As a follow-up, I bought several ViRGE cards in the past few months and recently had time to benchmark them.

They were all tested in a Pentium 233Mhz MMX system with 32MB of RAM. All cards have 4mb of memory. Results are as follows:

Diamond Stealth 3D 2000 (ViRGE, v1.04)
3dbench: 164.5
Quake I TimeDemo: 23.5fps
vspeed BVGA: 22.12
vspeed LFB: 19.17

Diamond Stealth 3D 2000 Pro Rev. A (ViRGE/DX, v1.01)
3dbench: 164.1
Quake I TimeDemo: 24.4fps
vspeed BVGA: 29.86
vspeed LFB: 63.77

Diamond Stealth 3D 2000 Pro Rev. G (ViRGE/DX, v1.14)
3dbench: 164.1
Quake I TimeDemo: 24.5fps
vspeed BVGA: 29.86
vspeed LFB: 63.77

STB Nitro 3D/GX (ViRGE/GX, v1.3)
3dbench: 164.6
Quake I TimeDemo: 24.6fps
vspeed BVGA: 30.60
vspeed LFB: 19.54

The only notable differences were among the vspeed scores, where the DX cards scored surprisingly high on the linear framebuffer memory speed tests. Not sure if that's a mistake, as the GX card scored about the same as the original ViRGE on the same tests.

SiliconClassics wrote:

The only notable differences were among the vspeed scores, where the DX cards scored surprisingly high on the linear framebuffer memory speed tests. Not sure if that's a mistake, as the GX card scored about the same as the original ViRGE on the same tests.

I would think that the difference may lie in the cards' BIOS. The LFB vspeed is quite plausible. Did you optimize the access for all cards with S3VBE, S3SPDUP and FastVid (or equivalent for you cpu)?

Of course you could also try MCLK but that would not explain a factor of three in the results. And my experience with my Nitro 3D is that is seems to reset the clocks on a mode change so it effectively blocks overclocking. It is clocked rather high to begin with so this shouldn't matter.

I didn't use any of those utilities, just booted straight to a clean DOS prompt (bypassed autoexec.bat and config.sys) and ran vspeed.

I'm surprised that LFB is so much faster on the DX cards than on the GX, as the GX was supposed to be a slightly faster chip. Perhaps it's due to the type of RAM on each card - is there any way to discern the kind of memory?

Come to think of it STB made a /GX card with EDO RAM, better check if they clock it anywhere near SDRAM cards.

Putas wrote:

Come to think of it STB made a /GX card with EDO RAM, better check if they clock it anywhere near SDRAM cards.

previous page. Clock speed is 75mhz.

Some discouraging news: I just tried to use the Stealth 3D 2000 Pro (Rev.G, ViRGE/DX) with my Forte VFX1 VR headgear and the Stealth's VESA feature connector does not seem to be passing a video signal to the headset. It works fine with my old Stealth 3D 2000 card, so there must be some difference in the new hardware.

Apparently, VESA feature connector pinouts were not completely standardized, and the ViRGE/DX card may use a scheme that differs from the ViRGE. Unfortunately, the images on that web site don't load. If they did, I could probably discern the pinout of the DX card and rig up some kind of adapter.

Any suggestions or am I stuck using the older ViRGE card?

I highly doubt they used a non-standard pin out.

Consult the product manual. I found that on some cards you have to enable the feature connector before it will output.

SiliconClassics wrote:

I didn't use any of those utilities, just booted straight to a clean DOS prompt (bypassed autoexec.bat and config.sys) and ran vspeed.

I'm surprised that LFB is so much faster on the DX cards than on the GX, as the GX was supposed to be a slightly faster chip. Perhaps it's due to the type of RAM on each card - is there any way to discern the kind of memory?

AFAIK DX and GX are identical except GX support EDO and SDRAM while DX only suppports EDO. The Nitro 3D uses a GX chip in combination with EDO (that 's what Putas and swaaye are talking about).
If you want to know exactly what RAM is used just google the inscription on the chips to get the data sheet. The Nitro 3D has quite fast chips (for a ViRGE) which is why it clocks so high.
Of course fast chips don't mean a high clock is used, it's just a precondition. You could check the memory timings/clocks with mclk. But I still think that the speed difference does not come from memory speed, I suspect the VGA bios of the Stealth cards does something different, maybe it already includes VBE 2.0? I'd really recommend to try the tools I mentioned to optimize performance and compare chips/ram and not BIOS code.

Anonymous Coward wrote:

I highly doubt they used a non-standard pin out.

Check the link in my previous post - the author has identified four different pinout schemes implemented among various video cards, and even identifies the ViRGE/DX as having its own peculiar pinout. Unfortunately the images depicting the pinouts don't load.

Anonymous Coward wrote:

Consult the product manual. I found that on some cards you have to enable the feature connector before it will output.

I don't have the manual, but I have tried using the FCON.EXE / FCON.COM utilities that enable the feature connector on the S3 Trio chipsets. They don't seem to work on the ViRGE.

Strange. Usually stuff that is "VESA" compliant follows a standard. The "S" in VESA means something after all. Perhaps just electrical standards in this case. Considering how uncommon it is for devices to use the feature connector, it would have been a pretty dumb move to non-standardize it.

If you can locate the pinouts again, you should be able to break apart the cable connector and re-crimp it.

Is it true that one is forced to have a ViRGE 325 in order to play specific S3D games that supposedly cannot be patched? Maybe he doesn't care about it but..

subhuman@xgtx wrote:

Is it true that one is forced to have a ViRGE 325 in order to play specific S3D games that supposedly cannot be patched? Maybe he doesn't care about it but..

No, the 325, DX & VX can play all S3D games.

The GX2 have issues with some games. Haven't tested the GX, MX and MX2.

To be more precise: There are DOS S3D games (at least Terminal Velocity) which don't work with anything except a 325 or VX out of the box.
If you use the S3D Toolkit Patcher (does not patch the executable, it is a launcher or TSR) you can use later chips as well. It seems there are different version of this tool, but even later versions which enable the use of later chips don't seem to support those properly so there may be graphical glitches.

My personal experience is that Terminal Velocity runs fine on a GX (STB Nitro 3D) using the Patcher.