User:Gargan/Benchmarking: Difference between revisions
No edit summary |
(→Plots) |
||
Line 47: | Line 47: | ||
===Plots=== | ===Plots=== | ||
[[File:Vanilla_average_gpu_temp.png|300px|highslide=01:Average GPU Temperature]] | |||
Vanilla_average_gpu_temp.png | [[File:Vanilla_average_gpu_load.png|300px|highslide=01:Average GPU Load]] | ||
Vanilla_average_gpu_load.png | [[File:Vanilla_average_dedicated_memory.png|300px|highslide=01:Average Dedicated Memory]] | ||
Vanilla_average_dedicated_memory.png | [[File:Vanilla_average_dynamic_memory.png|300px|highslide=01:Average Dynamic Memory]] | ||
Vanilla_average_dynamic_memory.png | [[File:Vanilla_average_fps.png|300px|highslide=01:Average FPS]] | ||
Vanilla_average_fps.png | |||
==Step 2.D. Fixes== | ==Step 2.D. Fixes== |
Revision as of 15:45, July 7, 2014
Methodology[edit | edit source]
My current method for benchmarking is as follows.
- Start logging with HWiNFO64
- Wait at least 15 seconds
- Start Skyrim
- Start logging with Fraps as soon as fps counter appears (F11 hotkey)
- Wait at least 15 seconds after the main menu appears
- Start a new game
- Stop Fraps logging a couple of seconds after the autosave (F11 hotkey again)
- Exit Skyrim
- Stop logging with HWiNFO64
This procedure allows me to gather all sensor data with HWiNFO and also gather frametimes from Fraps. I prefer using frametimes to calculate an instantaneous FPS instead of using the approximately average FPS that is logged with either HWiNFO or Fraps. This allows me to see stuttering that would otherwise be averaged out. Good information on the use of frametimes instead of FPS is here.
Computer Specifications[edit | edit source]
My current setup is as follows.
CPU | Intel Core i7-4771 |
Motherboard | MSI Z87-GD65 GAMING (MS-7845) |
Memory | G.Skill F3-1600C7D-16GTX (2x8GB) |
GPU | EVGA 04G-P4-2768-KR GeForce GTX 760 4GB |
HDD | Samsung SSD 840 EVO 1TB |
Vanilla[edit | edit source]
I tried to get my install as close to vanilla as I could. The only mods that aren't installed using Mod Organizer, are SKSE and ENBoost. So I temporarily removed ENBoost and I start Skyrim directly. I was able to get five extremely similar benchmarks. The pale green area is the maximum and minimum for all data used to compute the average.
Mean | Std. Dev. | Run 1 error | Run 2 error | Run 3 error | Run 4 error | Run 5 error | |
---|---|---|---|---|---|---|---|
GPU Temp [°C] | 49.683 | 5.528 | -0.0674 +/- 0.3347 | -0.3944 +/- 0.3806 | 0.2826 +/- 0.3506 | 0.1285 +/- 0.3162 | 0.0507 +/- 0.3219 |
GPU Load [%] | 61.037 | 8.410 | 0.1358 +/- 1.2257 | -0.1327 +/- 1.2689 | 0.0158 +/- 1.2279 | -0.0192 +/- 1.2809 | 0.0003 +/- 1.5897 |
Dedicated Mem. [MiB] | 978.498 | 115.843 | 0.8942 +/- 3.9577 | 3.2048 +/- 4.3953 | -3.4755 +/- 4.4320 | -0.3745 +/- 4.0854 | -0.2489 +/- 4.4787 |
Dynamic Mem. [MiB] | 26.853 | 4.253 | 0.2612 +/- 1.9814 | -0.6809 +/- 1.8211 | 0.7134 +/- 1.7002 | -0.1833 +/- 1.7222 | -0.1105 +/- 2.4164 |
FPS | 60.054 | 1.985 | -0.0437 +/- 3.0736 | -0.0150 +/- 2.9123 | 0.0209 +/- 3.2278 | 0.0236 +/- 3.2037 | 0.0143 +/- 3.4866 |
Plots[edit | edit source]
Step 2.D. Fixes[edit | edit source]
Mean | Std. Dev. | Run 1 error | Run 2 error | Run 3 error | Run 4 error | Run 5 error | |
---|---|---|---|---|---|---|---|
GPU Temp [°C] | 51.499 | 5.380 | -0.3926 +/- 0.3010 | 0.1348 +/- 0.3401 | 0.2577 +/- 0.2966 | 0.0675 +/- 0.3051 | -0.0674 +/- 0.3397 |
GPU Load [%] | 66.508 | 9.595 | 0.0586 +/- 1.3212 | -0.1638 +/- 1.4314 | 0.2112 +/- 1.3684 | -0.1218 +/- 1.4770 | 0.0158 +/- 1.4454 |
Dedicated Mem. [MiB] | 1081.530 | 76.287 | -1.2601 +/- 7.7566 | 8.8648 +/- 7.2176 | -10.4308 +/- 5.8692 | -0.5784 +/- 5.8017 | 3.4045 +/- 7.8428 |
Dynamic Mem. [MiB] | 25.764 | 2.420 | -2.1542 +/- 1.0214 | 0.6317 +/- 0.9645 | 0.6164 +/- 1.1522 | 0.1491 +/- 1.0547 | 0.7570 +/- 0.9573 |
FPS | 60.088 | 2.040 | 0.0103 +/- 3.3923 | -0.0049 +/- 3.5355 | -0.0009 +/- 3.3959 | -0.0304 +/- 2.9610 | 0.0259 +/- 3.4232 |
Plots[edit | edit source]
- Step 2d fixes average gpu temp.png
- Step 2d fixes average gpu load.png
- Step 2d fixes average dedicated memory.png
- Step 2d fixes average dynamic memory.png
- Step 2d fixes average fps.png