nakedpeanut
Expert Member
I had a little inspiration and decided to do a little bench mark session using Heaven Benchmark v2.1 on my XFX 5850 Black Edition.
The reference card is clocked at
Core: 725MHz
Memory: 1000MHz
Voltage: dunno? try google
The XFX BE Standard
Core: 765MHz
Memory: 1125MHz
Voltage: 1.087v
I initially had a few problems trying to get it to over clock.. But lets just say i had my sights set very high
haha.
So built up the clock speed and memory, from the standard 765/1125 to 890/1250 @1.16v stable in both furmark and within games. Could not get the Clock speed over 900MHz as my games would just crash although furmark seemed stable...
Did a bit of research and as i found, you usually have a trade off between Core and Memory speed. You get to a point where you cant continue to push both, you have to hold one back while pushing the other. So I decided to throttle back the memory just a little to 1150MHz and then got the core to 960MHz @1.16v. With my sights still set on 1GHz
(told you they were high) i decided to give it another go. This time i pulled back to memory to 1125MHz and got the core to 1000MHz @1.21v.
And so the benchmarking began, just to see what performance increase you do get with a decent OC.
My system (as is in my sig)
CPU: i5 750 @ 4GHz
Mobo: Asus P7P55D
G Card: XFX 5850 BE
RAM: 2 x 2 GB DDR3 1333 Kingston @ 8-8-8-22
Operating system: Windows 7 (build 7600) 64bit
Heaven Settings:
Render: direct3d11
Mode: 1920x1200 fullscreen
Shaders: high
Textures: high
Filter: trilinear
Anisotropy: 4x
Occlusion: enabled
Refraction: enabled
Volumetric: enabled
Replication: disabled
Tessellation: normal
Nice image with Tessellation on image (fairly big at 2.4MB)
So i decided to do 3 runs at each of the OCs and take the average to compare to the reference.
Ref: 725/1000 @ 1.087v (never lower the voltage, saw no need to)
OC 1: 765/1125 @ 1.087v
OC 2: 960/1150 @ 1.161v
OC 3: 1000/1125 @1.21v
Measurements were:
Min frame rate
Max frame rate
Avg frame rate
Score
Temperature and fan speed was set using MSI afterburner as so image
Results:
Ref:
Min: 12
Max: 73.7
Avg: 28.7
Score: 721.6
51 C @ 37% Fan Speed
OC1:
Min: 13.1
Max: 75.7
Avg: 30
Score: 756
52 C @ 40% Fan Speed
Average of 5.2% improvement over reference. (8.9% improvement in Min)
OC2:
Min: 16.1
Max: 90.7
Avg: 35
Score: 883.3
55 C @ 51% Fan Speed
Average of 25.5% improvement over reference. (34% improvement in Min)
OC3:
Min: 17.1
Max: 91.47
Avg: 35.4
Score: 892.7
57 C at 57 % Fan Speed
Average of 28.4% improvement over reference. (42% improvement in Min)
Then used the above to calculate the improvement per measurement, then averaged that. So we are able to see a very nice improvement of 19% between OC1 and OC2 which is about a 25% increase in Clock speed, and a 22% improvement between OC1 and OC3 which is a 31% increase in clock speed.
Looking at the price difference between it (R3400) and a reference card(R3000) was R400 which is an increase of 13%, you get a performance increase of 5% (non OC). It remains to be seen how well a reference OC's.
So I definitely think it is worth it if OC'ing. It's a very good improvement especially when you consider the rise in min frame rates which have more of an effect on games than the max frame rates.
So anyone want to lend me a 5870 or 5970 so i can compare
?
The reference card is clocked at
Core: 725MHz
Memory: 1000MHz
Voltage: dunno? try google
The XFX BE Standard
Core: 765MHz
Memory: 1125MHz
Voltage: 1.087v
I initially had a few problems trying to get it to over clock.. But lets just say i had my sights set very high
So built up the clock speed and memory, from the standard 765/1125 to 890/1250 @1.16v stable in both furmark and within games. Could not get the Clock speed over 900MHz as my games would just crash although furmark seemed stable...
Did a bit of research and as i found, you usually have a trade off between Core and Memory speed. You get to a point where you cant continue to push both, you have to hold one back while pushing the other. So I decided to throttle back the memory just a little to 1150MHz and then got the core to 960MHz @1.16v. With my sights still set on 1GHz
And so the benchmarking began, just to see what performance increase you do get with a decent OC.
My system (as is in my sig)
CPU: i5 750 @ 4GHz
Mobo: Asus P7P55D
G Card: XFX 5850 BE
RAM: 2 x 2 GB DDR3 1333 Kingston @ 8-8-8-22
Operating system: Windows 7 (build 7600) 64bit
Heaven Settings:
Render: direct3d11
Mode: 1920x1200 fullscreen
Shaders: high
Textures: high
Filter: trilinear
Anisotropy: 4x
Occlusion: enabled
Refraction: enabled
Volumetric: enabled
Replication: disabled
Tessellation: normal
Nice image with Tessellation on image (fairly big at 2.4MB)
So i decided to do 3 runs at each of the OCs and take the average to compare to the reference.
Ref: 725/1000 @ 1.087v (never lower the voltage, saw no need to)
OC 1: 765/1125 @ 1.087v
OC 2: 960/1150 @ 1.161v
OC 3: 1000/1125 @1.21v
Measurements were:
Min frame rate
Max frame rate
Avg frame rate
Score
Temperature and fan speed was set using MSI afterburner as so image
Results:
Ref:
Min: 12
Max: 73.7
Avg: 28.7
Score: 721.6
51 C @ 37% Fan Speed
OC1:
Min: 13.1
Max: 75.7
Avg: 30
Score: 756
52 C @ 40% Fan Speed
Average of 5.2% improvement over reference. (8.9% improvement in Min)
OC2:
Min: 16.1
Max: 90.7
Avg: 35
Score: 883.3
55 C @ 51% Fan Speed
Average of 25.5% improvement over reference. (34% improvement in Min)
OC3:
Min: 17.1
Max: 91.47
Avg: 35.4
Score: 892.7
57 C at 57 % Fan Speed
Average of 28.4% improvement over reference. (42% improvement in Min)
Then used the above to calculate the improvement per measurement, then averaged that. So we are able to see a very nice improvement of 19% between OC1 and OC2 which is about a 25% increase in Clock speed, and a 22% improvement between OC1 and OC3 which is a 31% increase in clock speed.
Looking at the price difference between it (R3400) and a reference card(R3000) was R400 which is an increase of 13%, you get a performance increase of 5% (non OC). It remains to be seen how well a reference OC's.
So I definitely think it is worth it if OC'ing. It's a very good improvement especially when you consider the rise in min frame rates which have more of an effect on games than the max frame rates.
So anyone want to lend me a 5870 or 5970 so i can compare
Last edited: