I spent some time tonight trying to get to the bottom of the frame rate issues that Brink suffers from. The most frustrating thing is that the performance is fine 70% of the time but as soon as you go outdoors it seems to plummet. Here are my findings.
Experiments 1 and 2 are performed at an outdoor location where the frame rate always drops below 60 FPS for me. There are many other such locations. Experiment 3 is performed at various locations that have satisfactory performance.
Hardware Used:
CPU: AMD 1090t @ various speeds
RAM: 8 GB, 2000 MHz
GPU: HD6970 x 2
OS: Windows 7 Ultimate 64
Drivers: 11.6
CAP: 11.6_cap1_june15
Resolution: 1920x1080
Experiment 1, r_shadows:
Performance is tested at lowest possible settings and compared to results at high settings. Both settings are tested with r_shadows turned off and on.
detail = high
crossfire = on
CPU = 3.25 GHz
seta image_anisotropy “8”
seta r_useAntiAliasing “1”
seta r_multiSamples “4”
seta r_shadows “0”
FPS = 48
detail = high
crossfire = on
CPU = 3.25 GHz
seta image_anisotropy “8”
seta r_useAntiAliasing “1”
seta r_multiSamples “4”
seta r_shadows “1”
FPS = 33
detail = low
crossfire = on
CPU = 3.25 GHz
seta image_anisotropy “1”
seta r_useAntiAliasing “0”
seta r_multiSamples “4”
seta r_shadows “0”
FPS = 49
detail = low
crossfire = on
CPU = 3.25 GHz
seta image_anisotropy “1”
seta r_useAntiAliasing “0”
seta r_multiSamples “4”
seta r_shadows “1”
FPS = 36
The minimum frame rate doesn’t depend much at all on texture detail, anti aliasing or anisotropic filtering. This indicated the performance is not GPU bound. Regardless of other settings, turning shadows off results in around 15 FPS gain.
Experiment 2, overclocking:
Performance is tested using the same settings but overclosking the CPU. All trials have;
detail = high
crossfire = on
seta image_anisotropy “8”
seta r_useAntiAliasing “1”
seta r_multiSamples “4”
seta r_shadows “0”
CPU = 2.5 GHz
FPS = 44
CPU = 3.25 GHz
FPS = 48
CPU = 3.75 GHz
FPS = 58
Overclocking the CPU increases the minimum FPS significantly. This indicates that the performance is limited by the CPU speed rather than GPU. The results from experiment 1 support this since the minimum FPS did not depend significantly on detail settings, anisotropic filtering or anti aliasing.
It would seem that a CPU frequency higher that 3.8 GHz is required in order to prevent the game from dipping below 60 fps with shadows turned off. An even higher frequency would be required with shadows turned on.
Experiment 3, crossfire:
Performance is tested with crossfire turned on and off at various locations. All runs have,
detail = high
CPU = 3.25 GHz
seta image_anisotropy “8”
seta r_useAntiAliasing “1”
seta r_multiSamples “4”
seta r_shadows “0”
crossfire = on
FPS = 94
crossfire = off
FPS = 116
crossfire = on
FPS = 101
crossfire = off
FPS = 106
crossfire = on
FPS = 119
crossfire = off
FPS = 122
These results are extremely disappointing. Crossfire has a negative impact on performance. The performance hit is usually negligible however there are locations where the performance drops up to 20% using crossfire. You are better off turning crossfire off for this game. I have also noticed that framerate is more variable with crossfire turned on.
Test locations:
These screenshots are taken looking directly at the floor at all test locations so that others may replicate the results if they desire.
Brink-posABCD
Brink-posEF
Brink-posGH
Brink-posIJ
All screenshots:
This is an album of all the screenshots.
All-Results
Summary:
-> Even with latest drivers there are numerous locations (mostly outdoors) with unacceptable frame rates.
-> Performance at these low FPS locations is severely CPU bound, overclock to over 3.8 GHz if at all possible.
-> seta r_shadows “0” will raise your minimum FPS by around 15 including at problem locations.
-> Crossfire scaling is still negative.
-> This is all very disappointing.