I conducted various tests at 3 clock settings:
175/166, which is stock
210/195, which was the fastest clock I could achieve without visual anomalies. The 210 core I feel is quite astounding! I've seen a few other Geforce2MX cards able to make 200 memory, but I was unable to do so here without visual pixel dropouts.
Let's first look at Quake3Arena v1.17, Demo001, at the
three standard resolution settings FAST, NORMAL, HIGH QUALITY, as well as HIGH
QUALITY with resolution bumped to 1024x768. I leave everything STOCK config.cfg
wise, just like ID meant it to be. This way, I'm comparing my results to my
results, not my config.cfg settings to yours, and we might as well have the eye
candy, hmm?? ;)
In most benchmarks run in Q3a, the FAST and NORMAL
settings are pretty pointless, usually showing basically CPU speed, as modern
video cards are not being taxed at these resolutions and bitdepths. Now that I
have a monitor capable of 1600x1200, this will probably be the last time I
bother with these lower resolution modes.. From stock to highest clock at the
High Quality and High Quality 1024x768, I saw an overall framerate increase of
~7% and ~17% respectively. With the texel engine limitations of the Geforce2MX,
in comparison to its big brother the Geforce2, we're seeing the biggest
framerate improvements due to mainly the increased memory clock, and to a lesser
extent, the core clockrate. All in all, these are quite significant gains to be
had from overclocking.
On to Madonion.com's 3Dmark 2000, version 1.1. Running
the default benchmark, which is @ 1024x768 16bit.
The graph is somewhat misleadingly scaled here. Overall, increases from stock 3Dmark score are 1% and 5% respectively for the two overclocked settings. I'm still somewhat wary of (no matter what Madonion claims) synthetic benchmarks, but this is probably the closest to the real world you can get in a D3D benchmarking suite. Still, I'm going to have to look into alternative methods of D3D testing.