Comparisons of CPU-GPU-RAM Overclocks and Comparing Bottlenecks (FM2 CPU)

[damric]

Here I am testing various platform overclocks and how well they scale while using one of my favorite games, Far Cry 2. It has a nice suite of built-in benchmarks, and today I'll be using the busy "Action Scene" to see how well CPU, GPU, and RAM overclocks can affect performance.

Platform:

Athlon II x4 760K
Asrock Extreme 6+ (BIOS 2.4)
2x4GB G.Skill RipjawsX DDR3-2133CL10
Radeon HD 7850 2GB
Monitor 60Hz 1080p Vsynch OFF
All in game detail quality settings set to maximum with 8x AA
Each test looped 3 times with the average recorded.

CPU-Z
GPU-Z
ASUS Voltage Tweak
Exact frequencies may vary slightly due to various bugs.

First let's see a run with all stock settings:

CPU at default/auto BIOS settings (3.8GHz stock with up to 4.0GHz Turbo enabled)
RAM at JEDEC settings (DDR3-1600CL11)
GPU at AMD stock settings (860MHz Core/1200MHz VRAM)

Results:
Average FPS: 38.77
Minimum FPS: 32.88
Maximum FPS: 49.89

Next I apply only a GPU overclock and test again:
CPU at default/auto BIOS settings (3.8GHz stock with up to 4.0GHz Turbo enabled)
RAM at JEDEC settings (DDR3-1600CL11)
GPU at 1200MHz Core 1450MHz VRAM (~40%OC)

Results:
Average FPS: 38.45
Minimum FPS: 31.78
Maximum FPS: 48.78

Next I apply only a CPU overclock and test again:
CPU at 4500 MHz (~18%OC)
RAM at JEDEC settings (DDR3-1600CL11)
GPU at AMD stock settings (860MHz Core/1200MHz VRAM)

Results:
Average FPS: 42.38
Minimum FPS: 36.32
Maximum FPS: 54.61

Next I apply both a CPU and GPU overclock, but no system memory overclock:
CPU at 4500 MHz (~18% OC)
RAM at JEDEC settings (DDR3-1600CL11)
GPU at 1200MHz (~40%OC) Core 1450MHz VRAM

Results:
Average FPS: 42.28
Minimum FPS: 35.93
Maximum FPS: 53.89

For the final test, I apply a full platform overclock:
CPU at 4500 MHz
RAM at DDR3-2400CL10 (~50%OC) with NB at 2600MHz (~44%OC)
GPU at 1200MHz (~40%OC) Core 1450MHz

Results:
Average FPS: 49.79
Minimum FPS: 41.00
Maximum FPS: 62.91

Analysis: Despite a few bugs, it is obvious that the GPU was totally bottlenecked by CPU and RAM, as adjusting GPU frequency resulted in no gains, while CPU and RAM overclocking brought substantial gains. Contrary to popular opinion, facts here indicate that overclocking RAM and memory controller can alleviate substantial bottleneck on the FM2+ platform, much more than just overclocking CPU and/or GPU.

 

[theFOoL]

in short: the faster the RAM the better FPS in games for APUs. System RAM acts as RAM for GPU for the APUs GPU. keep in mind that current APU 7850k is limited to 2133 RAM. go any higher and could have issues i think (unsure here hehe)

 

[damric]

in short: the faster the RAM the better FPS in games for APUs. System RAM acts as RAM for GPU for the APUs GPU. keep in mind that current APU 7850k is limited to 2133 RAM. go any higher and could have issues i think (unsure here hehe)

That is true, but this isn't an APU. This is a discrete CPU and GPU.

 

[emissary42]

Hi there, damric

How big of an impact had the higher NB-clock on the performance gains with the last config? (with memory @ DDR3-2400)

 

[damric]

Hi there, damric

How big of an impact had the higher NB-clock on the performance gains with the last config? (with memory @ DDR3-2400)

Due to the multiplier ratios higher NB was necessary or the DDR3 would underperform. Oddly enough, there wasn't much choice other than to take the NB up a couple of ticks in the BIOS. Seems to be doing alright, but not as good of memory performance as my FX-4100 had with DDR3-2000CL8 and 3000MHz NB.

 

[kn00tcn]

anything with a NB architecture will get a boost from the NB

FC2 was GPU limited in 2008, but not really on newer cards

my vote was for CPU in particular since the other option was memory, not NB/platform (& since it's an old athlon of course)

those numbers are disappointing though, you cant even do 60? i wonder how my q9550 3.6ghz compares... i have some benchmark runs from years ago on a 4870x2, but never tested at stock