Early PCI Board

Since some of this blog’s readers are very good at recognizing obscure hardware, I thought I’d post photos of one more somewhat exotic graphics card:

Early PCI Board (component side)

This board was one of the early PCI graphics adapters. It’s old enough that it doesn’t use the standardized PCI class ID. It’s based on the S3 928P chip and equipped with 1MB video memory.

The card as such is a fairly classic 1993 design: An accelerator chip (S3 928P), a DAC (Brooktree 485 at 135MHz), a separate clock chip (ICS2494M) plus a 14.318MHz oscillator, and memory (eight 128K chips). There’s very little else on this card (hint, hint). The card is equipped with a standard VGA connector, with the caveat that the cable must have one pin missing in order to fit (hello, needle-nose pliers!).

Early PCI Board (reverse side)

The board was manufactured in early 1994, possibly week 20, perhaps even slightly earlier. I believe the 928P chip was one of the very first PCI VGA controllers, and that is exactly why it had been chosen for this design.

This graphics card was made by a well known computer company, though it was probably never sold at retail. However, it is a production board and not a prototype. What is it?

Very Late Update: I finally got around to running lspci -vvnn -x on the board. This was not in the original PowerPC (too much hassle) but rather plugged into a SelectaDock III with a ThinkPad 600X attached, running an older Ubuntu. Here’s what I got:

08:04.0 VGA compatible unclassified device [0001]: S3 Inc. 86c928
 [Vision 928 VRAM] vers 0 [5333:88b0]
 Control: I/O- Mem- BusMaster- SpecCycle- MemWINV- VGASnoop-
          ParErr- Stepping- SERR- FastB2B- DisINTx-
 Status: Cap- 66MHz- UDF- FastB2B- ParErr- DEVSEL=medium
         >TAbort- <TAbort- <MAbort- >SERR- <PERR- INTx-
 Region 0: Memory at a8000000 (32-bit, non-prefetchable)
           [disabled] [size=64K]

00: 33 53 b0 88 00 00 00 02 00 00 01 00 00 00 00 00
10: 00 00 00 a8 00 00 00 00 00 00 00 00 00 00 00 00
20: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
30: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
This entry was posted in Graphics, PC hardware. Bookmark the permalink.

27 Responses to Early PCI Board

  1. Darkstar says:

    My guess is that it’s a NEC card from one of the later PC-9821’s.

    I’m pretty sure they used the ‘928 chip, so maybe they also used the PCI version (‘928P) in later models …

  2. Michal Necasek says:

    No, it’s not an NEC, and the card was not made a Japanese company (despite the fact that the graphics controller was made in Japan and the memory chips are from Hitachi).

  3. Stiletto says:

    How about Elsa Winner 1000 PCI 1MB? (er, if it exists)

  4. Michal Necasek says:

    …but clearly isn’t the same card 🙂 Here’s a hint–there’s one chip missing on this card, and that should say something about what sort of system the card was intended for.

    Incidentally, if someone knows what was the first PCI graphics card for PCs on the market (and when), I’m all ears. Possibly very late 1993, maybe early 1994.

  5. Calvin says:

    Is it for the first Power Macs?

  6. Michal Necasek says:

    No, but you’re thinking in the right direction.

  7. Rich Shealer says:

    Is it an 8514 clone? The early VGA spec had a keyed pin #9 and the 8514 used the VGA circuitry on the motherboard. Maybe this was for an early IBM PCI machine. I think the IBM ValuePoints had 486 and Pentium level CPUS with PCI in that time frame. The company where I worked in the 90â€ēs had sold a lot of the IBMs VPs to school districts, though none with 8514 graphics.

  8. Julien Oster says:

    I think I know what the missing chip is, and I think that makes it unlikely that this card was used in a PC of any sort.

    Though what kind of machine it actually is… I have no idea. Wikipedia says that PCI was developed for PCs, and you already excluded Power Macs which used it shortly after.

  9. Joshua Rodd says:

    This looks like an Elsa, which was a German PC manufacturer known for selling OS/2 preloads and for making boards with good OS/2 driver support.

  10. Julien Oster says:

    Oooh. Is it still somewhat closely associated with Macs, but more on a historical OS level? 8)

  11. Julien Oster says:

    Well I’m just gonna say it: is it for a NeXT machine?

    I found some references to Number 9 cards with a S3 928P and Bt495 combination, but you said “well known computer company”, so I guess that detail is wrong. So maybe it was made by NeXT themselves.

  12. Michal Necasek says:

    No, not NeXT… a much bigger company. Power Mac was the right direction, but not the Mac part 🙂

  13. Julien Oster says:

    Yeah, I just realized that I am not actually sure that there was a NeXT machine with PCI, might have been too early.

  14. Michal Necasek says:

    To answer my own question… probably the Matrox MGA-II chip, and the Matrox MGA Ultima may have been the first PCI graphics card for PCs. I couldn’t pin a date on it but it was advertised in the Dec ’93 of PC Magazine, and probably available either in late ’93 or early ’94. It’s mentioned as a new arrival in the Feb ’94 issue of PC Magazine.

  15. Michal Necasek says:

    The S3 chip was an 8514/A clone, but a weird one. It was architecturally extremely similar and had a very similar register interface, but it wasn’t compatible with any software written for 8514/A. I wrote about that a while ago in http://www.os2museum.com/wp/wp-admin/post.php?post=1859.

    And yes, this card was indeed used in an early IBM PCI machine.

  16. Michal Necasek says:

    You’re right about the missing chip, though it depends on how you define a “PC”.

    As usual, Wikipedia kind of lies… PCI was developed “on PCs” more than “for PCs”. In fact PCI was designed to be independent of the CPU architecture and was very quickly adopted by Alphas, PowerPCs, and all sorts of other machines. In that, it was unlike ISA or VL-Bus, but like Microchannel (which was used with several non-x86 architectures).

  17. Darkstar says:

    Okay, so the chip that’s missing is probably the one on the top left, which looks (from the layout) like it’s a BIOS ROM chip.
    So the VGA card has no BIOS, which means that these functions probably exist within the system firmware.
    The only machines which I know that work without BIOS chips on the VGA card are workstations. Since you already confirmed that it’s for an IBM system I would make my second guess a second-generation IBM RS/6000, which has a PowerPC CPU and thus no need for an (x86)BIOS chip, and it also nicely fits with your hint about the “Power” part. I can’t name an explicit model though since I only have one of these machines and that uses a different card.

  18. Michal Necasek says:

    Correct, there’s no BIOS chip because the card was supported in firmware.

    It was indeed intended for early IBM PowerPC-based systems. See what you can find out — very few of those systems used an S3 928. The later S3 864 was much more common.

  19. Julien Oster says:

    It’s difficult to make out the full part number, but it seems to be either 8183177 or 8185177. But I can only find references to S3 928 adapters with 8185109 (1MB) or 8185006 (2MB).

    At this point I’m just guessing, having almost no experience with PPC machines, but apart from the already mentioned RS/6000, which apparently did use S3 928 video adapters ( https://lists.debian.org/debian-user-de/2001/03/msg00534.html ), there’s also those: http://ps-2.kev009.com/rsinfo/powerpc.htm

    I don’t think it’s part of your ThinkPad. 😉

    Fun stuff, though: According to this manual, some Motorola Firmware for PowerPC-based machines has 486 real-mode emulation, specifically for the Video BIOS: http://www.ing.iac.es/~docs/external/vme/Moto-firmware.pdf (Page 1-2)

  20. Michal Necasek says:

    Yes, it’s the adapter with IBM part number 8185109 (1MB) 🙂 There are two IBM’ish looking numbers on the board itself, 8185177 and 8185280, not sure what those are.

    This particular card was found in an IBM Power Series 440 — see http://www.elhvb.com/mobokive/eprm/eprm1/f2589.htm . The Power Series 440 hardware was essentially identical to RS/6000 Model 7020 (40P). As I mentioned, the S3 928p is a very early PCI graphics card (S3 essentially tweaked an existing VLB design) and that’s why IBM used in in the original PowerPC systems. As soon as it was available, IBM switched to the much better S3 864.

    It is likely that IBM reused an existing board design, and that’s why there seems to be room for a BIOS chip. I suspect the board was manufactured by IBM, though I’m not certain about that.

    The IBM firmware does not include an x86 emulator (there’s just support for a few specific models of graphics cards), but the Microsoft ARC firmware does. So Windows NT for PowerPC can use graphics cards that the firmware of the machine can’t do anything with.

  21. Stiletto says:

    Good job, Julien 🙂

  22. ender says:

    Just curious: is it possible to get lspci -vv output for this card?

  23. Michal Necasek says:

    Yes, but not for a while. I’m away until late Feb.

  24. Michal Necasek says:

    Yes, but not sooner than about 3 weeks from now. And even then, it might take me a while. If you just needed the PCI ID, that would go a lot faster since the NT ARC disk shows that.

  25. ender says:

    No rush, I’m just curious what the output will be (also, use -vvnn instead of just -vv when running lspci).

  26. Michal Necasek says:

    Took me much longer than expected, but I finally updated the post with lspci output.

Leave a Reply

Your email address will not be published. Required fields are marked *