User:Z929669/Benchmarks

Benchmark conditions: STEP 2.2.1 Guide System Specs

The following table provides means +/- standard deviations for:

• GPU temperature
• GPU load
• Dedicated VRAM (on-card memory) - This caps at 1,996 MB on my system due to crossfireX, which adds memory used on BOTH cards. Effective VRAM is still 1 GB.
• System VRAM (swapped-to-system memory) - lower is better, > 500 MB produces moderate/severe stuttering
• FPS

Note that the windows background numbers are included in the first rows. This should be subtracted from all of the data following as a proxy for resources consumed only by Skyrim. The build is full Baseline STEP 2.2.1 with the exception that I used 2k rather than 1k textures for Skyrim HD and Serious HD (which costs a lot in terms of memory). Benchmark 1 (though STEP 1.F.2 in the Guide) shows great performance, as expected. Benchmark 2 also shows good performance, even though there is some bleed-over of dedicated VRAM into system. No noticeable stuttering occurs at this point.

Benchmarks 3 - 5 all show a marked increase in VRAM usage and a decrease in FPS. The frame drop is a mininal issue, as it still remains close to 30 FPS on average. The VRAM, however, is problematic, as noticeable moderate stuttering occurred in these runs. Texture optimization using DDSopt did not alleviate the problem, and neither did ATTK. The final solution was to run the normal maps (*_n.dds) through DDSopt with big texture overhauls, clamping their size to 1k. Reducing the VRAM bleed-over by the additional 100 MB was enough to noticeably reduce stuttering to an acceptably minimal level.

Run	Stat	GPU Temp. [?C]	GPU Load [%]	Dedicated VRAM [MB]	System VRAM [MB]	FPS
Windows background	Mean	52.5	0.5	85.4	32.6	n/a
	Std Dev	0.7	0.9	8.6	0.5	n/a
(1) After 1.F.2 Pure Vanilla	Mean	60.4	65.9	1,961.3	89.7	51.5
	Std Dev	1.6	7.5	21.6	12.6	8.6
(2) After 2.D Fixes DDSopt Vanilla HD Textures (1k Constrained)	Mean	51.1	66.6	1,876.6	85.2	52.7
	Std Dev	1.4	7.6	8.8	64.6	7.7