Have you tried winding back the video card drivers to the oldest supported?

Are you running in a pure DOS boot to run Quake or are you exiting Windows to DOS to play ?

Set vid_wait 0 and vid_nopageflip 1.

kolderman wrote on 2020-03-28, 23:34:

Have you tried winding back the video card drivers to the oldest supported?

I am running from MS-DOS, so there aren't any drivers involved.

Horun wrote on 2020-03-29, 02:45:

Are you running in a pure DOS boot to run Quake or are you exiting Windows to DOS to play ?

I am running from pure DOS

The Serpent Rider wrote on 2020-03-29, 08:38:

Set vid_wait 0 and vid_nopageflip 1.

I assume you mean that I should enter in Quake console?

It should affect my other systems based on 440BX/I815 also, as they ran the benchmark with same settings but producing way higher fps.

SubZero wrote on 2020-04-01, 11:49:

The Serpent Rider wrote on 2020-03-29, 08:38:

Set vid_wait 0 and vid_nopageflip 1.

I assume you mean that I should enter in Quake console?

It should affect my other systems based on 440BX/I815 also, as they ran the benchmark with same settings but producing way higher fps.

Try running one of the write combine enabling programs before starting Quake, such as mtrrlfbe or fastvid (see https://www.philscomputerlab.com/dos-graphics-boost.html). I'm not certain if I'm remembering 100%, but I think I recall reading that some BIOS enabled this automatically, while many others did not, which could explain the performance difference between your systems.

And yes, those are Quake cvars which you enter at the console (or you can specify on the command line.)

Edit: nevermind, I went back and reread your first post and see that you are already using them. Serves me right for replying too early and with insufficient coffee. Duh. Sorry.

Vid_wait will affect final score on any system if it's set to 1 or sometimes to 2. Vid_nopageflip set to 1 will most likely degrade Quake performance on P3 Intel chipsets, but you're free to try.

I’ve always heard bad things about VIA agp, so I have to wonder if that’s the issue. I’m sure it’s a huge pain to test but I’d be curious to hear if you had similar results with a PCI videocard.

Not worse as ali AGP

AGP implementation has little to no difference in DOS

The Serpent Rider wrote on 2020-04-03, 09:44:

AGP implementation has little to no difference in DOS

Ok but he’s not the first to report this:

Re: PCI card faster than AGP for DOS games?

Same or similar chipset.

I can report that results are similar with different agp card and pci card. I think I will give up on this issue since dos/VESA performance is still sufficient

Is bank interleaving enabled by default on this board?

Garrett W wrote on 2020-04-06, 14:45:

Is bank interleaving enabled by default on this board?

There is no such option in bios, so guess it is enabled by default

Yeah, you'd think that, but a lot of motherboards have it disabled actually! I'm not sure if it'd make any difference in DOS, but it did help performance under Windows. You can use SISoft Sandra to see if it is enabled and here's the patch in case you need to enable it:
http://www.georgebreese.com/net/software/#INT

I also just realized that I use this exact same board on my main Retro PC, but the specs are way different than yours. I use a VIA C3 (Nehemiah core) @1.5GHz and a Voodoo5. At that frequency, the C3 is roughly equivalent to a PIII 733 in 3D games, would it help if I tried Quake on this system and give you my results? Which timedemos did you try?

Garrett W wrote on 2020-04-06, 20:10:

Yeah, you'd think that, but a lot of motherboards have it disabled actually! I'm not sure if it'd make any difference in DOS, but it did help performance under Windows. You can use SISoft Sandra to see if it is enabled and here's the patch in case you need to enable it:
http://www.georgebreese.com/net/software/#INT

I also just realized that I use this exact same board on my main Retro PC, but the specs are way different than yours. I use a VIA C3 (Nehemiah core) @1.5GHz and a Voodoo5. At that frequency, the C3 is roughly equivalent to a PIII 733 in 3D games, would it help if I tried Quake on this system and give you my results? Which timedemos did you try?

Sure if you have time, please give it a run. I have only tested demo 1. If the logic is correct you should only get a bit more than 30 fps in 640x480.

I tried the link to the bank interleave patch withour luck. I also tried to google it without luck. Can you possibly provide it here?

100FPS vs 64FPS, the difference is too drastic for even a mis-tuned BIOS settings.

How was the memory setup, number of DIMM sticks and DS/SS?

I would suggest that you may have a mis-configured MTRRs even after running MTRRLFBE/FastVid. The only way to tell is to get a tool to dump out all the MTRRs in real DOS and make sure the VESA LFB is covered by the MTRRs as Write-Combined region. And, Quake actually is using LFB VESA mode instead of bank-switched mode.

kjliew wrote on 2020-04-06, 20:43:

100FPS vs 64FPS, the difference is too drastic for even a mis-tuned BIOS settings.

How was the memory setup, number of DIMM sticks and DS/SS?

I would suggest that you may have a mis-configured MTRRs even after running MTRRLFBE/FastVid. The only way to tell is to get a tool to dump out all the MTRRs in real DOS and make sure the VESA LFB is covered by the MTRRs as Write-Combined region. And, Quake actually is using LFB VESA mode instead of bank-switched mode.

In both cases I used 2x256MB dual sided sticks CL2.

In both cases I ran MTRRLFBE through "Phils Dosbench kit". Without MTRRLFBE enabled I get around 25 FPS in the VIA board. On the BX-board the patch was not giving any difference, full speed without the patch.