Our Modern Warfare 2 benchmark shows what appears to be a CPU bottleneck for the current overclocked configuration at resolutions below 2560x1600. The problem with that explanation is that the former build was clocked slower, and our previous tests have shown that the game cannot take advantage of the former build’s extra CPU cores.
To get to the bottom of this performance problem, we first looked at our June $2000 PC, which used similar cards and a similar CPU at similar speeds to the current build. Once we had a frame-of-reference, we tried various tuning tricks to reach the previous performance level.
We found that running uncore at its lowest compatible setting (twice the DRAM data rate) doubled our performance loss, while increasing memory speed and/or timings eliminated the loss. This means that running a lower uncore ratio in order to stabilize an overclock isn’t always a great move from a performance standpoint, and that memory bottlenecks do exist with some games.
We also found that DDR3-1134 CAS 6 netted similar performance to DDR3-1512 CAS 8, indicating that the bottleneck in this game is memory response time, rather than bandwidth (latency/frequency * 1000 = nanoseconds). Yet, we had to use completely different memory just to prove the insufficiency of the stuff we bought, and that’s something most builders can’t do.
- Luxurious Performance?
- Motherboard And Graphics
- CPU And Memory
- CPU Cooling and Case
- Power And Optical Drive
- Hardware Installation
- Test Settings
- Benchmark Results: 3DMark And PCMark
- Bechmark Results: SiSoftware Sandra
- Benchmark Results: Call Of Duty: Modern Warfare 2
- Benchmark Results: Crysis
- Benchmark Results: DiRT 2
- Benchmark Results: S.T.A.L.K.E.R.: Call Of Pripyat
- Benchmark Results: Audio And Video Encoding
- Benchmark Results: Productivity
- Energy, Efficiency, And Heat
- Value Conclusion