As the Mx440 is limited to 60fps in every resolution, I think it must be vsync or another setting. As you turn of vsync everywhere, I would change the refresh rate to 75hz and see if this affects your benchmarks to be sure.

pc-sound-legacy wrote on 2021-02-15, 17:58:

As the Mx440 is limited to 60fps in every resolution, I think it must be vsync or another setting. As you turn of vsync everywhere, I would change the refresh rate to 75hz and see if this affects your benchmarks to be sure.

Its weird because I think it is turned off (running 3dmark99 also explicitely states that vsync is turned off). I also checked with powerstrip and rivaruner. During the q3 benchmark I sometimes see it go to 70, 80, ... 100fps.

I can get it to 65fps average if I set everything to low (that made me think there is no 60fps cap somewhere).

Are you aware of any other settings that might influence this ?

Motherboard chipset limitation, overly conservative memory timings, L2 cache being disabled, chipset drivers not installed, are some ideas that spring to mind .

What motherboard you using and is your FSB at 100MHz ?

Your score seems accurate. I'm also looking at my results, and they're very different from Anand's. He got well over 100 FPS with an Athlon 750, which is definitely not the case for me. Only my 1 GHz Coppermine was able to break the 100 FPS mark.

I remember that I also had this dilemma in the past, when I compared my results to Anand's, and it seems that in his benchmarks he was using the lowest possible settings including Vertex lighting and low geometric details.
These settings put less pressure on the CPU, which is why scores are higher.

darry wrote on 2021-02-15, 18:34:

Motherboard chipset limitation, overly conservative memory timings, L2 cache being disabled, chipset drivers not installed, are some ideas that spring to mind .

What motherboard you using and is your FSB at 100MHz ?

Thx for the pointers

MSI MS6163 motherboard with Via Apollo chipset
100MHz FSB
Via 4in1 drivers installed in win98se
Played with BIOS memory timings (set to fastest).
L2 cache is enabled

RetroSpector78 wrote on 2021-02-15, 20:12:

Thx for the pointers […]
Show full quote

darry wrote on 2021-02-15, 18:34:

Motherboard chipset limitation, overly conservative memory timings, L2 cache being disabled, chipset drivers not installed, are some ideas that spring to mind .

What motherboard you using and is your FSB at 100MHz ?

Thx for the pointers

MSI MS6163 motherboard with Via Apollo chipset
100MHz FSB
Via 4in1 drivers installed in win98se
Played with BIOS memory timings (set to fastest).
L2 cache is enabled

The Via Apollo chipset may be part of the explanation . See https://www.anandtech.com/show/128/3

I am just able to crack 100fps on my Athlon 750 Classic in Quake 3 demo.

https://www.vogonswiki.com/index.php/Athlon_750_Benchmarks

mwdmeyer wrote on 2021-02-15, 20:57:

I am just able to crack 100fps on my Athlon 750 Classic in Quake 3 demo.

https://www.vogonswiki.com/index.php/Athlon_750_Benchmarks

You have a much better video card collection than I have 😀

My "old" fast AGP cards start with the Geforce2 MX and Geforce4MX.

Wonder if drivers have anything to do with it... "slow" CPU like the Pentium III 550MHz with "newer" drivers.
I saw the effect on the Riva TNT when I was using new drivers Vs 0ld drivers.
But I always picked the lowest versions of the drivers that were possible.

darry wrote on 2021-02-15, 20:51:

The Via Apollo chipset may be part of the explanation . See https://www.anandtech.com/show/128/3

Great ! 😀 Now all I need to do is move everything to an intel chipset board, find a script that will execute all of these benchmarks again in an automated unattended way and dump all that data on my Google Sheet 😀

Think I have a Compaq Pentium III desktop. That one will most likely have an intel chipset. Might give that a try.

Did you try lowering the details as I suggested? I’m pretty sure that there’s no way a 550MHz Coppermine can break the 100FPS mark when using high details, no matter what platform / drivers you use (btw, I think nvidia 12.41 are the best drivers for Quake 3, if I remember correctly - of course, you would need an older card).

Hi RetroSpector78!

I'm a big fan of your TY videos 😀
I agree that the issue should be related to the platform/chipset and detail settings, but could be worth checking if anything changes with the various Quake III patch versions; which one did you used?

My old benchmarks with a dual P3 700 overlocked with a GF2 and a G400
http://www.murc.ws/archive/index.php/t-26822.html
http://www.murc.ws/archive/index.php/t-26815.html

P3 1GHZ overclocked with GF3
https://www.compatdb.org/forums/topic/16588-q … s#comment-75735

Lol. Reading my posts. The joys of overclocking in my youth. 😀 Now I just do it when it's close to time to move to new hardware. heh

Try different versions of the via drivers and see if that makes a diff

Wow. Went from 1600x1200 in 2001 to 3840x1600 in 2021.

With newer video cards I find you can get dramatically better performance benchmarking in XP compared to windows 98. This is doubly true for nvidia cards.

DosFreak wrote on 2021-02-15, 23:33:

My old benchmarks with a dual P3 700 overlocked with a GF2 and a G400 http://www.murc.ws/archive/index.php/t-26822.html http://w […]
Show full quote

My old benchmarks with a dual P3 700 overlocked with a GF2 and a G400
http://www.murc.ws/archive/index.php/t-26822.html
http://www.murc.ws/archive/index.php/t-26815.html

P3 1GHZ overclocked with GF3
https://www.compatdb.org/forums/topic/16588-q … s#comment-75735

Lol. Reading my posts. The joys of overclocking in my youth. 😀 Now I just do it when it's close to time to move to new hardware. heh

Try different versions of the via drivers and see if that makes a diff

Wow. Went from 1600x1200 in 2001 to 3840x1600 in 2021.

In both of those cases we’re talking about dual CPU configurations, on SMP-aware operating systems. Quake 3 is known to be one of the first games (if not the first) to scale very well with SMP.
Again, a single P3 550E will never achieve over 100 FPS in Quake 3 with high details. It’s just not possible. 😀

PS: Your rigs were absolutely insane for the time... Damn. In January 2001 I still had an AMD K6-2 500 MHz (and I had just bought a cheap Riva TNT 2 M64) 🙁 Still, good times...

bloodem wrote on 2021-02-15, 21:28:

Did you try lowering the details as I suggested? I’m pretty sure that there’s no way a 550MHz Coppermine can break the 100FPS mark when using high details, no matter what platform / drivers you use (btw, I think nvidia 12.41 are the best drivers for Quake 3, if I remember correctly - of course, you would need an older card).

Some results :

• 640x480x16 low detail (textures & geometry) with Vertex lighting : 73fps (DRAM : turbo)
• 640x480x16 low detail (textures & geometry) with Lightmap lighting : 66fps (DRAM : turbo)
• 640x480x16 high detail (textures & geometry) with Lightmap lighting : 63fps. (DRAM : turbo)
• 640x480x16 high detail (textures & geometry) with Lightmap lighting : 61fps (DRAM : normal)
• 640x480x16 high detail (textures & geometry) with Lightmap lighting : 59fps (DRAM : setup defaults were 10ns != Normal)

Vertex lighting is definitely cheating 😀

What video card and drivers did you use? With a Pentium 3 550, I guess the best results are achieved with a GeForce 2/GeForce 3 & older drivers.

bloodem wrote on 2021-02-16, 09:20:

What video card and drivers did you use? With a Pentium 3 550, I guess the best results are achieved with a GeForce 2/GeForce 3 & older drivers.

The above tests were done with a Geforce4 MX440 with the Detonator 45.23 drivers
For the Geforce2 / Riva TNT cards I typically go with the 08.05 drivers.

If you have one, try a GeForce 2 GTS, with driver version 12.41 (which I think is one of the fastest for Quake 3)

If you use full game patched note that early versions had different demos despite same names. They are not comparable even if internals were the same.