I'm curious if the percent gain is reduced at resolutions of 1024x768 or greater.

For the most part is doesn't (see Incoming). Quake 2 fps didn't change at all. I didn't tested it, but Quake 3 fps will probably drop slightly, due to scenes heavy on fillrate (alpha textures).
Then again, with proper GeForce 2 MX PCI, results won't change at all or change just slightly.

I'm finally back at this Pentium Pro build and have put a proper GF2 MX400 in the system. Did you test several NVIDIA driver revisions and determine that 5.32 was the best for your Quadro card?

It's impossible to install Detonator 5.32 on Quadro 4, for obvious reason.

What I meant was, did you test several NVIDIA driver revisions and determine that 5.32 was the best for your TNT2 card; and did you test several NVIDIA driver revisions and determine that 28.32 was the best for your Quadro card?

For my PNY GF2MX400, these are the speed reports I see:

PowerStrip
GPU: 200 MHz
Memory: 167 MHz

NVTweak
GPU: 200 MHz
Memory: 334 MHz

RivaTuner
GPU: 100
Memory: 167 MHz

I'm guessing NVTweak is wrong on memory because it is assuming DDR when it is SDR memory? RivaTuner messed up on the GPU freq. again. I don't know why.

I don't think it's "the best" driver for TNT2. The whole point was to show how earlier drivers are better for slow CPUs. And hey, it's 5.32 vs 28.32, aka 23 version gap difference

For Quadro 4, 28.32 is the earliest supported driver, and I very much doubt that any newer driver will improve the scores in older games. Especially those with added GeForce FX support. With each new version, added driver overhead will just chop away your limited CPU resources even more.

RivaTuner messed up on the GPU freq. again. I don't know why.

Check the overclocking tab via driver (in RivaTuner). It will show the correct numbers.

The Serpent Rider wrote:

Check the overclocking tab via driver (in RivaTuner). It will show the correct numbers.

I'm not sure if I follow. Am I not at the right spot in this screenshot from previously? download/file.php?id=71951&mode=view

These are the options that the customise button provides. I started to wonder if the CPU rate is at some idle setting, so I ran GLQuake in window mode, but the clock rate still showed 100 MHz with the real-time graph.

You've clicked the wrong tab again.

lol! Let me try again. I'm running benchmarks right now, btw.

All the tabs in that other customise link also show the Graphics core running at 100 MHz.

No. Driver settings => System setting => Overclocking will show the correct numbers.

Anyway, that's not important. The card can overclocked via Powerstrip just as fine.

Ahhhh, thank you. I've attached a screenshot of this curious discrepancy. Off by a factor of 2.

I've run the series of benchmarks with driver version 45.23 and tonight will run it with version 6.50.

I have provided an updated comparison chart based on the Serpent Rider's observations. I have added a PNY GF2MX400 which runs at more typical clock speeds compared its ultra cheap derivative, Jaton. I ran the PNY GF2 using Detonator v45.23 for comparison against the Jaton, which was run previously using v45.23. I also ran the PNY GF2 using Detonator 6.50.

The PNY GF2 running at 200 MHz core and 167 MHz, 128-bit memory scores 45% faster than the cheaper Jaton GF2 running at only 120 MHz core and 120 MHz , 64-bit memory.

Using Detonator driver v6.50 instead of v45.23 offered a 5.5% improvement overall. There was one game in particular which appeared to have some benefit from the newer driver v45.23 and that game was Descent III.

These results beg the question - if I overclock the Jaton (120/120 MHz core/RAM) to that of the PNY (200/334), will the results equate?

Also, will the PIIOD333 now show greater framerate improvement compared to when using the PNY GF2 in games like GLHexen II, which previously showed no increase in speed with the Savage 4 (compared to the PPRO233)?

EDIT: I ran a few benchmarks with the Jaton overclocked. The maximum that RivaTuner and NVTweak would let me overclock is 180 MHz on the core and 180 MHz on the memory. I am using Detonator v6.50. Overclocked, the Jaton scored 38.1, whereas the PNY scores 40.1 fps. In GLHexen II, Jaton scored 40.1, whereas the PNY is at 48.8. Lastly, in Quake 2, the Jaton scored 43.1 fps, whereas the PNY scored 47.7 fps. Vsync enabled. So is it correct to assume that there is sufficient CPU power to keep scaling up with the GF2?

Next I should check is PIIOD. Will the PII be waiting around for the GPU, or GPU waiting for the CPU when using the PNY GF2?

scores 45% faster than the cheaper Jaton GF2

The PNY GF2 running at 200 MHz core and 334 MHz memory scores

200/166, SDR 128-bit.

Using Detonator driver v6.50 instead of v45.23 offered a 5.5% improvement overall.

Expendable 34.5% fps improvement is nothing to sneeze at.

Will the PII be waiting around for the GPU, or GPU waiting for the CPU when using the PNY GF2?

The major bottleneck would be CPU-PCI interactions. For example, AGP 1x interface (with the same card) can add ~33% more fps in Quake 3 (see PII results). It's not present in every game though.
I would say that all cards in the testing will receive performance boost with overclocking, but only in some games. Expendable is almost 100% limited by the CPU and the video card driver.

I've fixed my wording of the memory speeds. I was impressed to see these games run so quickly on such an old platform like the Pentium Pro.

How much you recon the PNY will overclock? Or is it pretty much at max right now? Do you think the Jaton can handle 180/180 MHz long term?

Tonight I will test to see if using the PNY instead of the Savage offers better scaling with the PIIOD.

How much you recon the PNY will overclock? Or is it pretty much at max right now?

The MX400 GPU can be overclocked up to 250Mhz (with active cooling) and the memory probably will handle 200Mhz or more.

Do you think the Jaton can handle 180/180 MHz long term?

GF2MX is not very power hungry, so it should be fine even with "chinese" design.

if I overclock the Jaton (120/120 MHz core/RAM) to that of the PNY (200/334), will the results equate?

No, at least not in high resolutions.

I also ran a few benchmarks with the PIIOD-333 to see how much improvement there was in frame rates. I selected two games which previously showed no improvement with the Jaton and two games which showed some improvement.

Previously, I found that GL Hexen II and Quake II showed zero improvement in frame rates when using the PIIOD-333 compared to the PPRO-233. They still show 0% improvement after using the PNY GF2 and Detonator v6.50.

I also tested MDK2 and Half-Life. Previously with the junky Jaton GF2 card, MDK2 showed a 9.5% increase with the PIIOD, while Half-Life showed a 19.3% increase. With the PNY GF2 and the PIIOD installed, MDK2 showed a 23.3% increase (from 30 fps to 37 fps) and Half-Life showed a 30.8% increase (14.8 fps to 19.3 fps).

The scaling with MDK2 and Half-Life were expected, but I do not understand why some games show no improvement.

I'll briefly try the PNY GF2 at 250 MHz and 200 MHz to see how much Half-Life improves, then I'll be wrapping up this project.

I would try Expendable also. Because S3 Savage 4 is not helping with CPU scaling in high resolution.

They still show 0% improvement after using the PNY GF2 and Detonator v6.50.

And they won't show any, with vsync on (which is broken anyway).