cbjaust wrote on 2021-02-01, 06:47:

Yes, I see the same problem with 3DMark 99 Max on modern AMD Ryzen platforms both with AMD and nVidia video cards. No fix found though.
I made a video showing the issue: https://youtu.be/bmdhu_dl94k

Don't have that problem on my Ryzen 3600X/RTX 2070

bloodem wrote on 2021-02-01, 07:01:

cbjaust wrote on 2021-02-01, 06:47:

Yes, I see the same problem with 3DMark 99 Max on modern AMD Ryzen platforms both with AMD and nVidia video cards. No fix found though.
I made a video showing the issue: https://youtu.be/bmdhu_dl94k

Don't have that problem on my Ryzen 3600X/RTX 2070

on Windows 10? I'm seeing the problem right now with a 3900X and a GTX 980 Ti in Windows 10. Here are some more with the same thing in Windows 7 with a 3600X:
https://hwbot.org/submission/4321563_cbjaust_ … d_4890_40_marks
https://hwbot.org/submission/4318678_cbjaust_ … d_5770_35_marks
https://hwbot.org/submission/4315558_cbjaust_ … ega_56_29_marks

Well, yeah, the latest Windows 10 version (20H2)
However, I forgot to mention that I'm using djVoodoo, so that might explain why it's working fine for me. 😀

Tested my Asus P3B-F with all my different P6 CPU's.
GC is an Asus V7100, GF2 MX 400.
Same fast memory stick with tightest memory timings in all the FSB speeds.
Made a nice CPU scaling chart in 3DMark 99 I'd say!

AMD K6-III+ 550MHz and ELSA Victory II (3dfx Voodoo Banshee) AGP

This is my P3:

- Supermicro P6SBA (i440BX, 100MHz FSB)
- 512 MB RAM (133@100 CL2)
- Pentium 3 700 MHz Slot-1
- Voodoo 3 3000 @ 3500 (latest official driver, vsync off)

Score: 6027, CPU score: 10843

I have done some benchmarking on 3DMark 99.
Here are some of the results on three different resolutions.
1024, 800 and 640.
More Grapics results wil follow in time.
Pc details in my signature.
All cards at default clock and stock official drivers.

My voodoo 5 5500 at my system

This are for now the 3Dmark 99 results . Videocards I have here at home which do work are benchmarked with the as far as I know the best available drivers.
So this is it for the time being.

Interesting to see is that the voodoo 5 5500 beats the Geforce 4 4200 at 640x480 and 800x600.

I have arranged the test results at 800x600 and all the test are with the same settings at 16bit colors.
All of them used on my AMD XP2400+ CPU with 512MB ddr333 on a Gigabyte GA-7VRXP mainboard.

Voodoo 2 cards are tested in combination with the voodoo 5, gf4 and S3 Savage 4 and the results are all well within a margin of error.

All cards were tested at stock speeds, except for the voodoo 3 3000 where clockspeeds are shown.

More complete update:

I have already read in this thread that you have to switch off VSync within the use of 3Dmark99 via the driver in order to get over 60FPS in the first two tests.
First I tried it via the global settings of the NVIdia driver (setting not left to the application) and switched off VSync.
But that didn't work.
Then I created an individual profile for the 3dmark99.exe (disabling VSync).

What else can I try?
My system:
Windows 98
NVidia FX5700 @ 53.04

Using RivaTuner did the trick - I forced Vsync=off via this tool and it worked 😀

tpowell.ca wrote on 2019-02-09, 14:40:

Hello. Has anyone experienced a problem where the benchmark is clearly running perfectly but you get 0fps? Oddly enough in my ca […]
Show full quote

Hello.
Has anyone experienced a problem where the benchmark is clearly running perfectly but you get 0fps?
Oddly enough in my case, this only happens with a VIA C3 Nehemiah cpu, all other hardware remaining unchanged.
Same motherboard, same video card, same drivers.

C3 Ezra-T and P-III CPUs run perfect.

Can confirm the issue on VIA C3 Nehemiah. My guess is that the software makes some inaccurate assumptions in how it implements the "Intel processor CPU optimizations" (most likely referring to MMX code), and based on those assumptions (that probably give different results on a VIA C3) the FPS division calculation is wrong.

bloodem wrote on 2021-02-01, 07:01:

Don't have that problem on my Ryzen 3600X/RTX 2070

UPDATE:
Ha! Installed 3DMark99 again on my Ryzen build (Windows has been reinstalled since my last attempt), and now I have the same "0 FPS" issue on the Ryzen as well (with or without dgVoodoo). This is indeed strange!
Unfortunately, without the 3DMark source code it would be very hard to debug, as you need to dive deep into assembly language...

My "new" Diamond Viper V770 Ultra on my XP2400+ CPU.
Diamond driver 368.

Not the least disapointed with these results.

My PIII 600 scores 4800 in 3D Mark 99 in 640x480-1024x768 regardless of bit depth with Geforce FX5600. With Voodoo2 it is 3100 3D Marks in 640x480x16bit. I didn't switch off vsync though.

AlexZ wrote on 2021-07-21, 16:12:

My PIII 600 scores 4800 in 3D Mark 99 in 640x480-1024x768 regardless of bit depth with Geforce FX5600. With Voodoo2 it is 3100 3D Marks in 640x480x16bit. I didn't switch off vsync though.

But... you are using a p3 600mhz with a fx5600.
It doesnt matter to the graphics card which res you use... this is all the frames the cpu can offer the graphics card to render.

Aebtdom wrote on 2021-07-22, 12:11:

But... you are using a p3 600mhz with a fx5600.
It doesnt matter to the graphics card which res you use... this is all the frames the cpu can offer the graphics card to render.

He would get a higher score with a GeForce 2 / GeForce 3 with period correct drivers. 😀
From my benchmarks: a Pentium 3 Katmai 600 (OC @ 672 MHz / FSB 112) with a GeForce 3 Ti 200 I got 6388 points.
I think even a Riva TNT2 Pro would get a very high score.

bloodem wrote on 2021-07-22, 14:17:

He would get a higher score with a GeForce 2 / GeForce 3 with period correct drivers. 😀
From my benchmarks: a Pentium 3 Katmai 600 (OC @ 672 MHz / FSB 112) with a GeForce 3 Ti 200 I got 6388 points.
I think even a Riva TNT2 Pro would get a very high score.

I'm using the latest drivers for Windows 98 which is not the best choice. Once I have more time I will test multiple driver versions.

AlexZ wrote on 2021-07-22, 16:05:

I'm using the latest drivers for Windows 98 which is not the best choice. Once I have more time I will test multiple driver versions.

Even the oldest drivers that work with the FX5600 are too new. Although this is less of a problem with Pentium 3 compared to Socket 7.
Anyway, it depends on what you play. Benchmarks are only part of the story. As long as the games you're playing run OK, you've got nothing to worry about. 😀

bloodem wrote on 2021-07-22, 17:09:

Even the oldest drivers that work with the FX5600 are too new. Although this is less of a problem with Pentium 3 compared to Socket 7.
Anyway, it depends on what you play. Benchmarks are only part of the story. As long as the games you're playing run OK, you've got nothing to worry about. 😀

Unfortunately there is no Geforce 3 on sale here. FX5200-5600 are extremely cheap, that's why I got it. With PIII 733Mhz CPU, 133FSB score went up to 5700. But unfortunately system is not 100% stable with 133Mhz FSB.

Rerun of my Diamond Viper V770 Ultra.