Alternative to Voodoo 3 would be a stretch, there is a reason 3dfx won the early accelerator wars for sure. There are too many compromises with the Rendition card. While it supports VESA 2.0, it performs poorly in many vga benchmarks and notably Doom, although there is a newly modified executable that resolves that, FastDoom. It’s also a bit picky about bus mastering and dma transfers and this doesn’t work well on all motherboards. The Gigabyte GA486AM/S for example has issues with the AM5x86 in WB and DMA enabled on the Rendition. Without DMA, performance takes a severe hit as seen below.

If you want to check out the first 3D accelerated version of Quake, and push your socket 3 Quake experience a bit further, in DOS, it really is worth checking out. I think vQuake looks really nice, especially compared to software mode and even against GLQuake, particularly the lighting when compared to GLQuake. I’ve seen other tests that put the V2000 versions of the Rendition ahead of the original Voodoo in Direct3D and I’d like to test this myself. I think this card, like the Voodoo, is much more appropriate for socket 7 builds but I enjoy pushing socket 3.

While others have commented about poor signal quality I can’t say I’ve seen the same with my Diamond S220 Rev. A. It looks great in dos and Windows, GUI performance isn’t lacking but I’ve not tested it via any 2d benchmarks.

Here are some Quake scores I pulled together. I’ve put 320x240 in the mix because, for whatever reason, 320x200 won’t fill the screen on an LCD or a CRT. So IMO, most would play it in 320x240. I’ve also included 640x480 results to give a comparison against GLQuake. All tests were done as close to the dosbench standard as possible with no sound.

POD 100 L1 WB Gigabyte GA486AM/S 1024k L2

Dosbench Quake 320x200 19.2
Dosbench Quake vid _mode 11 320x200 21.6
Dosbench Quake vid _mode 12 320x240 18.9

Quake108 320x200 21.5
Quake108 vid_mode 11 320x200 24.1
Quake108 vid_mode 12 320x240 21.3
Quake108 vid_mode 640x480 7.0

VQuake nodma 320x200 29.0
VQuake nodma 320x240 26.0
VQuake nodma 640x480 16.2

VQuake dma 320x200 39.3
VQuake dma 320x240 35.9
VQuake dma 640x480 24.9

AM5x86-160 L1 WT Gigabyte GA486AM/S 1024k

VQuake dma 320x200 27.4
VQuake dma 320x240 25.3
VQuake dma 640x480 17.3

AM5x86-180 L1 WB Biostar UUD 256k L2

VQuake nodma 320x200 23.4
VQuake nodma 320x240 21
VQuake nodma 640x480 13.3

VQuake dma 320x200 30.4
VQuake dma 320x240 27.4
VQuake dma 640x480 19.4

These numbers are dismal.
The rest of the details you added only make it sound worse.
Thanks for spending the time on this.

voodoo 1 and 2 never did it for me.
when they came to market, games were not the main motivation factor for me anymore, so they had limited appeal.
At the time i salivated over pro-grade 3d accelerators. Financially that was a fantasy.
So when nvidia came up with Riva, TNT and especially later with GF and Q - the good base line capabilities combined with affordable price opened the door to a world that was out of reach until then.

In 1996/97 what I *really* wanted in a graphics card was high resolution at good refresh rates and true colour. I didn't understand the point of blowing hundreds on a dedicated 3D card to play games that I didn't like anyway. Plus the Creative 3D Blaster had already come and gone so 3D hardware seemed like a passing fad.

Pretty much my thoughts.
But to each their own.
Not judging.

@pshipkov, curious if you had tried a sound card on the Biostar 8433UUD at 60MHz FSB? I couldn’t run the FSB:PCI at 1:1 without distorted sound which puts it at a bit of a disadvantage against the LS486 boards.

Is that with specific sound card model, or any card ?
I don't remember hearing bad sounds from it so far with sb16 and awe64g.
But also, it is not like I did detailed sound profiling.

pshipkov wrote on 2021-11-23, 16:19:

Is that with specific sound card model, or any card ?
I don't remember hearing bad sounds from it so far with sb16 and awe64g.
But also, it is not like I did detailed sound profiling.

I tried an Awe64 and an ESS, both ISA cards. Windows startup sound was first I noticed.

Will check soon.

Btw, yesterday was kind of a bad day for retro hardware here.
Plugged Voodoo3 PCI in Biostar UUD board and now both are real quiet now.
Kids, don't do this at home.

Oh no, it’s the worst feeling when something like that happens, maybe it will come back.

The motherboard is back online. It was a failing RTC module - ran out of battery. Spooked me for good.
But feels like the Voodoo3 card is collateral damage.
With the prices of this junk out of control these days will have to try to repair it somehow.
First step will be compare voltages on components to working card - in case it is electrical issue.
If signal one - i don't know - no oscilloscope here, so chances will be slim.

Just checking, you’ve tried the Voodoo3 on another board? It won’t post on a UMC based socket 3 board to my knowledge.

Yes, i tried it on several other boards - 487 SIS, VIA SS7, 440BX, etc.
The reason i shoved it in the Biostar board was to see if it will light-up with POD100 CPU in the socket.

Does it have a bios chip that could be re-flashed? Not in front of mine at the moment. But maybe that’s all it needs.

It does, but is not socketed. I have a programmer for these so it will be checked as well. Also have sockets for this type of EEPROMS.

Btw, no sound problems on the UUD @ 180/200 MHz. At least with SB16 CT2230.
But i know that it was fine with other sound cards too.

pshipkov wrote on 2021-11-24, 02:58:

Btw, no sound problems on the UUD @ 180/200 MHz. At least with SB16 CT2230.
But i know that it was fine with other sound cards too.

Which version of UUD and can I get a screenshot of Bios Chipset settings?

Tried on the two boards here - versions 2.0 and 3.1.
2.0 has the 1024Kb L2 cache mod.

3.1 board runs with all settings on max except DRAM Read WS = 1 and IBC DEVSEL# Decoding = Medium.
2.0 board runs with all settings on max except CACHE WAIT STATES = 3-2-2-2, DRAM Read WS = 1, IBC DEVSEL# Decoding = Medium.

Screenshots in this post.
What OS you use ?

pshipkov wrote on 2021-11-20, 20:40:

But i provided you a 2.09 bios already. Did you forget, or i don't remember ? […]
Show full quote

But i provided you a 2.09 bios already.
Did you forget, or i don't remember ?

Not sure what is the difference between the two.
I see the X model showing up more frequently on PCI cards.
In functionality P and X appear identical.

Some relevant reading material for you:
Stealth64 DRAM T VLB Trio64 86c764-P vs. 86c764X
Re: Preferred UniVBE version for Diamond Stealth 64 DRAM T VLB (Trio64)

I need the BIOS for the -P card, not the X card. Using the X card's v2.09 BIOS on my -P card makes it so that I an no longer soft-reset. Hence, I am looking for v2.02, which is the latest BIOS sticker I have seen on the -P VLB cards. There may be a later one, but I haven't seen a photo of it.

pshipkov wrote on 2021-11-22, 04:10:

These numbers are dismal.
The rest of the details you added only make it sound worse.
Thanks for spending the time on this.

I suppose they aren't great when compared with a voodoo2 in GLQuake at 800x600, but 640x480 at 19.4 fps is certainly better than software mode. If it looks better than software mode, I can see some interest in this product. Can the card go to 800x600? And at 320x200 - 30.4 fps. Isn't the record for software mode at this resolution on an Am5x86 only 20.6 fps?

pshipkov wrote on 2021-11-24, 08:42:

3.1 board runs with all settings on max except DRAM Read WS = 1 and IBC DEVSEL# Decoding = Medium.

What OS you use ?

I’m using Windows 98.

How’d you get the 3.1 board to run with 2111 cache settings at 60MHz FSB? I’ve got 12NS modules, 256k total, and only 3111 is stable in Dos, Windows requires 3222.