Sometimes identifying old hardware is a bit tricky. Consider these two graphics cards:
The PCB is the same, the BIOS chips look the same, the DAC is slightly different but an 80 MHz part in both cases, memory is the same… but wait—the graphics controller is different! The top specimen is equipped with the original S3 86C911 accelerator, while the bottom one uses the newer and improved S3 86C924 with added 24-bit true color acceleration.
What the card really is and who made it is an interesting question. The one identifying mark is “P/N 2150D”, which according to TH99 means this is a S3-911 based graphics card by ESCOM AG (clearly not quite right in the second case!). Given that ESCOM was a German system builder who generally did not design PC components, it’s rather likely that ESCOM sold the card but it was made by some other company.
These graphics cards were both made in the second half of 1992 and are good examples of early S3-based PC graphics accelerators. Both use an 80 MHz DAC and 1 MB VRAM (not DRAM). And they show that more or less no-name (probably Taiwanese) OEMs could build relatively high-end graphics cards based on the S3 chips back in 1992.
The VGA function on these cards is quite slow, in fact noticeably slower than even cheap basic SVGA cards such as Acumos/Cirrus Logic AVGA2/CL-GD5402 from the same era. The accelerator presumably did rather better—after all, S3’s line of graphics accelerators which started with the 911 was quite successful.
For the interested, here are close-ups of the two chips:
The S3 924 was more or less a superset of the 911, which is why it was obviously easy to take an existing 911-based design and upgrade it to the 924.
If someone knows how to decipher a date code from the 911 chip, please leave a comment. It would be interesting to know if the 911 is actually older than the 924 or not.
Hi, is it possible to send me a copy privately or to some other site of your 924 bios? I’m working on a emulator alongside a friend of mine and we’ve already got the 911 one, but we lack the 924.
It is possible. But it will take a while because I’m traveling right now, and once I’m back, I need to find the card which might not be trivial (still recovering from last year’s move).
No luck so far… I found the 911 card, but the 924 wasn’t in the same box where I expected it to be. So I have no idea where the 924 is, and it will surface eventually but it could take a while.
Okay, that took a long time, but the S3 924 VGA BIOS is now here: http://www.os2museum.com/files/S3924AMI.BIN
Note that the card has two 32KB BIOS chips. They are both ST M27C256B EPROMs, and they have identical contents. Also note that there are two separate ROMs inside the BIOS image.
I have the EISA version of the 923; it’s P/N 2151. Horribly slow in DOS; half that of a mach32 EISA. I’d love to know what the jumpers and switches are for..
One bios chip has a “BRICK” sticker. It was a Norwegian OEM brand in the 90s.
Not to necropost but how does one deal with the 24-bit acceleration of the 924 (also used with the 80x), at least on Windows 3.1? From my logs, it seems that the s3_24.drv driver (80x/924 true color driver), in bitblt (cmd 6), it does a 4-phase to bitblt and swaps the x/y coordinates from src to dst and viceversa, but I still get glitches while moving windows or writing on Write (Win3.1). Any clue or some in-depth information?
I fear that I don’t have any in-depth knowledge beyond what is in the datasheets. Those chips had no 24bpp acceleration, right? So everything had to be done with 8bpp operations?
You might potentially find something useful in the OS/2 S3 driver because the source code was published as part of the OS/2 DDK, although I don’t know exactly which chips it supports (it should be 801, 805, likely 924). The driver operation was also reasonably well explained in the DDK documentation.
Actually the old NT DDKs (3.1/3.5/3.51) also include S3 drivers. That could be easier to dig into than the Windows 3.1 driver for which no source code probably exists (does it?).