Fantastic article, I like it very much. My first 3d card was a s3 virge.
Thank you very much.

Good write up. Since the Trio 3D is just renamed ViRGE, why not finish the story with it and how it fixed the blend mode problems/etc?

furan wrote on 2021-04-15, 00:30:

Good write up. Since the Trio 3D is just renamed ViRGE, why not finish the story with it and how it fixed the blend mode problems/etc?

It didn't?

Putas wrote on 2021-04-15, 02:41:

It didn't?

Here is a link to you saying it did:
Re: S3 Trio3D/2x - is this just a rebadged ViRGE or more?

Heh, a quick look at my screenshots, and I knew better in 2011. Many blending errors were fixed indeed.

Putas wrote on 2021-04-15, 18:08:

Heh, a quick look at my screenshots, and I knew better in 2011. Many blending errors were fixed indeed.

My memory does this too. I would really like to try one out, maybe see if I can patch S3D stuff to work with it.

My Toshiba Tecra 550CDT laptop has an integrated S3 Virge. In my experience games that worked correctly without distracting transparency glitches did generally run a bit better than software rendering, but not by too much and rarely can I afford to put the resolution higher than 240p. The forced dither pattern especially at low resolutions can also look pretty bad at times.

Though, there are games like Resident Evil 1 at 240p that run well, so that's cool. 480p in that game is playable but noticeably choppier.

Any idea where the name ViRGE comes from?

Virtual Reality Graphics Engine aka ViRGE

matze79 wrote on 2021-04-16, 07:23:

Virtual Reality Graphics Engine aka ViRGE

Ah thanks. Searching for that, apparently it's on the ViRGE Wikipedia page, I assumed it was something obscure.

S3 Virge is up to 2x-3x times faster than PMMX 200 when rendering 3D with identical settings (16-bit color, filtering, etc).

https://retro.swarm.cz/s3-virge-325-vx-dx-gx- … tors-deep-dive/

There is very little difference between ViRGE DX and GX chips. The 3D core is exactly the same. GX just has a better memory controller that allows board manufacturers using synchronous memory chips (SDRAM/SGRAM) in addition to EDO RAM. Sadly, due to a missing texture cache, ViRGE chips are very inefficient in using synchronous memory chips. SDRAM/SGRAM-based ViRGE GX have 15-25% performance hit when drawing textured triangles in comparison with EDO-based ViRGE GX running at the same clock.

Hello diagon_swarm.
I have found this comparsion SGRAM and EDO version, but it's Virge DX, not Virge GX. What is strange, SGRAM version totally win by large margin, and even on lower clocks.
Are you sure about this? Or... what do you say happened here?
Another interesting topic... they've used SGRAM modules and PCB even with DX chips. maybe they had excessive DX chips and PCBs in later stage. This seems as rarity. What is strange, seems SGRAM works like charm, higher perfomance on lower clocks. Even on Virge DX chip! Which shouldn't have even improved controller for SGRAM.

https://www.clous.cz/hercules-terminator-3d-dx/
https://www.clous.cz/s3-virge-dx-sgram/

SGRAM version (60/60 MHz)

Final Reality
Overal score: 2.74 RM
Robots: 9.3 FPS
City scene: 9.4 FPS

EDO RAM version (67/67 MHz)

Overal score: 2.18 RM
Robots: 6.9 FPS
City scene: 7.5 FPS

W.x. wrote on 2024-07-18, 19:56:

Hello diagon_swarm. I have found this comparsion SGRAM and EDO version, but it's Virge DX, not Virge GX. What is strange, SGRAM […]
Show full quote

https://retro.swarm.cz/s3-virge-325-vx-dx-gx- … tors-deep-dive/

There is very little difference between ViRGE DX and GX chips. The 3D core is exactly the same. GX just has a better memory controller that allows board manufacturers using synchronous memory chips (SDRAM/SGRAM) in addition to EDO RAM. Sadly, due to a missing texture cache, ViRGE chips are very inefficient in using synchronous memory chips. SDRAM/SGRAM-based ViRGE GX have 15-25% performance hit when drawing textured triangles in comparison with EDO-based ViRGE GX running at the same clock.

Hello diagon_swarm.
I have found this comparsion SGRAM and EDO version, but it's Virge DX, not Virge GX. What is strange, SGRAM version totally win by large margin, and even on lower clocks.
Are you sure about this? Or... what do you say happened here?
Another interesting topic... they've used SGRAM modules and PCB even with DX chips. maybe they had excessive DX chips and PCBs in later stage. This seems as rarity. What is strange, seems SGRAM works like charm, higher perfomance on lower clocks. Even on Virge DX chip! Which shouldn't have even improved controller for SGRAM.

https://www.clous.cz/hercules-terminator-3d-dx/
https://www.clous.cz/s3-virge-dx-sgram/

SGRAM version (60/60 MHz)

Final Reality
Overal score: 2.74 RM
Robots: 9.3 FPS
City scene: 9.4 FPS

EDO RAM version (67/67 MHz)

Overal score: 2.18 RM
Robots: 6.9 FPS
City scene: 7.5 FPS

It could be possible that the S3 ViRGE/DX is using 35ns (or something closer) EDO RAM chips in a different package. I do have a Compaq S3 ViRGE/GX with proper 10ns SGRAM chips clocked at 75 MHz, it performs quite slower.

W.x. wrote on 2024-07-18, 19:56:

I have found this comparsion SGRAM and EDO version, but it's Virge DX, not Virge GX.

There needs to be some other difference. The supposed SGRAM card is still using EDO, the only memory type supported by the DX. See also Re: S3 Virge DX and GX memory . Maybe the "EDO" card they used operates at the slow 2-clock burst cycles instead of the faster single-clock cycles.

EDIT: This post said "the only memory type supported by the GX", which should obviously have been "the only memory type supported by the DX". Fixed it.

mkarcher: Yes, I think the 1-cycle vs. 2-cycle EDO can be the answer for this.

Hello. I couldn't get there from july 2024.
Yes, after few days, I've found out, it is really EDO ram, but in different (32-bit) packaging.
The card overclocks like crazy, 80 mhz core and 80 mhz (EDO) memory, power strip didn't let me more with slider.
Card still ran Final Reality's Robots.

I've tried other Virges, and of course, they had problems to go over 60-65 Mhz, usually artefacted or freezed.

This EDO 32-bit package late Virge DX, can go to monster overclock, from 50/50 to 80/80, and maybe even more (Powerstrip didnt let me).

Sorry for causing confusion with SGRAM.
Even Clous have it incorrently on page, that those are SGRAM. Those are really EDO, even his card.

Interesting read, I had a S3 ViRGE 4mb in 1997/98 but I don’t remember specific model number.

Unfortunately I was part of the 3d-decelerator crowd, mostly because of my experiences with Descent2. The 3d accelerated version looked better but was not smooth at all. Ultimately I played using the software rendered version until I purchased a Rendition Verite 2100 a few months later.

My system was a Cyrix P-200 with 48mb ram which after reading your comments about ViRGE needing a fast CPU probably contributed to the issues I experienced.

W.x. wrote on 2024-09-01, 01:07:

This EDO 32-bit package late Virge DX, can go to monster overclock, from 50/50 to 80/80, and maybe even more (Powerstrip didnt let me).

Powerstrip has a configuration file in which you can adjust these limits.

Putas wrote on 2024-09-01, 09:47:

W.x. wrote on 2024-09-01, 01:07:

This EDO 32-bit package late Virge DX, can go to monster overclock, from 50/50 to 80/80, and maybe even more (Powerstrip didnt let me).

Powerstrip has a configuration file in which you can adjust these limits.

Thank you. I couldn't figure out, how to do it, so after 10 minutes, I've gave up. But I knew, I could unlock scale. But it was probably different version, where it could be set from options. I'll check configuration file next time, I will be doing some testing. But, didn't have many cards, that can go out of the scale to this day. 😀 I guess it will have limit at 82 or 84 mhz, it's not worth of retesting. 80/80 Virge DX is 60 % overclock!

Other than configuration file, you can also just double click on the Core or RAM icon under each slider and enter a new percentage. It's much easier then .ini editing.