker/whoiam.html


 In the new testkernels of 1.1.50 and above, the AMD-singlechip
 ethernetadapters are supported. With a pentium, they ought to then see
 900K/second ftps +(assuming an NCR PCI scsi controller) at about 20%
 cpu load. (AMD Lance).

 Anything based on the AMD PCnet/PCI chip should work at the time
 being. In the US the Boca board costs under US$ 70

 Geoffry Coram reported in the news that he got his 3com 590 TPO to
 work. He had to get the alpha driver from
 http://cesdis.gsfc.nasa.gov/linux/drivers.  Other drivers would be
 there as well.  Note
 http://cesdis.gsfc.nasa.gov/linux/drivers/vortex.html

 Donald Holmgren said he successfully attached his DEC DE435 (PCI) card
 to the local network on thin coax (BNC).  The DE435 driver checks the
 twisted pair connection first, then switches to the alternate port
 (jumper selectable as AUI or BNC) if the 10BaseT port fails.
 Jim Cusick uses the Boca BEN 1PI card on a thin coax network.  It
 works just fine.  You might want to check out:
 http://cesdis.gsfc.nasa.gov/linux/misc/boca-failure.html for details
 on the early failures of this card.  My second card, after sending one
 back for replacement, was marked "PN 4186".  The old one that did not
 work was "PN 4185".  Mandate this newer model when you order from you
 vendor.  At $ 70, the card is a good deal.

 Dave Platt recommends to stay off the Boca BEN1PI card at all costs.
 It would be unreliable due to design flaws, and Boca seems unable to
 really fix the problem. The 3Com 3c590 "Vortex" PCI card is available
 in a combo version (10BaseT, thin coax, and AUI).  The Linux driver
 for this card is not yet part of the release kernel, but is available
 from http://cesdis.gsfc.nasa.gov/linux/drivers/vortex.html and can be
 patched into the later 1.2.x kernels (as well as 1.3.x) without much
 difficulty.  The Linux driver does not support the interface
 autodetect feature of this card - you must use the DOS utility to
 configure the card for the interface you wish to use (thin coax in
 this case).  Once you've done that, the Linux driver will use the
 correct interface.

 He has been using a 3c590 for several weeks, and it is working fine.

 Dave Kennedy said he got two of the above Boca boards and they work
 fine under light load, but under heavy work like ftping two 16M files
 into both directions, they failed. He sent the boards back to Boca for
 a hardwarefix. After they soldered a couple of things
 (diodes/resistors) onto the card and sent them back, the cards worked
 fine regardless of load. The two cards have been in 7/24 use in two
 P90 systems without problems for 6 months now.

 Craig does not recommend it since Boca seems not to follow the AMD
 specs but he has been running them for 2 weeks without problems. He
 tested his NFS performance and has been moving large files to and from
 server (16M, 8M).  He also tried to do all his workin localy using his
 data files mounted by NFS and has had no problems. Performance seems
 to be 100 percent better (wrt to NFS performance) over his NE2000 ISA
 board. (editors note: but so would probably have been the ISA SMC
 Elite Ultra?)


 6.1.  3com-3c590-tpo


 Someone on usenet mentioned ht used the 3Com-3C590-TPO (EtherLink III
 - PCI).  He had to get the "3c59x.c" driver and "vortex.patch" to make
 it work with his 1.2.8 Linux kernel.


 6.2.  DEC435 PCI NIC


 The DEC435 PCI NIC is said to work great with the drivers included in
 the Slackwaredistribution - I'd say they are in the standard-kernel?


 7.  Motherboards

 The people who answered were using the following boards:


 7.1.  ASUS


 o  [email protected] - successful.

 o  [email protected] - half-successful, works, but...

 o  [email protected] (Ulrich Teichert), - successful.

 o  [email protected] - successful

 o  [email protected] - successful

 o  [email protected] - successful - but trouble with the serial port

 o  [email protected] and his friend - successful after solving
    IDE-puzzle

 o  Lars Heinemann ([email protected]) successful

 o  Michael Will ([email protected]) - successful.


 7.2.  Micronics P54i-90


 [email protected] succesful [email protected] successful
 [email protected] successful


 7.3.  SA486P AIO-II


 [email protected] successful



 7.4.  Sirius SPACE


 [email protected] - successful


 7.5.  Gateway-2000

 [email protected] - no problems except the soundcard he tries to swap
 [email protected] - successful, but...  robert
 logan ([email protected]) - flawless.  James D. Levine
 ([email protected]) - flawless.


 7.6.  Intel-Premiere

 [email protected] - successful [email protected] - successful
 [email protected] - successful (Premier-II)


 7.7.  DELL Poweredge SP4100 [email protected] - successful

 7.8.  [email protected] - successful when turning off plug and
 play DELL OptiPlex Gl+ 575

 7.9.  Comtrade Best Buy PCI / PCI48X MB Rev 1.0

 [email protected] - "Works, I believe it has buggy Saturn chipset.
 I would also like to add: I strongly recommend not buying from
 Contrade.  Their service is horrible. "



 7.10.  IDeal PCI / PCI48X MB Rev 1.0

 [email protected] - "Did not work with PCI48X motherboard"


 7.11.  CMD Tech. PCI IDE / CSA-6400C

 [email protected] - "Works"


 7.12.  GA-486iS (Gigabyte)

 [email protected] - success with problems.


 7.13.  GA-586-ID (Gigabyte) 90 Mhz Pentium PCI/EISA Board

 [email protected] - succesful


 7.14.  ESCOM 486dx2/66 - which board?

 Works perfect except the ftape-streamer (archive)


 7.15.  J-Bond with i486dx2/66

 Drew Eckhardt ([email protected]) uses Diamond Stealth 64
 VRAM with 4M of memory (964 based). It works great, he usualy runs it
 at 1024x768 72hz in 32bpp; 16 and 8bpp also work. He needed to get the
 X311u2S3.tgz server from ftp.xfree86.org; people with 968 based
 Diamond boards will definately need to do this.



 7.16.  super micro 011895 03:50 SUPER P54CI-PCI rev 1.3 (Opti)


 Manuel de Vega Barreiro

 o  board    super micro 011895 03:50 SUPER P54CI-PCI rev 1.3

 o  Opti chipset: 82c557,82c556,82c558,82c621.

 o  4 PCI, 4 ISA Slots (4x16 Bit)

 o  ZIF Socket for CPU (120,100,90,75 mHz)

 o  4 72 pin-SIMMs (max 128Mb)

 o  cache 256,512,1024 Kb L2-cache

 o  Ami WinBIOS in Flash-Eprom (101094-VIPER-P)

 o  onboard: EIDE for 4 drives

 o  Pentium with 90Mhz, 8M (now 16M) RAM and 256K L2-cache.

 o  1 maxtor 540 Mb, 1 st3122A 1Gb

 o  Number Nine 9GXE64pro with 2Mb

 o  Sound blaster 16 + cdrom Matsushita

 o  17" microscan 5ep ADI monitor

    I run  linux 1.1.57  (now 1.2.1) without problems.  dosemu0.53 work
    fine (com. software like kermit and xtalk) XFree86 3.1 at 1024x768
    resolution



 8.  reports on success



 8.1.  GigaByte GA486-AM with AMD Am5x86-133-WB @ 160MHz (40MHz PCI)


 GigaByte GA486-AM


 o  AMD Am5x86-133-WB @ 160MHz (40MHz PCI)

 o  BIOS as of 11/07/95 (Rev.A)

 o  256KB 2nd level cache (15ns)

 o  48MB RAM (Mixed 60/70ns)

 Hercules Terminator 64/VIDEO (S3 765 or "Trio 64V+")

 Sound Blaster 16

 o  Panasonic CR563 CD-ROM drive

 Silicon 4Ser/3Par I/O

 o  Mouse

 o  Terminal

 o  Terminal

 o  Modem (14k4)

 o  HP Laserjet III

 Mitsumi CD-ROM controller

 o  FX001D drive

 Longshine 1MBit Floppy controller

 o  IOMega Tape Insider 250

 o  3,5" Floppy

 o  5,25" Floppy

 No Network card, because the 4 ISA slots are full, and I don't have a
 PCI card.  I (now) use kernel 2.0.22 with APM enabled, and the hard
 drives power down and up properly without panics.  The system is 24hrs
 up a day and still running. Kernel compilation takes between 5 and 7
 minutes, depending on options.



 8.2.  California Graphics - Sunray II Pro


 Guido Trentalancia ([email protected]) reported the California
 Graphics - Sunray II Pro with Triton chipset to work well with
 Pentium100, Hd: Conner cfs420a, Conner cfs210a, crunching numbers at
 147492 dhrystones/second.


 8.3.  Micronics P54i-90 ([email protected])


 Pentium with 90Mhz, 32M RAM and 512K L2-cache. Works extremely well (a
 kernel recompile takes 10 minutes :-).


 The board includes:

 o  UART - two 16550A high speed UARTS

 o  ECP - one enhanced parallel port

 o  Onboard IDE controller

 o  Onboard floppy controller

 Pros: Currently, I'm using it with an Adaptec 1542CF and a 1G Seagate
 drive, No problems. Graphics is ATI Graphics Pro Turbo (PCI). Very
 fast. The serial ports can keep up with a TeleBit T3000 modem (38400)
 without overruns.  Caching above 16M does occur. There are 3 banks of
 SIMM slots (2 SIMM's per bank), with each bank capable of 64M each (2
 32M 72-pin SIMM's). Each bank must be filled completely to be used
 (I'm only using bank 0 with 2 16Mx72-pin SIMM's). The CPU socket is a
 ZIF type socket. The BIOS is Phoenix, FLASH type.

 Drawbacks: RAM is expandable to 192M, but the L2 cache is maxed at
 512K. While the graphics are very fast, there is currently no XF86
 server for the Mach64 (well, actually there is, but it doesn't use any
 of the accelerator features; it's just an SVGA server). I don't know
 if the onboard IDE hard drive controller works; I'm prejudiced against
 a standard that won't allow my peripherals to operate across
 platforms, so I didn't buy an IDE disk; instead, I got a Seagate
 31200N and a NEC 3Xi.

 Mitch


 8.4.  Angelo Haritsis ([email protected]) about SA486P AIO-II:

 The motherboard I eventually bought (in the UK) is one supporting 486
 SX/DX/DX2/DX4 chips. It is called SA486P AIO-II. Features include:


 o  Intel Saturn v2 chipset

 o  Phoenix BIOS (flash eprom option)

 o  NCR scsi BIOS v 3.04.00

 o  256K 15ns cache (max 512) write back and write through

 o  4 72-pin SIMM slots in 2 banks

 o  3 PCI slots, 4 ISA


 o  On-board NCR 53c810 scsi controller

 o  On-board IDE / floppy / 2 x 16550A uarts / enhanced parallel

 I bought it from a company (UK) called ICS, (note I have no
 connections whatsoever with the company, just a happy customer). I use
 a 486/DX2-66 CPU.

 Before I had a VLB 486 m/board with a buslogic BT-445S controller that
 I was borrowing. I have 2 scsi devices: 1 barracuda 2.1GB ST12550N
 disk and a Wangtek 5525ES tape drive.  I was expecting a lot of
 adventures by switching to the new motherboard, esp after hearing all
 these non-success stories on the net. To my surprise everything worked
 flawlessly on the 1st boot! (1.1.50). And it has been doing so for
 about a month now. I did not even have to repartition the disk:
 apparently the disk geometry bios translation of the 2 controllers is
 the same.  Linux has had no problems at all. SCSI is visibly much
 faster as well (sorry, I have no actual performance measurements).

 The only problems (related to Drew's linux ncr53c7,810 scsi driver -
 thanks for the good work Drew!) are:

 o  no synchronous transfers are yet supported => performance hit

 o  disconnect/reconnect is disabled => disk scsi ops "hold" during
    certain slow scsi device opeartions (eg tape rewind)

 o  tagged queuing is not there (?) => performance hit

 If you get Windows complainingg about 32-bit disk driver problems,
 just disable 32-bit disk access via Control Panel. This should not
 hurt performance. (What I did is remove the WDCTRL driver from my
 SYSTEM.INI).

 All else is fine. I tried the serial ports with some dos/windows s/w
 and worked ok. The IDE/floppy work ok as well. I have not tried the
 parallel yet. The motherboard is quite fast and so far I am very
 pleased with the upgrade. I have not yet tried a PCI graphics board. I
 will later on. I am using an old ISA S3 which is fine at the moment.

 PS: the NCR drivers in the 2.0.x kernels should have no problems of
 that kind anymore. please consult the SCSI-HOWTO for further and
 hopefully more uptodate information.


 8.5.  [email protected] about his Micronics M5Pi

 Micronics M5Pi motherboard with 60 MHz Pentium, PCI bus having the
 following components:


 16Mb RAM/512k cache
 onboard IDE, parallel, 16550A UARTS
 2 X 340MB Maxtor IDE Hard Drives
 Soundblaster 16 SCSI-II
 Toshiba 3401B SCSI CD-ROM
 Archive Viper 525MB SCSI Tape Drive
 Viewsonic 17 monitor
 Cardex Challenger PCI video card (ET4000/W32P)
 A4-Tech Serial Mouse



 Everything works great, Slackware installation was very easy, I can
 run Quicken 7 for DOS under DOSEMU. I run X at 1152x900 resolution at
 67Hz.
 8.6.  Simon Karpen ([email protected]) with Micronics
 M54pi

 I have had no problems with the above board, the on-board PCI IDE
 (hopefully soon will also have SCSI), and an ATI Mach32 (GUP) with 2MB
 of VRAM.


 8.7.  Goerg von Below ([email protected]) about DELL Poweredge



 - Intel 486DX4/100
 - 16 MB RAM
 - DELL SCSI array (DSA) with Firmware A07, DSA-Manager 1.7
 - 1 GB SCSI HD DIGITAL
 - NEC SCSI CD-ROM
 - 2 GB internal SCSI streamer
 - 3-Com C579 EISA Ethernet card
 - ATI 6800AX PCI VGA subsystem, 1024 MB RAM

 CAVE! DELL SCSI Array controller (DSA) runs only with firmware Rev. A07 !
 A06 is buggy, impossible to reboot !
 To get it: ftp dell.com , file is /dellbbs/dsa/dsaman17.zip



 Apart from this firmware-problem there where no problems for the last
 2 months, running with linux 1.1.42 as primary nameserver, newsserver
 and www-server on internet.


 8.8.  [email protected] about Gateway2000 P-66

 Gateway2000's P5-66 system with Intel's PCI motherboard, with 5 ISA
 slots and 3 PCI slots.  The only PCI card I am using is the # 9 GXe
 level 12 PCI card (2 MB VRAM and 1 MB DRAM). This card was bought from
 Dell. Under Linux I am using the graphics in the 80x25 mode only (I am
 waiting for some XFree86 refinements before using it in 1280x1024
 resolution), but under DOS/Windows I have used the card in
 1280x1024x256 mode without problems. Etherlink 3C509 Ethernet card,
 Mitsumi bus-interface card, Adaptec 1542C SCSI interface card and
 additional serial/parallel ports card (which makes the total of serial
 ports 3).

 I have total of 32 MB RAM (recognized and used by both Linux and DOS).
 There is also a bus mouse (Microsoft in the PS2 mode).

 No problems so far.


 8.9.  James D. Levine ([email protected]) with Gateway2000

 Gateway 2000 P5-60 with an Intel Mercury motherboard, AMI-Flash-BIOS,
 (1.00.03.AF1, (c)'92) 16M RAM, on-board IDE controller and an ATI AX0
 (Mach32 Ultra XLR) PCI display adapter. He had absolutely no problems
 with the hardware so far but has not tried anything fancy, such as
 accelerated IDE drivers or SCSI support.


 8.10.  [email protected] with SPACE

 SPACE-board, 8MB RAM, S3 805 1MB DRAM PCI 260MB Seagate IDE-hard disk
 because of lack of NCR53c810-Driver, 0.99pl15d, does seem to work
 well.

 8.11.  [email protected] with INTEL

 17 machines running a 60Mhz-i586 on Intel-Premier-PCI-Board


 8.12.  Jermoe Meyers ([email protected]) with Intel Premiere


 Motherboard - Intel Premiere Plato-babyAT 90mhz with Buslogic bt946c
 w/4.86 mcode w/4.22 autoSCSI firmware, (note, mine came with 4.80
 mcode and 4.17 autoSCSI firmware. (interrupt pins A,B,C conform to
 respective PCI slots!) ATI Xpression (Mach64) - using driver from
 sunsite, (running AcerView 56L monitor).

 The motherboard has 4 IDE drives, Linux (Slackware 2.0) sees the first
 two and everything on the Buslogic as it emulates an adaptec 1542.
 Uh, yes, Dos sees them all.  Buslogic is VERY accomodating in regards
 to shipping upgraded chips (you will have to know how to change PLCC
 (plastic leaded chip carrier) chips, 3 of them.  Though, don't let
 that scare you :-) it's not that tough.  Get a low end PLCC removal
 tool, and your in business.  You also might want to "flash upgrade
 your system bios from Intel's IPAN BBS, a trivial process.  Whats even
 more interesting is I also have a Sound Blaster SCSI-2 running a scsi
 CDROM drive off it's adaptech 1522 onboard controller.  So thats 4 IDE
 drives (2 under Linux) and 2 SCSI-2 controllers.


 I hope this helps others who are struggling with PCI technology use
 Linux!  Jerry ([email protected])


 8.13.  Timothy Demarest ([email protected]) Intel Plato Premiere II

 My system is configured as follows:  16Mb 60ns RAM, 3Com Etherlink-III
 53C809 ethernet card (using 10base2), ATI Mach 64 2Mb VRAM, Toshiba 2x
 SCSI CDROM, NCR 53c810 PCI SCSI, Syquest 3270 270Mb Cartridge Drive,
 Viewsonic 17 monitor, Pentium-90 (FDIV Bug Free).  Running Slackware
 2.1.0, Kernel 1.2.0, with other misc patches/upgrades.

 Everything is functioning flawlessly.  I dont recommend the Syquest
 drives.  I have used the 3105 and the 3270 and both a very, very
 fragile.  Also, the cartridges are easily damaged and I have had
 frequent problems with them.  I am in the process of looking for
 alternative removable storage (MO, Zip, Minidisc, etc).

 Some information you might need:


 8.13.1.  Flash Bios upgrades

 Flash Bios updates can be ftp'd from
 wuarchive.wustl.edu:/pub/MSDOS_UPLOADS/plato. The current version is
 1.00.12.AX1.  The BIOS upgrades *must* be done in order.  1.00.03.AZ1
 to 1.00.06.AX1 to 1.00.08.AX1 to 1.00.10.AX1 to 1.00.12.AX1.  The
 Flash BIOS updates can also be downloaded from the Intel BBS.  I do
 not have that number right now.


 8.13.2.  NCR 53c810 BIOSless PCI SCSI

 If you are using an NCR 53c810 BIOSless PCI SCSI card in the Plato,
 you may have trouble getting the card to be recognized.  I had to
 change one of the jumpers on the NCR card:  the jumper that controls
 whether there is 1 or 2 NCR SCSI cards in your system must be set to
 "2".  I dont know why, but this is how I got it to work.  The other
 jumper controls the INT setting (A,B,C,D).  I left mine at A (the
 default).


 8.13.3.  apart from that - plug and play!

 There are no settings in the motherboard BIOS for setting the NCR
 53c810.  Dont worry - once the card is jumpered correctly, it will be
 recognized!  So much for PCI Plug-n-Play!


 8.14.  [email protected] with ASUS

 ASUS-PCI-Board (SP3) having:



 o  -- Asus PCI-Board with AMD 486/dx2-66 and 16M RAM

 o  -- Fujitsu 2196ESA 1G SCSI-II

 o  -- Future Domain 850MEX Controller (cheap-SCSI-Controller, almost a
    clone to Seagate's ST01... want's to use ncr53c810 as soon as the
    driver comes out

 o  -- ATI Graphics Ultra (the older one with Mach-8 Chip, ISA-Bus)

 o  -- Slackware 1.1.1

 He just exchanged the boards, plugged his cards in, connected the
 cables, and it worked perfect. He does not use any PCI-Cards yet,
 though.


 8.15.  [email protected] with ASUS

 ASUS-PCI-Board with 486DX66/2, miro-crystal 8s PCI driven by the
 S3-drivers of XFree86-2.0, using the onboard SCSI-Chip. No problems
 with compatibility at all.



 8.16.  Lars Heinemann ([email protected]) with ASUS

 ASUS PCI/I-486SP3 Motherboard w/ 486DX2/66 and 16M RAM (2x8),
 miroChrystal 8S/PCI (1MB) S3, Soundblaster PRO, Adaptec 1542b (3.20
 ROM) SCSI host adapter with two hard disks (Fujitsu M2694ESA u.
 Quantum LPS52) and a QIC-150 Streamer attached.  No problems at all!


 8.17.  [email protected] with ASUS

 ASUS PCI/I-486SP3 / i486DX2-66 / 8 MB PS/2 70 ns BIOS: Award v 4.50
 CPU TO DRAM write buffer: enabled CPU TO PCI write buffer: enabled PCI
 TO DRAM write buffer: disabled, unchangeable CPU TO PCI burst write:
 enabled Miro Crystal 8s PCI - S3 P86C805 - 1MB DRAM

 Quantum LPS 540S SCSI-Harddisk on NCR53c810-controller.


 8.18.  robert logan ([email protected] with GW/2000)


 Gateway 2000 4DX2-66P 16 Megs RAM, PCI ATI AX0 2MB DRAM (ATI GUP).  WD
 2540 Hard Disk (528 Megs) CrystalScan 1776LE 17inch. (Runs up to
 1280x1024) Slackware 1.1.2 (0.99pl15f)

 It is giving no problems. He uses SLIP for networking and an Orchid-
 Soundwave-32 for niceties, awaiting the NCR-Driver.  The only problem
 he has is that the IDE-Drive could be much faster on the PCI-IDE. It
 is one of the new Western Digital fast drives and in DOS/WfW it
 absolutely screams - on Linux it is just as slow as a good IDE-Drive.


 8.19.  [email protected] and his friend use ASUS

 Archie and his friend have rather similar configurations:


 o  ASUS PCI-SP3 board (4 ISA, 3 PCI)

 o  Intel 486DX2/66

 o  Genoa Phantom 8900PCI card (friend: Tseng 3000/W32i chipset)

 o  Maxtor 345 MB IDE hard drive

 o  Supra 14.4 internal modem

 o  ViewSonic 6e monitor (Archie)

 o  NEC Multisync 4fge (friend)

 o  Slackware 1.2.0

 The onboard-SCSI is disabled. First there were problems with the IDE-
 drive: ``on the board there's a jumper which selects whether IRQ14
 comes from the ISA bus or the PCI bus. The manual has an example where
 they show connecting it to PCI INT-A. Well, we did that just like the
 example... but then later our IDE drive would not work (the IDE
 controller is on board). Had to take it back. The guys at NCA were
 puzzled, then traced it back to this jumper. I guess the IDE
 controller uses IRQ14 or something? That's not documented anywhere in
 the manual. Other than that, seems to be kicking ass nicely now.
 Running X, modeming, etc. (for the Supra you have to explicitly tell
 the kernel that the COM port has a 16550A using setserial (in
 Slackware /etc/rc.d/rc.serial))''.


 8.20.  Michael Will with ASUS-SP3 486 (the old one)


 used the following:


 o  ASUS PCI-SP3-Board with 486dx2/66 and 16M RAM

 o  NCR53c810-SCSI-II chip driving a 1GB-Seagate-SCSI-II disk and a
    Wangtec-tape

 o  ATI-GUP PCI Mach32 Graphics card with 2M VRAM running perfectly
    with XFree86(tm)-3.1 8bpp and 16bpp

 o  Linux kernel 1.1.69

 It runs perfectly and I am content with the speed, the ATI-GUP-PCI
 (Mach32) does not give as good benchmarks as expected, though. Since I
 got the money by now, I got me an ASUS-SP4 with P90 which gives me
 better throughput on Mach32-PCI...  If I had even more money I'd get
 me another 16M of RAM and a Mach64-PCI with 4M RAM, though... I still
 keep on dreaming :-)


 8.21.  Mike Frisch ([email protected]) Giga-Byte 486IM



 o  Motherboard: Giga-Byte 486IM

 o  Configuration: 4 ISA slots (2 double as VLB) and 4 PCI slots

 o  CPU: Intel 486DX/33

 o  BIOS: Award 4.50G

 o  PCI EIDE Disk Controller: Giga-Byte GA-107 (CMD 640x PCI Multi-I/O)

 o  PCI Video card: ATI Graphics eXpression PCI 2MB DRAM

 o  Linux Kernel: 1.2.9

 o  Linux Dist'n: Highly modified Slackware 2.2.0

 I have been running this board 24 hours a day for the past 5-6 months.
 It has worked flawlessly for me under DOS/Windows, OS/2 Warp, and
 Linux (with Linux being run usually 24 hours a day).



 8.22.  Karl Keyte ([email protected]) Gigabyte GA586 Pentium


 o  PCI/EISA Board Gigabyte GA586-ID 90MHz Pentium (dual processor, one
    fitted)

 o  32M RAM

 o  SCSI - no scsi-NCR-chip on-board, using Adaptec 1542C,

 o  PCI ATI GUP 2M VRAM

 o  Adaptec 1742 EISA SCSI controller

 o  Soundblaster 16

 o  usual I/O


 Everything under DOS AND Linux works perfectly.  No problem
 whatsoever.  A VERY fast machine!  BYTE Unix benchmarks place it about
 the same as a Sun SuperSPARC-20 running Solaris 2.3.  The PC is faster
 for integer arithmetic and process stuff (including context
 switching).  The SPARC is faster for floating point and one of the
 disk benchmarks.



 8.23.  [email protected] with G/W 2000

 He uses a Gateway 2000 with no problems, except the soundcard (which
 one?). He is trading it in for a genuine soundblaster in hopes that
 will help.


 8.24.  Joerg Wedeck ([email protected]) / ESCOM

 originaly buyed a 486 DX2/66 from ESCOM (which board?) with onboard
 IDE and without (!) onboard NCR-SCSI-chip. ISA-adaptec 1542cf scsi-
 controller instead spea v7 mercury lite (s3, PCI, 1MB), ISA-
 Soundblaster-16, mitsumi-cdrom (the slower one).  Everything except
 the archive-streamer works with no problems.  The spea-v7 works
 perfectly since XFree86-2.1

 He abandoned the Intel-board in favour of an ASUS-SP3-g and has some
 problems with PCI-to-Memory burstmode which is crashing only on Linux,
 "looking like a deadlock in the swapper". If you have any information
 on this, please eMail the maintainer of the PCI-HOWTO.

 After turning off the PCI-to-Memory posting feature it just works
 perfect.

 Rather than sending him mail please read his http-homepage at
 "http://wsiserv.informatik.uni-tuebingen.de/ jw" where he keeps
 information about his PCI-system, too.


 8.25.  Ulrich Teichert / ASUS


 ASUS-PCI board with AMD486dx40 (but actually running at 33Mhz?!)  His
 ISA-ET3000 Optima 1024A ISA works nice. No problems with Quantum540S
 SCSI Harddisk attached to the onboard NCR53c810.



 9.  Reports of problems



 9.1.  Compaq PCI systems, especially Presarios


 Patrick Yaner ([email protected]) reported a Compaq-speciality to
 me. It seems they are mapping the PCI BIOS data area to an obscure
 area of memory, one that Linux (or OS2) cannot access. It can usually
 find it, but it can't get in, and gives a message on startup
 (something like "pcibios_init: entry in high memory area, unable to
 access").  Although this is alright with the display (which is on the
 PCI bus) and the IDE controller (also PCI), it means any other PCI
 devices -- such as an Ethernet card -- cannot be detected by Linux.

 Compaq offers a driver for DOS at
 ftp://ftp.compaq.com/pub/softpaq/Drivers/SP1116.ZIP

 but using this with linux would mean using the program that boots
 linux from DOS, instead of LILO.  Note that Compaq occasionally
 updates the software in this archive, so the file
 ftp://ftp.compaq.com/pub/softpaq/allfiles.html (also available as
 allfiles.txt) might be handy in checking to see that they haven't
 upgraded.

 Oddly, this information can also be found in the SCSI HOWTO, although
 the Pressarios come with IDE built in.


 9.2.  VLSI Wildcat PCI chipset like in Zeos P120 box


 Paul Bame ([email protected]) reported:

 The Wildcat PCI chipset works fine in late 1.3 and all 2.0 kernels.



 9.3.  [email protected] G/W 2000


 Gateway 2000 G/W 2000 4DX2/66 PCI ATI-Graphics-Ultra-Pro IDE of
 indeterminate make

 It works well - only the IDE-Card runs in ISA-compatibility-mode, and
 works a lot faster when switched into PCI-Mode by a DOS-program...
 thus it's not that fast in Linux, and a patch would be nice.


 9.4.  [email protected] (Frank Hofmann) / ASUS


 He uses the ASUS-board with 16MB-RAM, ISA-based S3/928, and the
 onboard-IDE-controller with a Seagate ST4550A harddisk. He's had no
 trouble with the newer Linux-kernels.

 His problem:

      using X, my mouse is not responding the way I was used to before. It's
      sometimes behind movement and makes jumps if moved quickly. I think
      this was discussed In a Linux newsgroup before (I don't know which
      one) and is due to the use of 16550 serial chips for the onboard
      serial interfaces. After two weeks, I got used to it :-)


 Reducing the threshold of the 16550 should help. There should be a
 patch to setserial available somewhere, but I do not know where.



 9.5.  [email protected] (Axel Mahler) / ASUS


 ASUS PCI/I-486SP3 Motherboard (Award BIOS 4.50), 16 MB RAM the on-
 Board NCR Chip is disabled, he had the Genoa Phantom/W32 2MB for PCI
 and a Adaptec AHA-1542CF (BIOS v2.01) connected to:

 o  an IBM 1.05 GB Harddisk

 o  a Toshiba CD-ROM (XM4101-B)

 o  a HP DAT-Streamer (2GB)

 when creating the filesystems, 'mke2fs' (0.4, v. 1.11.93) hung and
 installation was impossible. After replacing the Genoa Phantom/W32 2MB
 PCI with an ELSA Winner 1000 2MB PCI it worked perfectly.  He tested
 it with an old Eizo VGA-ISA and it worked as well, so the problem was
 in the Genoa-PCI-card.


 9.6.  Frank Strauss ([email protected]) / ASUS

 ASUS SP3 Board i486DX2/66 NCR53c810 disabled Adaptec 1542B in ISA Slot
 with 2 hard drives (200MB Maxtor, 420MB Fijutsu), SyQuest 88MB and
 Tandberg Streamer ELSA Winner 1000 PCI, 1MB-VRAM Soundblaster Pro in
 ISA Slot at IRQ 5 Onboard IDE disabled Onboard serial, parallel, FD
 enabled

 After a reset, the machine sometimes 'hangs' (soft and hard-reset the
 same) - this is probably not related to the Adaptec and the Soundcard,
 because even without these the system sometimes fails to come up. But
 if it runs, (and the ELSA-WINNER-1000-PCI-message appears) it runs ok.


 The two serial ports are detected as 16550 as they should, but at some
 mailbox-sessions there was heavy data-loss at V42bis... The problem
 seems to be in the hardware...


 CPU>-PCI-Burst seems to work well with DOS/MS-Windows

 CPU->PCI-Burst does not work properly with linux0.99p15, Messing up
 when switching the virtual-consoles, crashing completely when calling
 big apps like ghostview, or xdvi, leaving the SCSI-LED on (!).

 (I suspect these apps would be using a lot of CPU->PCI-burst because
 of the big heap of data to transmit to the PCI-Winner-1000)

 After disabling CPU->PCI-Burst, it works well, the Winner-1000 at
 1152x846 (not much font cache with 1MB) does 93k xstones. OpaqueMove
 with twm is more than just endureable :-)

 He has got a SATURN.EXE which he loads under DOS before starting
 Linux, helping to turn on burst without hangs...

 Someone stated that these problems might go away when turning off
 "sync negotiation" on the Adaptec - I do not know if this is possible
 with the adaptec1542B too? But I guess so.

 With CPU->PCI-Burst it yielded 95k xstones, so he considers it as not
 too grave to do without. His only problem is that he would like to run
 his Winner-1000 at 1152x900 which fails because it seems to take any
 x-resolution higher than 1024pixels as a 1280pixel-resolution, thus
 wasting a lot end resulting in a y-resolution of 816pixels... but this
 is probably no PCI-related problem. It should have gone away with
 XFree86-2.1


 9.7.  [email protected] / ASUS



 o  BOARD  ASUS PCI/I-486 SP3      RAM: 16MB (4x4M-SIMM)

 o  CPU    486DX33 CPU

 o  BIOS   Ver. 4.50 (12/30/93)

 o  Floppy         Two floppy drives (1.2 and 1.44), using ASUS on-
    board floppy controller

 o  SCSI tried both WD7000 SCSI controller and Adaptec 1542CF and
    worked.

 o  Two SCSI 320M hard drives

 o  SCSI NEC84 CDROM drive

 o  SCSI QIC150 Archive tape drive

 o  Video - Tseng ET4000 ISA graphics card

 o  Sound PAS16 sound card

 o  Printer attached to on-board ASUS parallel port

 He has nothing in the PCI-Slots yet, but wants to buy a PCI-Video-
 Card, currently uses WD7000 SCSI controller but will switch to the
 NCR-Chip onboard as soon as the driver is out.

 Everything works perfectly - the first serial port which has a 14.4K-
 Modem attached does hang occasionally when reconnecting with the modem
 after having used it previously.  He says that would not be unique to
 ASUS but rather a bug in the SMC-LSI device with its 16550UART. The
 logitech-serial-mouse on the second port works fine. Setting down the
 threshold of the 16550 for the mouseport would definitely help, one
 does seem to need a special patched setserial for that? I have not got
 the information yet, please contact me if you know more!



 9.8.  [email protected] / GigaByte


 o  Board - GA-486iS from Gigabyte w/ 256Kb 2L-Cache, i486-DX2

 o  Bios - AMI, 93/8

 o  SCSI - no scsi-NCR-chip on-board, using Adaptec 1542C,

 o  Video - ELSA Winner 1000

 o  Linux 0.99pl14 + SCSI-Clustering-Patches / Slackware 1.1.1


 All seems to go well, but he has not tried neither networking,
 printing or a streamer yet. Before applying the clustering- patches he
 had some problems with hangs triggered by "find", but this no longer
 is the case - perhaps it was an older kernel-bug.

 The ELSA-Winner-1000 sometimes hangs, with very strange patterns on
 the screen resolved only by rebooting... The dealer has told him it
 was a bug in the ELSA-Card, but the manufacturer claims it had solved
 the problem. The bug is not reproducible so he does not plan to take
 any action at the moment.

 All in all the machine seems to work very well under heavy text
 processing (emacs, LaTeX, xfig, ghostview) usage.  Interaction is
 surprisingly responsive, little difference between it and the 3-4X as
 expensive Sun he works on...

 CPU->PCI-Burst is still disabled because the bios does not support the
 PCI-things well?

 A problem with his new modem (v32 terbo) arose: it looses characters.
 Especially when using SLIP it complains a lot about RX and TX errors.
 As soon as he runs X it gets unusable. He said he activated FIFO and
 RTS/CTS with stty, but to no avail...



 9.9.  Steve Durst ([email protected]) with UMC 8500 mainboard

 Running Linux 1.2.12 on the UMC8500-100Mhz motherboard with the
 dreaded CMD PCIO640B (E)IDE controller, when booting the screen
 wiggles a few seconds, as if the Diamond Stealth64-DRAM (S3 864) has
 to warm up first, but he can live with that.



 9.10.  Tom Drabenstott ([email protected]) with Comtrade / PCI48IX


 PCI48IX Motherboard Rev. 1.0. Made by ??? documentation copyrighted by
 "exrc". The BIOS says not very much about PCI.

 His E-315E Super IDE UMC (863+865) ISA-Controller-card does have
 problems. (It is a multifunction controller-card). It seems to work
 well under DOS/OS2 but not under Linux.


 10.  General tips for PCI-Motherboard + Linux NCR PCI SCSI


 This was compiled by Angelo Haritsis ([email protected]) from various
 people's postings:


 10.1.  DON'Ts:


 Do *NOT* go for combination VLB/PCI motherboards. They usually have a
 lot of problems. Get a plain PCI version (with ISA slots as well of
 course).  A lot of bad things have been heard about OPTI chipset PCI
 motherboards.  Someone hints: "Avoid the OPTi (82C596/82C597/82C822)
 chipset based motherboards like the TMC PCI54PV".

 (I know of at least one person having no problems with his TMC PCI54PV
 motherboard. He just had to put the NCR53c810 addonboard into slot-A
 which is the only slot capable of busmastering as it seems.)

 Rumours say that Intel chipset PCI motherboards will have problems
 with more than one bus-mastering PCI board. I have not tried this one
 yet on mine and have nothing to suggest. I also heard that the Saturn
 II chipset is problematic, but this is the one I use and it is
 perfectly ok! Advice: Try to negotiate a 1-2 week money back agreement
 with your supplier, in case the motherboard you get has problems with
 the use you plan for it.


 10.2.  SIMM slots


 Go for 72-pin only SIMMs for speed: Some (all?) of the mainboards
 which take 30 pin SIMMs use a 32 bit main memory interface, and will
 be significantly slower than the Intel based boards which all use a 64
 bit or permantly interleaved memory interface.  You might want to keep
 that in mind.


 10.3.  Praised PCI Pentium motherboard


 The P90 Intel motherboard with the Intel Premiere II chipset (aka
 Plato). Get the latest BIOS which has concatenated NCR scsi BIOS
 3.04.00. Otherwise DOS won't see your scsi disk(s) if you use a BIOS-
 less 53c810 based controller.  NCR SCSI BIOS exists in the AMI BIOS of
 the plato after version 1.00.08 (or maybe verion 1.00.06). This BIOS
 is FLASH upgradeable so you should be able to get the upgrade on a
 floppy from your supplier. The current version is 1.00.10 and has all
 early problems fixed.

 (Bios files should be available at ftp.demon.co.uk:/pub/ibmpc/intel,
 but I did not check that myself. the Autor.)


 10.4.  irq-lines


 The value in the interrupt line PCI configuration register is usually
 set manually (for compatability with legacy ISA boards) in the
 extended CMOS setup screens on a per-slot or per-device basis.  Older
 PCI mainboards also force you to set jumpers for each PCI slot/device
 which select how PCI INTA and perhaps INTB, INTC, and INTD are mapped
 to an 8259 IRQ line, Obviously, if these jumpers exist on your board,
 they must match the settings in the extended CMOS setup.  Also note
 that some boards (notably Viglens) have silkscreens and instruction
 manuals which disagree with the wiring, and some experimentation may
 be in order.


 10.5.  Info about the different NCR 8xx family scsi chips:


 All NCR 8XX Chips are dircet connect PCI bus mastering devices, that
 have no preformance difference wether on motherboard or add in option
 card.  All devices comply with PCI 2.0 Specification, and can burst 32
 bit data at the full 33 MHz (133Mbytes/Sec)


 10.5.1.  53C810

 53C810 = 8 bit Fast SCSI-2 (10 MB/Sec) Single ended only Requires
 Integrated Mother board BIOS 100 pin Quad Flat Pack (PQFP) Worlds
 first PCI SCSI Chip, Volumes make it the most inexpensive.


 10.5.2.  53C815

 53C815 = 8 bit Fast SCSI-2 (10 MB/Sec) Single Ended only Support ROM
 BIOS interface, which makes it ideal for add-in card Designs. 128 Pin
 QFP


 10.5.3.  53C825

 53C825 = 8 bit Fast SCSI-2, Single ended or Differential 16 bit Fast
 SCSI-2 (20 MB/Sec), Single ended or Differetial Also has support for
 external Rom, making it a good candidate for add in cards. 160 pin QFP
 Not supported by linux yet. (See section below on news about the 825).
 Must have devices with wide or differential scsi to use these
 features.



 10.6.  future of 53c8xx

 There are 4 new devices planned for announcement late this year and
 into early next year.  Footprint compitible with 810 and 825 with some
 new features.

 All the Chips require a BIOS in DOS/Intel applications.  The 810 is
 the only chip that needs it resident on the motherboard. Latest NCR
 SCSI BIOS version: 3.04.00 The bios supports disks >1GB, indeed up to
 8G under MS-LOSS.


 10.7.  Performance of the 53c810

 C't magazine's DOS benchmarks showed that it was significantly faster
 than the Buslogic BT-946, one user noted a 10-15% performance increase
 versus an Adaptec 2940, and with a very fast disk it may be 2.5X as
 fast as an Adaptec 1540.



 10.8.  News about NCR53c825 support

 works. period.


 10.9.  Frederic POTTER ([email protected]) about Pen-
 tium+NCR+Strap_bug

 On some Intel Plato board,  the NCR bios doesn't recognize the board,
 because it needs to see the board as a "secondary SCSI controller",
 and because on most SCSI board the jumper to select between
 primary/secondary has been ironed to primary (to spare 1 cent,
 presumably).

 Solution:

 near the NCR chip, they are 3 via ( kind of holes ) with a strap like
 that
                 O--O  O

         this mean primary is selected as default setting. For the Plato Intel
         Mainboard, it should be like that

                 O  O--O

         The best solution is to get rid of the strap and to put a 2 position
         jumper instead.



 10.10.  PCIprobe in the latest Linux Kernels by Frederic Potter


 Frederic Potter has added a PCI-Probe into the latest kernels. If you
 do a "cat /proc/pci" it should list all your cards. If you own cards
 which are not properly recogniced, please contact him via mail as
 "[email protected]".

 See arch/i386/kernel/bios32.c and include/linux/pci.h in the kernel
 source for more information on PCI-Probe-Stuff.


 10.11.  Other PCI Devices

 What other PCI-cards are supported? Apart from various graphicscards,
 I would like to know about other cards like ethernet, framegrabber, or
 the TSET boards Cyclades is about to beta-test at the moment:


 10.11.1.  Cyclades: a 16-port PCI RISC-based multiport card.



 The product is called Cyclom-Ye, and has the following
 characteristics:


 o  PCI host card based on the PLX chip-set. This host card supports 8
    to 32 serial ports, utilizing 8 or 16-port external boxes.

 o  SCSI II cable.

 o  8 or 16-port external boxes with RJ45 or DB25 connectors (your
    choice).  You can start with 8 ports and expand to 32, by just
    adding more boxes. Each external box contains 2 or 4 CD-1400 RISC
    Serial controllers (each CD-1400 controls 4 serial ports).

 o  Up to 4 Host cards can be installed in the PC system, allowing a
    maximum of 128 serial ports per system.


 The product is being in the beta-test phase at July the 26th, 1995,
 and should be available by Octobre or something. eMail them at
 [email protected].



 11.  Conclusion


 If you have some moneny to put into your machine, you'd be well off
 with a Pentium90, ASUS-SP4, which is what I use at the moment. If you
 can afford 32M RAM that would be much better than 16M RAM.

 Real soon now the upcoming standard will be the Triton Chipset with
 support for special SIMMS called EDODRAM, and SRAM. Both will be more
 expensive than PS2-RAM, and at the time of writing (28-June-1995) SRAM
 is not available. While EDO-DRAM is more expensive, this is not
 because of the production costs, they are said to be the same.

 For a highperformance system I would still choose an ASUS-TP4/XE with
 EDO-DRAM, but if you do not need to use it at the moment, I d rather
 wait some more.

 For Graphic-boards I'd say the best cheap board fitting perfectly on a
 good Multisync-15 like the Samsung SyncMaster 15Gli, is the SPEA V7
 Mirage P64 with Trio64 Chipset and 2M DRAM. For more sophisticated
 Display like the Iiyama-IDEK 8617A-T I think the PCI Mach64 ATI-GUP-
 Turbo (not the cheaper GUP-Turbo-Windows) would be a good choice, with
 4M RAM you can have truecolor in higher resolutions. It is well
 supported in the XFree86(tm)-3.1.1, and there are commercial X-Servers
 available of which I'd recommend Accelerated/X by Roell, which
 supports the Mach64 very well and fast.

 For SCSI I'd take the DPT rather than the (much cheaper and very fast)
 NCR53c810 in case you plan to use SCSI-Tapes a lot. The NCR53c810
 driver on Linux does lack disconnect/reconnect support, thus blocking
 the SCSIbus on operations like "mt rewind", "mt fsf" etc. It bears a
 performance penalty on tar-operations - but check out Drews new alpha
 drivers before making a decision, perhaps it does solve all the
 problems.

 For building servers, the DPT would be the controller of choice anyway
 because of all the nifty hardware cache (with elevator sorting on
 accesses, so cache it is not a silly thing even in a Linux enviroment
 where the OS does the caching) and RAID-Support up to raid level 5.

 If you do not want to spend that much money on computer equipment
 (e.g.: you are having a life) you might go for an ASUS-SP3-SiS with
 AMD-DX2/66 or DX4/100. The SPEA V7 Mirage P64 PCI with 2M DRAM would
 be a good choice, since it uses the Trio64 S3 Chip, which is well
 supported by XFree86(tm)-3.1.1, quite cheap to buy and fast, too.

 Another fine card since XFree86(tm)-3.1 is the fast and cheap
 et4000/w32-PCI-card.


 12.  Thanks

 I want to thank the following people for supporting this document:

 o  David Lesher ([email protected]) for extensive help with the
    english language

 o  Nathanael MAKAREVITCH ([email protected]) for translating into
    french

 o  Jun Morimoto ([email protected]) for translating into
    japanese

 o  Marco Melgazzi ([email protected]) for translating into
    italian

 o  Donald Becker ([email protected]) for ethernet-
    informations

 o  Drew Eckhardt ([email protected]) for SCSI-informations

 o  Zhahai Stewart ([email protected]) for help with the intro section

 and many more peole adding information mostly by mail and by posts,
 some of them will be named here:

 [email protected],
 [email protected],
 [email protected] (Working at the PCI-NCR53c810-Driver),
 [email protected],
 [email protected],
 [email protected],
 [email protected],
 [email protected] (iX-Magazine),
 [email protected],
 [email protected],
 [email protected],
 [email protected],
 [email protected],
 [email protected],
 [email protected],
 [email protected],
 [email protected],
 [email protected],
 [email protected],
 [email protected],
 Ulrich Teichert, [email protected],
 [email protected],
 [email protected]
 [email protected],
 [email protected],
 Gert Doering ([email protected]),
 James D. Levine ([email protected]),
 Georg von Below ([email protected]),
 Jerome Meyers ([email protected]),
 Angelo Haritsis ([email protected]),
 [email protected] and his friend [email protected].



 13.  copyright/legalese

 (c)opyright 1993,94 by Michael Will - the GPL (Gnu Public License)
 applies. See last section about this.

 If you sell this HOWTO on a CD or in a book I would be happy to have a
 copy for reference.


 ([email protected])

 Contact me, either via eMail or call +49-7071-969063.

 Trademarks are owned by their owners. There is no warranty on the
 information in this document.

 For german users I am offering tested, preinstalled / preconfigured
 and supported Linux-PCI-machines. Call me at 07071-969063.


 14.  GPL - Gnu Public License



                     GNU GENERAL PUBLIC LICENSE
                        Version 2, June 1991

  Copyright (C) 1989, 1991 Free Software Foundation, Inc.
                           675 Mass Ave, Cambridge, MA 02139, USA
  Everyone is permitted to copy and distribute verbatim copies
  of this license document, but changing it is not allowed.

                             Preamble

   The licenses for most software are designed to take away your
 freedom to share and change it.  By contrast, the GNU General Public
 License is intended to guarantee your freedom to share and change free
 software--to make sure the software is free for all its users.  This
 General Public License applies to most of the Free Software
 Foundation's software and to any other program whose authors commit to
 using it.  (Some other Free Software Foundation software is covered by
 the GNU Library General Public License instead.)  You can apply it to
 your programs, too.

   When we speak of free software, we are referring to freedom, not
 price.  Our General Public Licenses are designed to make sure that you
 have the freedom to distribute copies of free software (and charge for
 this service if you wish), that you receive source code or can get it
 if you want it, that you can change the software or use pieces of it
 in new free programs; and that you know you can do these things.

   To protect your rights, we need to make restrictions that forbid
 anyone to deny you these rights or to ask you to surrender the rights.
 These restrictions translate to certain responsibilities for you if you
 distribute copies of the software, or if you modify it.

   For example, if you distribute copies of such a program, whether
 gratis or for a fee, you must give the recipients all the rights that
 you have.  You must make sure that they, too, receive or can get the
 source code.  And you must show them these terms so they know their
 rights.

   We protect your rights with two steps: (1) copyright the software, and
 (2) offer you this license which gives you legal permission to copy,
 distribute and/or modify the software.

   Also, for each author's protection and ours, we want to make certain
 that everyone understands that there is no warranty for this free
 software.  If the software is modified by someone else and passed on, we
 want its recipients to know that what they have is not the original, so
 that any problems introduced by others will not reflect on the original
 authors' reputations.

   Finally, any free program is threatened constantly by software
 patents.  We wish to avoid the danger that redistributors of a free
 program will individually obtain patent licenses, in effect making the
 program proprietary.  To prevent this, we have made it clear that any
 patent must be licensed for everyone's free use or not licensed at all.

   The precise terms and conditions for copying, distribution and
 modification follow.

                     GNU GENERAL PUBLIC LICENSE
    TERMS AND CONDITIONS FOR COPYING, DISTRIBUTION AND MODIFICATION

   0. This License applies to any program or other work which contains
 a notice placed by the copyright holder saying it may be distributed
 under the terms of this General Public License.  The "Program", below,
 refers to any such program or work, and a "work based on the Program"
 means either the Program or any derivative work under copyright law:
 that is to say, a work containing the Program or a portion of it,
 either verbatim or with modifications and/or translated into another
 language.  (Hereinafter, translation is included without limitation in
 the term "modification".)  Each licensee is addressed as "you".

 Activities other than copying, distribution and modification are not
 covered by this License; they are outside its scope.  The act of
 running the Program is not restricted, and the output from the Program
 is covered only if its contents constitute a work based on the
 Program (independent of having been made by running the Program).
 Whether that is true depends on what the Program does.

   1. You may copy and distribute verbatim copies of the Program's
 source code as you receive it, in any medium, provided that you
 conspicuously and appropriately publish on each copy an appropriate
 copyright notice and disclaimer of warranty; keep intact all the
 notices that refer to this License and to the absence of any warranty;
 and give any other recipients of the Program a copy of this License
 along with the Program.

 You may charge a fee for the physical act of transferring a copy, and
 you may at your option offer warranty protection in exchange for a fee.

   2. You may modify your copy or copies of the Program or any portion
 of it, thus forming a work based on the Program, and copy and
 distribute such modifications or work under the terms of Section 1
 above, provided that you also meet all of these conditions:

     a) You must cause the modified files to carry prominent notices
     stating that you changed the files and the date of any change.

     b) You must cause any work that you distribute or publish, that in
     whole or in part contains or is derived from the Program or any
     part thereof, to be licensed as a whole at no charge to all third
     parties under the terms of this License.

     c) If the modified program normally reads commands interactively
     when run, you must cause it, when started running for such
     interactive use in the most ordinary way, to print or display an
     announcement including an appropriate copyright notice and a
     notice that there is no warranty (or else, saying that you provide
     a warranty) and that users may redistribute the program under
     these conditions, and telling the user how to view a copy of this
     License.  (Exception: if the Program itself is interactive but
     does not normally print such an announcement, your work based on
     the Program is not required to print an announcement.)

 These requirements apply to the modified work as a whole.  If
 identifiable sections of that work are not derived from the Program,
 and can be reasonably considered independent and separate works in
 themselves, then this License, and its terms, do not apply to those
 sections when you distribute them as separate works.  But when you
 distribute the same sections as part of a whole which is a work based
 on the Program, the distribution of the whole must be on the terms of
 this License, whose permissions for other licensees extend to the
 entire whole, and thus to each and every part regardless of who wrote it.

 Thus, it is not the intent of this section to claim rights or contest
 your rights to work written entirely by you; rather, the intent is to
 exercise the right to control the distribution of derivative or
 collective works based on the Program.

 In addition, mere aggregation of another work not based on the Program
 with the Program (or with a work based on the Program) on a volume of
 a storage or distribution medium does not bring the other work under
 the scope of this License.
   3. You may copy and distribute the Program (or a work based on it,
 under Section 2) in object code or executable form under the terms of
 Sections 1 and 2 above provided that you also do one of the following:

     a) Accompany it with the complete corresponding machine-readable
     source code, which must be distributed under the terms of Sections
     1 and 2 above on a medium customarily used for software interchange; or,

     b) Accompany it with a written offer, valid for at least three
     years, to give any third party, for a charge no more than your
     cost of physically performing source distribution, a complete
     machine-readable copy of the corresponding source code, to be
     distributed under the terms of Sections 1 and 2 above on a medium
     customarily used for software interchange; or,

     c) Accompany it with the information you received as to the offer
     to distribute corresponding source code.  (This alternative is
     allowed only for noncommercial distribution and only if you
     received the program in object code or executable form with such
     an offer, in accord with Subsection b above.)

 The source code for a work means the preferred form of the work for
 making modifications to it.  For an executable work, complete source
 code means all the source code for all modules it contains, plus any
 associated interface definition files, plus the scripts used to
 control compilation and installation of the executable.  However, as a
 special exception, the source code distributed need not include
 anything that is normally distributed (in either source or binary
 form) with the major components (compiler, kernel, and so on) of the
 operating system on which the executable runs, unless that component
 itself accompanies the executable.

 If distribution of executable or object code is made by offering
 access to copy from a designated place, then offering equivalent
 access to copy the source code from the same place counts as
 distribution of the source code, even though third parties are not
 compelled to copy the source along with the object code.

   4. You may not copy, modify, sublicense, or distribute the Program
 except as expressly provided under this License.  Any attempt
 otherwise to copy, modify, sublicense or distribute the Program is
 void, and will automatically terminate your rights under this License.
 However, parties who have received copies, or rights, from you under
 this License will not have their licenses terminated so long as such
 parties remain in full compliance.

   5. You are not required to accept this License, since you have not
 signed it.  However, nothing else grants you permission to modify or
 distribute the Program or its derivative works.  These actions are
 prohibited by law if you do not accept this License.  Therefore, by
 modifying or distributing the Program (or any work based on the
 Program), you indicate your acceptance of this License to do so, and
 all its terms and conditions for copying, distributing or modifying
 the Program or works based on it.

   6. Each time you redistribute the Program (or any work based on the
 Program), the recipient automatically receives a license from the
 original licensor to copy, distribute or modify the Program subject to
 these terms and conditions.  You may not impose any further
 restrictions on the recipients' exercise of the rights granted herein.
 You are not responsible for enforcing compliance by third parties to
 this License.

   7. If, as a consequence of a court judgment or allegation of patent
 infringement or for any other reason (not limited to patent issues),
 conditions are imposed on you (whether by court order, agreement or
 otherwise) that contradict the conditions of this License, they do not
 excuse you from the conditions of this License.  If you cannot
 distribute so as to satisfy simultaneously your obligations under this
 License and any other pertinent obligations, then as a consequence you
 may not distribute the Program at all.  For example, if a patent
 license would not permit royalty-free redistribution of the Program by
 all those who receive copies directly or indirectly through you, then
 the only way you could satisfy both it and this License would be to
 refrain entirely from distribution of the Program.

 If any portion of this section is held invalid or unenforceable under
 any particular circumstance, the balance of the section is intended to
 apply and the section as a whole is intended to apply in other
 circumstances.

 It is not the purpose of this section to induce you to infringe any
 patents or other property right claims or to contest validity of any
 such claims; this section has the sole purpose of protecting the
 integrity of the free software distribution system, which is
 implemented by public license practices.  Many people have made
 generous contributions to the wide range of software distributed
 through that system in reliance on consistent application of that
 system; it is up to the author/donor to decide if he or she is willing
 to distribute software through any other system and a licensee cannot
 impose that choice.

 This section is intended to make thoroughly clear what is believed to
 be a consequence of the rest of this License.

   8. If the distribution and/or use of the Program is restricted in
 certain countries either by patents or by copyrighted interfaces, the
 original copyright holder who places the Program under this License
 may add an explicit geographical distribution limitation excluding
 those countries, so that distribution is permitted only in or among
 countries not thus excluded.  In such case, this License incorporates
 the limitation as if written in the body of this License.

   9. The Free Software Foundation may publish revised and/or new versions
 of the General Public License from time to time.  Such new versions will
 be similar in spirit to the present version, but may differ in detail to
 address new problems or concerns.

 Each version is given a distinguishing version number.  If the Program
 specifies a version number of this License which applies to it and "any
 later version", you have the option of following the terms and conditions
 either of that version or of any later version published by the Free
 Software Foundation.  If the Program does not specify a version number of
 this License, you may choose any version ever published by the Free Software
 Foundation.

   10. If you wish to incorporate parts of the Program into other free
 programs whose distribution conditions are different, write to the author
 to ask for permission.  For software which is copyrighted by the Free
 Software Foundation, write to the Free Software Foundation; we sometimes
 make exceptions for this.  Our decision will be guided by the two goals
 of preserving the free status of all derivatives of our free software and
 of promoting the sharing and reuse of software generally.

                             NO WARRANTY

   11. BECAUSE THE PROGRAM IS LICENSED FREE OF CHARGE, THERE IS NO WARRANTY
 FOR THE PROGRAM, TO THE EXTENT PERMITTED BY APPLICABLE LAW.  EXCEPT WHEN
 OTHERWISE STATED IN WRITING THE COPYRIGHT HOLDERS AND/OR OTHER PARTIES
 PROVIDE THE PROGRAM "AS IS" WITHOUT WARRANTY OF ANY KIND, EITHER EXPRESSED
 OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
 MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE.  THE ENTIRE RISK AS
 TO THE QUALITY AND PERFORMANCE OF THE PROGRAM IS WITH YOU.  SHOULD THE
 PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF ALL NECESSARY SERVICING,
 REPAIR OR CORRECTION.

   12. IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING
 WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MAY MODIFY AND/OR
 REDISTRIBUTE THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR DAMAGES,
 INCLUDING ANY GENERAL, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING
 OUT OF THE USE OR INABILITY TO USE THE PROGRAM (INCLUDING BUT NOT LIMITED
 TO LOSS OF DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY
 YOU OR THIRD PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER
 PROGRAMS), EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE
 POSSIBILITY OF SUCH DAMAGES.

                      END OF TERMS AND CONDITIONS

         Appendix: How to Apply These Terms to Your New Programs

   If you develop a new program, and you want it to be of the greatest
 possible use to the public, the best way to achieve this is to make it
 free software which everyone can redistribute and change under these terms.

   To do so, attach the following notices to the program.  It is safest
 to attach them to the start of each source file to most effectively
 convey the exclusion of warranty; and each file should have at least
 the "copyright" line and a pointer to where the full notice is found.

     <one line to give the program's name and a brief idea of what it does.>
     Copyright (C) 19yy  (name of author)

     This program is free software; you can redistribute it and/or modify
     it under the terms of the GNU General Public License as published by
     the Free Software Foundation; either version 2 of the License, or
     (at your option) any later version.

     This program is distributed in the hope that it will be useful,
     but WITHOUT ANY WARRANTY; without even the implied warranty of
     MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
     GNU General Public License for more details.

     You should have received a copy of the GNU General Public License
     along with this program; if not, write to the Free Software
     Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.

 Also add information on how to contact you by electronic and paper mail.

 If the program is interactive, make it output a short notice like this
 when it starts in an interactive mode:

     Gnomovision version 69, Copyright (C) 19yy name of author
     Gnomovision comes with ABSOLUTELY NO WARRANTY; for details type `show w'.
     This is free software, and you are welcome to redistribute it
     under certain conditions; type `show c' for details.

 The hypothetical commands `show w' and `show c' should show the appropriate
 parts of the General Public License.  Of course, the commands you use may
 be called something other than `show w' and `show c'; they could even be
 mouse-clicks or menu items--whatever suits your program.

 You should also get your employer (if you work as a programmer) or your
 school, if any, to sign a "copyright disclaimer" for the program, if
 necessary.  Here is a sample; alter the names:

   Yoyodyne, Inc., hereby disclaims all copyright interest in the program
   `Gnomovision' (which makes passes at compilers) written by James Hacker.

   (signature of Ty Coon), 1 April 1989
   Ty Coon, President of Vice

 This General Public License does not permit incorporating your program into
 proprietary programs.  If your program is a subroutine library, you may
 consider it more useful to permit linking proprietary applications with the
 library.  If this is what you want to do, use the GNU Library General
 Public License instead of this License.