I've done some experimentation with different driver versions and benchmark scores while I was troubleshooting other issues.  I haven't tested every driver version for the RTX 30X0 series, but I've tested a few.  So far, it seems that driver version 457.30 has given me my best benchmark scores.  In some of the various 3DMark benchmarks, 457.30 has given me scores that are 200-300 points higher than later drivers.
 
To beat the benchmark scores I get at 73F/23C with driver version 457.30, I need to use much cooler ambient temperatures with later drivers.  However version 457.30 less stable than later drivers in some benchmarks.
 
Has anyone else had similar results with that driver version?

Release day drivers will get you the best benchmarks since they have a way more agressive boost up.
But they will also crash in games a lot more.

Sajin
3dmarks scores that combine a cpu result, and a gpu result to get the final score can vary as much as 3% from run to run, so you might think different drivers are actually helping when that may not be.

If its a GPU only benchmark with little influence by the GPU will the results vary much if the card rusn at the same temperature and settings?

bavor

Sajin
3dmarks scores that combine a cpu result, and a gpu result to get the final score can vary as much as 3% from run to run, so you might think different drivers are actually helping when that may not be.

If its a GPU only benchmark with little influence by the GPU will the results vary much if the card rusn at the same temperature and settings?

Shouldn’t see much difference.