User:Ah42/Benchmarks

I was not satisfied with the combined output of FRAPS and GPU-Z, since Skyrim can become fairly CPU-bound depending on configuration, and these do not record this information. I also disliked having to remember to start/stop the monitoring. Instead, I chose to use Skyrim Performance Monitor to record all data. This has the added benefit of not needing a baseline (i.e. idle) benchmark for RAM usage, as SPM only tracks usage by the Skyrim (+enboost) processes. In partnership with some formulae and LibreOffice, the output was distilled to provide the Average() and StdDev() from the moment of RAM + DiskIO Spike when choosing a New Game, through the program quitting (using console "qqq" at the first auto-save)

The data monitored with this method:

CPU Load
GPU Load
GPU Temperature (1)
Disk IO (2)
Threads
System RAM (used by Skyrim) (3)
Dedicated VRAM (used by Skyrim)

Notes:

(1) Ambient temperature fluctuated during the multi-day benchmark period, as the weather went from summer extreme to autumn mild. Starting (minimum) GPU temperature is provided, and all other temperatures are listed as relative to this base.
(2) Official BSA's are on a RAMDisk, and symlinked to from my Data directory which is on SSD. Mod Organizer is also on SSD. This appears to cause SPM not to record Disk IO for the Official BSA's, which means the mod-added IO stands out more starkly against the baseline.
(3) SPM tracks total System RAM. Minimum used RAM (i.e. at the start of logging) has been subtracted to find Skyrim RAM usage