I notice in the PTS version 1.60 that ETQW seems to disable v-sysnc or the fps cap built into the game during the first test. But on the 2nd and 3rd test it either turns v-sync on or enables a fps cap of 60.
====================================
ET: Quake Wars Demo (Run 1 of 3)
====================================
4938 frames in 65.3 seconds = 75.6 fps
====================================
ET: Quake Wars Demo (Run 2 of 3)
====================================
4938 frames in 83.4 seconds = 59.2 fps
====================================
ET: Quake Wars Demo (Run 3 of 3)
====================================
4938 frames in 83.1 seconds = 59.5 fps
####################################
ET: Quake Wars Demo:
1680 x 1050
75.6 Frames Per Second
59.2 Frames Per Second
59.5 Frames Per Second
Average: 64.76 Frames Per Second
Averaging three results and you end up with an incorrect value for average performance between the 3 tests.
I have not checked the 1.8beta yet but I noticed this and didn't see any other posts off hand about the issue.
Just did some more testing. The Doom3 bench is also going to be off of a normal run. You will need to set the usecache option for the demo as each subsequent run after the first run will be higher frames than the first run as the first run will do the caching for the second and third run.
====================================
Doom 3 (Run 1 of 3)
====================================
2148 frames rendered in 27.1 seconds = 79.3 fps
====================================
Doom 3 (Run 2 of 3)
====================================
2148 frames rendered in 26.7 seconds = 80.4 fps
====================================
Doom 3 (Run 3 of 3)
====================================
2148 frames rendered in 25.8 seconds = 83.4 fps
########################################
Doom 3:
Resolution: 1680 x 1050 - Quality: Ultra
79.3 Frames Per Second
80.4 Frames Per Second
83.4 Frames Per Second
Average: 81.03 Frames Per Second
########################################
If you were to use the usecache from the get go on the timedemo you shouldn't have the gradual increase in frames per second. It will also produce closer frames to between each run instead of having a 4fps delta, wich I know isn't huge but just saying most runs with usecache = ~.5 fps variance.
====================================
ET: Quake Wars Demo (Run 1 of 3)
====================================
4938 frames in 65.3 seconds = 75.6 fps
====================================
ET: Quake Wars Demo (Run 2 of 3)
====================================
4938 frames in 83.4 seconds = 59.2 fps
====================================
ET: Quake Wars Demo (Run 3 of 3)
====================================
4938 frames in 83.1 seconds = 59.5 fps
####################################
ET: Quake Wars Demo:
1680 x 1050
75.6 Frames Per Second
59.2 Frames Per Second
59.5 Frames Per Second
Average: 64.76 Frames Per Second
Averaging three results and you end up with an incorrect value for average performance between the 3 tests.
I have not checked the 1.8beta yet but I noticed this and didn't see any other posts off hand about the issue.
Just did some more testing. The Doom3 bench is also going to be off of a normal run. You will need to set the usecache option for the demo as each subsequent run after the first run will be higher frames than the first run as the first run will do the caching for the second and third run.
====================================
Doom 3 (Run 1 of 3)
====================================
2148 frames rendered in 27.1 seconds = 79.3 fps
====================================
Doom 3 (Run 2 of 3)
====================================
2148 frames rendered in 26.7 seconds = 80.4 fps
====================================
Doom 3 (Run 3 of 3)
====================================
2148 frames rendered in 25.8 seconds = 83.4 fps
########################################
Doom 3:
Resolution: 1680 x 1050 - Quality: Ultra
79.3 Frames Per Second
80.4 Frames Per Second
83.4 Frames Per Second
Average: 81.03 Frames Per Second
########################################
If you were to use the usecache from the get go on the timedemo you shouldn't have the gradual increase in frames per second. It will also produce closer frames to between each run instead of having a 4fps delta, wich I know isn't huge but just saying most runs with usecache = ~.5 fps variance.
Comment