http://kegel.com/wine/yagmarkdata/wine-1.1.44-19-vs-wine-1.1.44-72.txt shows yesterday's wine's performance compared to today's. Highlights:
Comparing wine-1.1.44-19 with wine-1.1.44-72 benchmark_variable wine-1.1.44-19 wine-1.1.44-72 ratio 3dmark06_3DMark_Score 3377.00 3388.00 1.00 3dmark2000_3DMark_Result 16662.00 16917.00 1.02 3dmark2001_3DMark_Score 15691.00 17633.00 1.12 heaven2_d3d9_FPS 9.31 11.08 1.19 heaven2_gl_FPS 16.12 16.21 1.01
Golly, looks like http://www.winehq.org/pipermail/wine-cvs/2010-May/066833.html was a win.
Here are a few notes about the raw results.
A couple benchmarks (really short ones?) seem to have strongly bimodal distributions on Wine. For instance, here are ten measurements on each OS of 3dmark06_CPU2_Red_Valley:
wine-1.1.44-19 1.33 1.36 1.36 1.36 1.37 1.37 1.37 1.37 1.37 1.37 wine-1.1.44-72 1.30 1.31 1.31 1.31 1.31 1.32 1.32 1.32 1.32 1.38 Vista 1.39 1.39 1.39 1.39 1.39 1.39 1.39 1.39 1.39 1.39
For some reason, today's wine usually had the slow result, whereas yesterday it usually had the fast result. Vista didn't suffer from this, and was slightly faster, too. Why does Wine return a slow result sometimes?
3dmark2000_Game_1_Helicopter_High_Detail was quite variable under Wine (and slightly better today), but hardly varied at all on Vista: wine-1.1.44-19 133.70 133.90 141.60 145.00 145.40 151.60 159.30 161.80 164.40 173.60 wine-1.1.44-72 139.50 144.70 154.10 157.40 157.90 162.40 170.50 178.80 180.80 198.60 Vista 281.60 282.50 282.90 283.20 283.40 283.60 283.70 283.80 284.00 284.50
What can explain such variability?
heaven2_d3d9_FPS, on the other hand, didn't vary much on Wine - and was quite a bit better today: wine-1.1.44-19 9.23 9.28 9.30 9.30 9.30 9.31 9.31 9.32 9.32 9.32
wine-1.1.44-72 11.05 11.07 11.07 11.07 11.08 11.08 11.08 11.09 11.10 11.10 Vista 18.37 18.37 18.38 18.38 18.38 18.38 18.38 18.38 18.38 18.38
Perhaps I should strip down my system and run fvwm95 rather than Gnome, etc., to see if I can reduce the noise a bit more. - Dan
On Tue, May 11, 2010 at 3:27 PM, Dan Kegel dank@kegel.com wrote:
Why does Wine return a slow result sometimes?
Solar flares! ;)
I can't stay I notice this discrepancy in actual game play day after day though.
I personally would be looking at memory available as the cause. In general you want to make sure you have plenty free as any excessive swapping causes performance loss. You can test this yourself by trying a game with 1gb free memory and then trying the game again with only enough to fit the games memory requirements (or even force it over). The difference will be easily noticeable.
On Tue, May 11, 2010 at 7:27 AM, Dan Kegel dank@kegel.com wrote:
http://kegel.com/wine/yagmarkdata/wine-1.1.44-19-vs-wine-1.1.44-72.txt shows yesterday's wine's performance compared to today's. Highlights: Comparing wine-1.1.44-19 with wine-1.1.44-72 benchmark_variable wine-1.1.44-19 wine-1.1.44-72 ratio 3dmark06_3DMark_Score 3377.00 3388.00 1.00 3dmark2000_3DMark_Result 16662.00 16917.00 1.02 3dmark2001_3DMark_Score 15691.00 17633.00 1.12 heaven2_d3d9_FPS 9.31 11.08 1.19 heaven2_gl_FPS 16.12 16.21 1.01 Golly, looks like http://www.winehq.org/pipermail/wine-cvs/2010-May/066833.html was a win. Here are a few notes about the raw results. A couple benchmarks (really short ones?) seem to have strongly bimodal distributions on Wine. For instance, here are ten measurements on each OS of 3dmark06_CPU2_Red_Valley: wine-1.1.44-19 1.33 1.36 1.36 1.36 1.37 1.37 1.37 1.37 1.37 1.37 wine-1.1.44-72 1.30 1.31 1.31 1.31 1.31 1.32 1.32 1.32 1.32 1.38 Vista 1.39 1.39 1.39 1.39 1.39 1.39 1.39 1.39 1.39 1.39 For some reason, today's wine usually had the slow result, whereas yesterday it usually had the fast result. Vista didn't suffer from this, and was slightly faster, too. Why does Wine return a slow result sometimes? 3dmark2000_Game_1_Helicopter_High_Detail was quite variable under Wine (and slightly better today), but hardly varied at all on Vista: wine-1.1.44-19 133.70 133.90 141.60 145.00 145.40 151.60 159.30 161.80 164.40 173.60 wine-1.1.44-72 139.50 144.70 154.10 157.40 157.90 162.40 170.50 178.80 180.80 198.60 Vista 281.60 282.50 282.90 283.20 283.40 283.60 283.70 283.80 284.00 284.50 What can explain such variability? heaven2_d3d9_FPS, on the other hand, didn't vary much on Wine - and was quite a bit better today: wine-1.1.44-19 9.23 9.28 9.30 9.30 9.30 9.31 9.31 9.32 9.32 9.32
wine-1.1.44-72 11.05 11.07 11.07 11.07 11.08 11.08 11.08 11.09 11.10 11.10 Vista 18.37 18.37 18.38 18.38 18.38 18.38 18.38 18.38 18.38 18.38 Perhaps I should strip down my system and run fvwm95 rather than Gnome, etc., to see if I can reduce the noise a bit more.
- Dan
On my laptop I also had some weird results in some game depending on when I started it. In my case I suspected that the GPU wasn't at the maximum clock speeds yet. Depending on what GPU you are using you might also have '2d' and '3d' clocks. Try to force it to maximum clocks using nvidia-settings.
Roderick
On Tue, May 11, 2010 at 1:27 AM, Roderick Colenbrander thunderbird2k@gmail.com wrote:
On my laptop I also had some weird results in some game depending on when I started it. In my case I suspected that the GPU wasn't at the maximum clock speeds yet. Depending on what GPU you are using you might also have '2d' and '3d' clocks. Try to force it to maximum clocks using nvidia-settings.
Thanks, I'll try.
Looks like nvidia-settings can only read the clocks, not set them? nvidia-settings -q GPUPerfModes -t; nvidia-settings -q GPUCurrentPerfLevel -t; nvidia-settings -q GPUCurrentClockFreqs -t displays their current value.
Setting them appears to require changing kernel module parameters (either in xorg.conf or /etc/modprobe.d/mumble). Related pages: http://tutanhamon.com.ua/technovodstvo/NVIDIA-UNIX-driver/ https://bugs.launchpad.net/ubuntu/+source/linux-restricted-modules-2.6.24/+b...
Setting the clocks must be performed using the gui. You need to have the Coolbits option in your xorg.conf. It reminds me to still update nvclock which can do this from cli fine ;)
Roderick
On Tue, May 11, 2010 at 3:45 PM, Dan Kegel dank@kegel.com wrote:
On Tue, May 11, 2010 at 1:27 AM, Roderick Colenbrander thunderbird2k@gmail.com wrote:
On my laptop I also had some weird results in some game depending on when I started it. In my case I suspected that the GPU wasn't at the maximum clock speeds yet. Depending on what GPU you are using you might also have '2d' and '3d' clocks. Try to force it to maximum clocks using nvidia-settings.
Thanks, I'll try.
Looks like nvidia-settings can only read the clocks, not set them? nvidia-settings -q GPUPerfModes -t; nvidia-settings -q GPUCurrentPerfLevel -t; nvidia-settings -q GPUCurrentClockFreqs -t displays their current value.
Setting them appears to require changing kernel module parameters (either in xorg.conf or /etc/modprobe.d/mumble). Related pages: http://tutanhamon.com.ua/technovodstvo/NVIDIA-UNIX-driver/ https://bugs.launchpad.net/ubuntu/+source/linux-restricted-modules-2.6.24/+b...
-
Dan Kegel -
Edward Savage -
Roderick Colenbrander