Interesting. Never heard of either S-MOS or PIX o.0.

Not much on interwebs either, quick google only came up with:
http://www.informedusa.com/t/pix3daccelerator11.20.html
https://www.edn.com/s-mos-systems-pix-3d-rend … softs-direct3d/
http://www.micahfreedman.com/projects/animation/smos.php

Seems S-MOS were working with Apple (Rave) and possibl MS at the time, but doesn't seem like it left the deveopment stage. Also interesting is that advert you posted in PCGamer is clearly for a development board! which I would have thought is somewhat strange to be advertising in an 'end-users' publication. Stuff like that was usually advertised through other channels.

Looks a interesting design, similar perhaps to PowerVR? If I'm reading that correct would have perfomed the polygon setup (usually done on CPU around about that time I think) and would utilise the RAM. Can't see any info about rasterisation though other than mentioning the 66million tris. I assume this was pumped directly to the QuickDraw/DirectDraw frame buffers, similar to how the PCX operated?

Good stuff.

The only number I see is 66 MPixels.
Since the Pix finishes with rasterization, the pixel color writes have to be calculated and written by the host and then dumped to the video card's frame buffer.
If I guessed right, it would be quite a unique partial 3d accelerator solution.

Where might this put it performance wise with its peers?

If the quoted fillrate is sustainable even with many vertex parameters, in what it does it would be up there with best consumer chips of 1996.

https://archive.org/details/Pcmania45/pcmania … ge/n29/mode/2up

"Finalmente, PIX, de S-MOS, es un nuevo API compatible con Direct 3D, BRender y Renderware, que puede trabajar en una tarjeta normal. Basta con insertar una placa especial en un slot PCI que contiene el chip SPC1515. Esto permite afiadir aceleracion 3D a cualquier tarjeta grafica"

"Finally, PIX, from S-MOS, is a new API compatible with Direct 3D, BRender and Renderware, which can work on a normal card. Simply insert a special board into a PCI slot that contains the SPC1515 chip. This allows adding 3D acceleration to any graphics card"

There's the part number, shown by Chadti99 in the original post: S-MOS SPC1515.

[EDIT] More:
https://pc.watch.impress.co.jp/docs/article/9 … 001/epson3d.htm
https://www.electronicproducts.com/graphics-c … -in-transition/
https://www.osp.ru/pcworld/1997/05/157452
https://vintageapple.org/macuser/pdf/MacUser_ … 05_May_1997.pdf 32-33 (pg 28-29)

Looks like the "Microtech 3D Xplosion" for Apple Macintoshes got out to at least one MacUser magazine reviewer.

An earlier one in the family, pre-PIX, seems to have been called the S-MOS SPC1500.
https://www.edn.com/edn-access-12-22-94-graph … ring-3-d-to-pc/
https://stason.org/TULARC/pc/3d-graphics-card … _Processor.html

Very interesting! I'm also amazed that the card had a website printed on it back in 1996. If I saw one of those I would have thought for SURE it was a modern made retro-compatibility card. It really looks like something modern and simple.

We need more information on this!

So you guys might not believe this, not sure I do 🤣, I managed to find one of these. Now to test and try and track down any demos or drivers. It’s gonna be a long shot but I’m going to contact anyone I can track down involved.

Wow, awesome ~

That's pretty crazy! Hope you have luck finding software.

Cool that someone actually has gotten their hands on one . This is a piece of history .

If this thing uses system memory for textures and even if it can do DMA to push rendered frames directly into a PCI or AGP video card's frame buffer (first linked press release says it uses system memory for that too, so presumably buffers there and then copies to VGA framebuffer), I can't imagine performance being that great .

I mean this is the time of 33 MHz 32-bit PCI able to do 133MHz MB/s in burst mode (not taking into account overhead) in half duplex and system memory being 66MHz SDRAM at 64 bits wide which does about 533 MB/second (not counting overhead). All of this would also be subject to North bridge <-->South bridge limitations .

I am no expert and I may well be completely off the mark on this, but the concepts sound like a bandwidth limited disaster on multiple fronts . I'm guessing it is probably no coincidence that this thing, in the portrayed implementation at least, probably had issues that made it disappear into the bowels of history.

The PCX1 and PCX2, for example, pulled it off without a dedicated frame buffer, but they had on-board texture memory.

That being said, I would be extremely curious the see what it could (can?) actually do . Hopefully, at least some software has survive somewhere .

EDIT : https://www.computer.org/publications/tech-ne … -rssi-1993-2000
This says that 100MB/second over PCI was effectively attainable . I would guess that was highly chipset PCI implementation dependent.

EDIT2: Corrected glaring typo (MHz which should have been MB/s)

133MB/s for PCI at 33MHz.

Cheers,

That's a very interesting card and looks even simpler. I wonder how it actually performs as it lacks dedicated memory and whether or not it has some kind of caching.

Wish you good luck!

Seems a joint development with Tseng Labs (ET6000) had been planned

https://books.google.co.uk/books?id=Pj0EAAAAM … 2%20pix&f=false

pentiumspeed wrote on 2022-04-26, 22:53:

133MB/s for PCI at 33MHz.

Cheers,

Theoretical. Without any overhead.

PC Hoarder Patrol wrote on 2022-04-27, 12:52:

Seems a joint development with Tseng Labs (ET6000) had been planned

https://books.google.co.uk/books?id=Pj0EAAAAM … 2%20pix&f=false

That would lead to a card similar to Apocalypse 5D- not helping the Pix 3D in other way but saving some PCI communication.

Yep, I can’t imagine performance being all that great. They probably took a look at the competition, namely 3dfx, and decided to can the project.

If it weren't for the handful of articles that are clearly real, and the fact it's mentioned on an Internet-Archived page from SMOS in 1996, I would almost think this card was a hoax.
The ad looks borderline fake and the font used on the silkscreen looks very much like the font KiCAD uses by default.

Chadti99 wrote on 2022-04-27, 15:20:

They probably took a look at the competition, namely 3dfx, and decided to can the project.

Considering just how many crappy 3d accelerators were in the market both before and after the Voodoo it seems difficult to believe a company would develop a 3d chip and give it support for all these APIs and choose not to release it. I guess it has to be true though, the PIX disappears from S-MOS's website between the April 1997 archival and the May 1997 archival.
See under Standard Products, "PIX 3D Graphics Rendering Engine" vanishes between archives.
https://web.archive.org/web/19970412152710/ht … ww.smos.com:80/
https://web.archive.org/web/19970524161434/ht … ww.smos.com:80/
Unfortunately they didn't archive any of the potentially informative links from the PIX home page itself.
https://web.archive.org/web/19970412152848/ht … /pix/index.html
The PCX was an okay 3d accelerator but it wasn't all that common, if this were cheaper (no memory chip, probably cheaper?) it could have found a niche, even if it was just in OEM systems where the onboard 2d was some cirrus chip with no 3d.

Error 0x7CF wrote on 2022-04-28, 00:45:

Considering just how many crappy 3d accelerators were in the market both before and after the Voodoo it seems difficult to believe a company would develop a 3d chip and give it support for all these APIs and choose not to release it.

Not surprising considering that 3Dfx got a head start having engineers with SGI expertise.

S-MOS was deep into other areas and the GPU was just a "test the waters in this hot market" effort that didn't pay off in the end. Epson soon gobbled up the company in the years after the GPU was developed and the GPU idea was discarded entirely, to focus on building components for portable electronics and consoles.