I found some results on Vogons, but its not a large downloadable dataset.
Re: My 3DMark01 Mega Thread

Across threads, the methodology seems inconsistent. I would suggest aligning the 3DMark version to the GPU and driver capabilities, otherwise you're testing a variably mixed CPU and GPU combination of software and hardware rendering. Hence I'd shift to the 2000 version for DX7 cards, etc.

Any suggestions on how to collate all results from all threads into a single source of truth?

Not my results but here is some for you: https://www.512bit.net/3dmark2001se.html & https://hw-museum.cz/article/12/the-ultimate- … -2000---2002-/5

Trying to find a universal list is going to be damn near impossible if not for all the different cpu & motherboard varieties and I have found people dont even like to use the default settings & resolutions half the time. (not talking about the two links listed above but in the many forum posts I have seen)

Those early 3D Marks were heavily CPU dependent. Same video card would get wildly different results on e.g. a 933 MHz Pentium III and 3.06 GHz Pentium 4

I suggest searching your videocard on HWBot.org and comparing with similar systems. There's a result there with a Sempron 2200+ and a Radeon 9500 scoring about 1000 less than yours, so I'd say you are in the ballpark of where you should be.

Errius wrote on 2026-05-06, 19:50:

Those early 3D Marks were heavily CPU dependent. Same video card would get wildly different results on e.g. a 933 MHz Pentium III and 3.06 GHz Pentium 4

I would expect that to happen if the 3DMark generation does not match the GPU generation.

If you put a DX7 GPU on 3DMark 2003 it might run, but all the DX8/9 stuff will be pushed to software.

Grem Five wrote on 2026-05-06, 19:34:

Not my results but here is some for you: https://www.512bit.net/3dmark2001se.html & https://hw-museum.cz/article/12/the-ultimate- … -2000---2002-/5

Trying to find a universal list is going to be damn near impossible if not for all the different cpu & motherboard varieties and I have found people dont even like to use the default settings & resolutions half the time. (not talking about the two links listed above but in the many forum posts I have seen)

Thanks. That is enough to start with.

I wonder how 512.bit arrived at the rarity score?

One of my cards that I feel is rare, because I've only ever seen one, is a Radeon 7000 128bit VRAM. It's not represented in any benchmark lists currently and I'm looking forward to pitting against my TNT2 Ultra in 3Dmark99 because that's the closest match on paper.

I would suggest you put more info on screenshots... (at least GPU configuration/driver versions + it's clocks, and CPU/RAM settings)

Validating score to get idea if you setup anything wrong is going to be hard.
Main issue is that not many people share mass GPU test results (that actually allow you to judge if performance of one GPU is within expectations), and second would be actual performance level that is comparable to your testing setup. Even having the same board and CPU, just different OC settings WILL make comparing stuff inaccurate. You will be able to pick up GART being weird, but finding for example a driver issue (which unnaturally decreases performance) - won't be as easy.

You could limit yourself to pure GPU bound metrics (like Fillrates), example :

^BUT that's useless for picking issues with shader performance on later cards (or AGP speed limitations).

Here's my 9500 score (12800) :

and GF3 Ti 200 (~8050) :

Note : 3DMark 03 by default simply doesn't run/skips shader tests on DX7 card (marked as "unsupported").