Have You Seen These Cards?

While cataloging my junk pile, I came across two graphics cards that I could not identify. Both are high-end ISA graphics accelerators from the early 1990s and both utilize Texas Instruments GPUs (long before the term “GPU” was first used, but very appropriate).

The first card was clearly made by SPEA (that is, SPEA Software AG), a German company known for its high-end graphics card. Based on the chip markings, the card must have been manufactured in the second half of 1991.

SPEA accelerator, component side

The DAC is a Texas Instruments TLC34075, presumably a 135MHz part (very fast for 1991!) and marked “prototype”. The accelerator chip is a 32MHz TMS34020, the newer TIGA GPU. The card uses VRAM, and does not appear to have any VGA compatibility or a BIOS. It has one 15-pin connector, presumably analog VGA, and one 9-pin connector, possibly a digital monitor connector.

SPEA accelerator, reverse side

I have not been able to determine what this card is, or what its specifications are beyond what’s obvious by looking at the chips.

The other board is even more of a mystery. It’s slightly older, probably manufactured in early 1991. There are no markings identifying the manufacturer. The GPU is the original TMS34010, the DAC is a 125MHz Brooktree Bt458, and there’s a Bt432 clock chip. Nothing too unusual there.


What’s definitely unusual is the use of a Xilinx XC3020 FPGA. The Vermont Microsystems, Inc. VMI VCAD chip suggests that the board may have been made by Vermont Microsostems, a high-end CAD hardware and software specialist.

Again there’s no apparent VGA compatibility. There are two 15-pin connectors, presumably analog VGA.


The card might be a prototype, for several reasons: The VRAM is very sparsely populated (5 chips out of 25), there’s extensive manual patching, and there are several unusual connectors on the board which could be used for debugging and testing. The FPGA utilization is unusual, but doesn’t necessarily mean it’s not a production board.

If anyone can identify either of these vintage graphics accelerators, please leave a note!

Update: The top card has been identified as one of SPEA HiLite accelerators, most likely an early SPEA Graphiti HiLite given the age of the card. The bottom card is certainly a Vermont Microsystems product, but the exact model is unknown. Likely a X/Series, but possibly a Cobra or some other model.

This entry was posted in Graphics, PC hardware. Bookmark the permalink.

38 Responses to Have You Seen These Cards?

  1. Darkstar says:

    The first one is probably a SPEA Grafiti Hilite:
    (nope, I don’t know anything more about it, just did some creative Googling 😉

    as of yet, I have no idea of the second one… but I like a good challenge 😉

  2. Darkstar says:

    The other one *could* be a descendant or pre-production version of the SPEA Graphiti Gallery HE-P35 here: http://www.vgamuseum.info/index.php/palcals-collection/63-professional-cards/500-spea-graphiti-gallery-he-p35
    Note the same memory type (SIL modules), same BT chip, although the HE-P35 has no FPGA.

    You could try asking the guys at vgamuseum.info, maybe they know something. Can’t you plug it in and try and dump the BIOS ROM from it? That might give some clues

  3. Michal Necasek says:

    The problem is that I could not find any mention of such card anywhere else. InfoWorld, drivers, release notes, etc. That makes me very skeptical. There is also a pic of this card on Wikipedia, but with no description.

  4. Michal Necasek says:

    The Bt DACs and SIL VRAM chips were very common in high end cards. This card can’t be it, it’s far too old (1986 vs. 1991). I’m not sure my card has a BIOS but I will check again.

  5. Darkstar says:

    Well, they had (have?) a trademark and a logo (see http://trade.mar.cx/spea_software) so it’s probably real. But finding drivers for such an old beast will be almost impossible 🙁

  6. Michal Necasek says:

    Sorry, I wasn’t clear. SPEA was a well known company, but “Grafiti HiLite” is what I couldn’t find any mention of.

    And I can’t find any BIOS on either board, unfortunately.

  7. Raijinzrael says:

    Them don’t have Video BIOS because them weren’t designed to work as primary display adapters, but as secondary monitor render viewports driven by special display drivers tailored for professional CAD software packages.

    A very common configuration was to use the primary video adapter to display the program interface and enter commands for the CAD application, while the TIGA accelerator, connected to a secondary high resolution monitor, renders and displays the result of the operation.

    Was common that the cards where supplied with drivers for the most popular CAD/CAM software packages, like HEIDI drivers for Autocad and 3D Studio.

    Sorry for the double post.

  8. Michal Necasek says:

    Yes, they’re TIGA boards (well, presumably, since they have the appropriate TI 340×0 processors), and yes, they don’t have a BIOS because they’re not meant to be EGA/VGA compatible or provide BIOS level support. That’s not the question. The question is, what exactly are those cards, especially the one with the Vermont Microsystems chip?

  9. Calvin says:

    I recall seeing information on NuBus graphics cards for Mac with very similar TI TMS GPUs – their instructions supposedly map to QuickDraw very well.

    I tried to find more on 68kMLA, where I remembered seeing some info on them, but phpBB’s search sucks.

  10. Calvin says:

    (and many NuBus graphics cards do have Bt DACs AFAIK – think it was a Radius card that uses a Bt+TMS, but don’t quote me on this)

  11. Michal Necasek says:

    Brooktree was simply one of the handful of companies who had high-performance DACs (Brooktree also supplied the workstation market). Looking at my junk pile, nearly all of the higher-end graphics cards from the early 1990s used Brooktree DACs. The other companies I can think of were Inmos (their DAC design was used in the IBM VGA), Music Semiconductor (found in old ATIs), and Texas Instruments.

    I don’t know about the TMS340x0 GPUs and Macs, but I’m not surprised that they were used there as well. The TMS340x0 was essentially a RISC processor with special graphics support, very generic, fully programmable, and therefore highly flexible. But in the end it couldn’t compete with the 8514/A and its clones (S3, Mach8) on either price or performance.

    It was typical that TIGA boards used two completely different types of memory, the framebuffer VRAM and standard DRAM used for the TMS GPU’s RAM. I think some boards had up to 16MB RAM for the TMS chip, an insane amount back in 1990 or so. After 25 years, we’re now back to the original idea of a fully programmable GPU 🙂

  12. Erik says:

    The reason you’re not finding any references to “Grafiti Hilite” is because the cards are called “Graphiti Hilite”. There’s plenty of more or less vague references to and pictures of these boards on the web. The top card is almost certainly a prototype version of a Graphiti Hilite card. Your card being a prototype (as the writing on the TLC34075 chip indicates) could explain a few differences to the production cards. In particular, the area in the very top right of the card (where the letters “EOLK” appear on your specimen) looks different.

    It seems from various anecdotes around the web that these SPEA Graphiti Hilite cards, mostly popular in Europe and especially Germany, were produced from 1992-94. The first Usenet mention on Google Groups is from 1993, from someone trying to sell his half-year old card.

  13. Erik says:

    BTW, there should be a Windows 3.1 driver for the Graphiti Hilite floating around. The file is called BWHIL245.EXE, is 266351 bytes and dated 18-08-94. If you could get that to install and recognize the card that would be pretty good confirmation that the card is what it appears to be 🙂

  14. Erik says:

    Apologies for the triple posts… But the second card is indeed likely to be made by Vermont Microsystems, Inc. I found this Usenet post from 1996 referencing a VMI card with a Xylinx chip:

    My Vermont Microsystems video card uses a RAM-based Xylinx part. Everytime you boot the computer you download the firmware to the card; the idea being they could update it to add new functionality or fix bugs.

  15. Michal Necasek says:

    Given how unusual FPGAs are, it’s pretty much certain it’s a Vermont Microsystems card. Now I wonder what the model was, or if it was a released product at all. I might have some MCA board from VMI around, will have to check…

  16. Michal Necasek says:

    Nice detective work! Right now I don’t have a system available, but I’m hoping that will change in a week or two. Then I’ll try the driver and see if it does anything. Without a driver, I can’t tell if the card works at all.

  17. Stiletto says:


    Another difference between yours and the VGAMuseum cards is that yours is marked P41/R11 on the reverse, and theirs are marked P41/R12.

    It’s always possible it’s a slightly different model SPEA card?

    Looking into your Vermont Microsystems card, without a doubt it’s made by them. There’s printing on the PCB directly above some of the … RAM? chips but I can’t make it out.

  18. Stiletto says:

    Okay, so having read some ancient press releases:

    I believe what you have there is a Vermont Microsystems Inc. (VMI) X/Series, possibly the X/Series 2-D.

    The press releases imply that the VCAD chip was created in 1989 and only the X/Series models shipped with it. However, the Brooktree chip on the card dates 1991, so I can’t be sure.

    Here’s drivers for Windows 3.0 for the X/Series.

    This guy can maybe tell you more.

    Of course, I could be wrong.

    Getting that printing in the top left (after “Made in the USA”) might be a hint, can’t make it out…

  19. Michal Necasek says:

    Looking at the images, the board is definitely a SPEA Graphiti HiLite, but the real question is “what exactly is/was SPEA Graphiti HiLite” 🙂 How to tell the exact model I have no idea. The number of memory chips might matter, or the DAC, or who knows what.

  20. Michal Necasek says:

    Not much of a hint, there’s just MADE IN U.S.A. 1091-07 01. I’m guessing 1091 is a date code. There’s additional marking which reads T0100BL-C.

    It now occurs to me that all those funny connectors could be intended to hold a daughter card with additional memory or possibly even additional processing chips.

    The hints that VMI in fact designed the IBM PGA are intriguing…

  21. Darkstar says:

    Okay, so I spelled “Graphiti” wrong, but my link was correct 😉

    Still no drivers though, but I’ll keep looking. Please continue to post unknown hardware on your blog, it’s interesting, entertaining and addictive 🙂

  22. A. Kohl says:

    The Spea cards were mainly used for CAD applications in Germany – I will check for drivers in my archive.

  23. Stiletto says:

    Liberated bwhil245.exe:
    SPEA BigWin Driver v2.45 for Windows 3.1x for SPEA Graphiti HiLite Series (rel.0894)

    The last paragraph in this blog assisted me.

  24. Michal Necasek says:

    Sweet. I’m glad I never needed to use Driver Guide…

  25. GL1zdA says:

    The first one is SPEA HiLite 1024. Source: c’t 11/1992 – get yourself the Blu-ray archive: http://shop.heise.de/zeitschriften/ct/ctrom – it has a good search engine and will give you information on many things that were never put on the Internet. One of my best purchases. In Deutsch of course.

  26. Michal Necasek says:

    Interesting… so is it a Graphiti HiLite or HiLite 1024? And does c’t say anything about the specs, or when it started selling? If it was really first mentioned in Nov ’92 then my board could really be a prototype or a very early revision.

  27. GL1zdA says:

    The 11/92 article is a comparison of several Windows accelerators, including TIGAs.

    There were several HiLites. The first one is TI 34020, 2 MB VRAM, 1 MB RAM for programs, can do 1280×1024 with 256 colors. Drivers for: CAD, 3D-Studio, Animator Pro, Windows can make use of it through the standard TIGA driver or the custom SPEA BigWin driver. Price: 4435 DM. They call it Graphite HiLite here when it was announced in c’t 12/1991.

    There’s was a cheaper version with only 1 MB VRAM, up to 1158×870 with 16 colors announced in 9/1992, price: 2839 DM. This is the HiLite 1024 – in 11/1992 they state the maximum resolution is 1024×768. They have a photo of it and it looks exactly like the one you have.

  28. Michal Necasek says:

    Thanks for the info. I ordered the c’t archive, but it’ll take a few days to get here.

    Let’s see… the markings are very hard to read (the chips are slanted the wrong way), but my card has two sets of memory, both in ZIP (not SIP!) form factor. There are eight smaller TMS44C256-80SD chips, that ought to be 1MB DRAM, presumably for the GPU. Then there are sixteen larger TMS44C251-10SD chips, which is dual-ported VRAM. At 16x256Kx4 bits that’s 2MB.

    So my board should in fact be a Graphite HiLite, capable of 1280x1024x8bpp. Nice! And Nov-Dec ’91 is about when my board appears to have been made, so it ought to be one of the early units. Heck, I think I got my first graphics card with 2MB video memory about 5 years later…

  29. Michal Necasek says:

    Oh, and just to clarify… It’s a “Graphiti HiLite”, not “Graphite HiLite”, at least according to the driver package (bwhil245).

  30. Raijinzrael says:

    Nope, this card isn’t a Graphiti, since those ones never came with the VCAD chip, as you can see in the photos. This card is member of the X/Series, and, looking in the driver package, seems to be more or less compatible with the Cobra series.


    With this driver you should be able to use it in Windows 3.0/3.1… Looks like Win3.1 is the last windows version supported, as Windows 95 include drivers only for the most updated Cobra cards. There is also a product that can make this card work with AutoCAD, called AutoMATE.

    Surprisingly, CADKey could drive this card directly:

  31. Michal Necasek says:

    I don’t think anyone suggested that the card with the VCAD chip is a HiLite? The other card (which is unquestionably made by SPEA) is…

    Hmm, AutoMate was VMI’s own package, wasn’t it? There was even a lawsuit related to it (VMI vs. Autodesk).

  32. Raijinzrael says:

    AutoMate is like SciTech Display Doctor but for AutoCAD, a set of commercial ADI drivers made by VMI with support for graphic accelerators… The lawsuit was because AutoDesk stole code from VMI to make their own set of display driver that could be bundled with AutoCAD.

    Probably your X/Series card came with a licensed copy of AutoMate, so it could be used by AutoMate enabled Apps.

  33. Stiletto says:

    Raijinzrael, way ahead of you with the second card. The only remaining question in my mind is “which X/Series”? Apparently there were several different models (see: “X/Series” “Vermont Microsystems” Google Book search results)

    My money is on the entry-level card…

  34. Hi,
    I designed the initial version of firmware running on that TI34010 Vermont Microsystems card, it was called the Cobra. We built and launched the very first TI34010 card available (in a footrace between all the CAD card vendors, who had been given the TI chip at the same time) sometime in 1987. This was a revolutionary time, the TI chip had enough power for us to put the highest resolution logic onto a single card and the vector draw rates were about 50 X faster.
    SPEA was a major competitor from Starnberg near Munich, later sucked up by Diamond. At one point, SPEA and VMI had talks of acquisition. I have several of these in my garage with a bunch of other stuff, I am certain. Prior to this card, VMI had produced the IBM High resolution graphics cards on an OEM contract. VMI took these and made their own line of cards, all two-card sets until this TI chip allowed us to fit on one card. There was a very hot market for high performance Graphics cards for CAD and scientific use between 1985 and 1990. The introduction of the VGA, with its’ extremely low price point, combined with much faster main CPUs took the wind out of that market between 1990 and 1995 and many of these companies went belly up: VMI, Number Nine, Control Systems, SPEA, many others disappeared or were acquired. Around 1995, users again started to exceed the capability of the main board architectures when gaming became popular, fueling another market demand for high performance graphics cards which lasts to this day. VMI had top engineers: many of these people went to SGI and other leading vendors of the day and eventually became leaders at Nvidia today.

  35. Michal Necasek says:

    Very cool. The TI34010 is historically very interesting in that architecturally it’s closer to the modern GPUs, but it was too far ahead of its time and couldn’t compete against the much simpler but adequate and very cheap fixed-function 2D accelerators. Technology often moves in circles (or on an upwards spiral, we like to believe).

    I did a bit of work in graphics a while ago and remember the number of vendors going from dozens down to three. It does make things simpler 🙂

  36. @David Ressler –

    I’ve got an IM1024 (pre-TI34010) and would be over the moon to have drivers, diags, and/or docs for it. Not sure you’ll see this, but on the off chance you do, I’d love it if you could get in touch with me (possibly through Michal, if the comment section doesn’t leave you any breadcrumbs there… he has my email address).

  37. Andreas Kohl says:

    The latest driver and firmware for the Spea HiLite is from August 1994. I have here one of the later models from late 1993. Spea’s SuperDisk 1 includes an installation program which should also show the correct model. All the drivers and tools from Spea together should be 15 MB. Digital monitors are not supported – only different multifrequency and some fixed frequency CRT from Spea, which are Sony or Hitachi OEM models.

Leave a Reply

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

This site uses Akismet to reduce spam. Learn how your comment data is processed.