Scores drop further compared to previous benchmarks after the significant overclock.
Scores drop further compared to previous benchmarks after the significant overclock.
So I'm a bit puzzled because my stable overclock seems to be yielding slightly poorer benchmark results now. Before the change: Intel 6600k at 3.5ghz (standard voltage) with an EVGA GTX 960 SSC running around 1292mhz. After the change: Intel 6600k at 4.5ghz (1.3V) with an EVGA GTX 960 in precision X mode at 110% power, clocking 1500mhz. No scores after overclocking. Catzilla: 8291, Firestrike: 6661, Cinebench R15: CPU (602cb) OpenGL (125.70 fps). After OC scores: Catzilla 8283, Firestrike 6641, Cinebench R15: CPU (697cb) OpenGL (151.50 fps). The benchmark numbers actually made sense, does it really matter who ran the tests first? Only notepad and CPUID HWMonitor were active. Temperatures stayed very low, never exceeding 50°C. Any questions about my build can be found in my signature.
It's completely feasible to achieve a stable overclock that still yields lower scores if the stability isn't consistent or if overheating occurs. This doesn't typically occur with AMD CPUs due to their high instability, whereas Intel CPUs are more capable of self-adjustment and will actually reduce their clock speed when overheating, even after being overclocked. This tendency is one reason why early Haswell reviews suggested it was a good overclocking experience, but in reality, the CPUs would downclock during testing.
It's completely feasible to achieve a stable overclock that appears solid, but if the stability isn't there it could lead to lower scores. This doesn't usually occur with AMD CPUs due to their high instability, whereas Intel CPUs tend to adjust automatically—underclocking themselves when overheating, even after being overclocked. This tendency explains why early Haswell reviews seemed to praise its performance, while in reality the CPUs were actually slowing down during testing. Nvidia GPUs behave similarly, particularly when overheating. I recommend using monitoring tools to track temperatures and clock speeds closely while stressing the system. I bet something is throttling, and if nothing slows down, it's likely the voltage is too low.
It's quite feasible to achieve a stable overclock that appears reliable, yet may result in lower scores if the stability isn't consistent or if overheating occurs. This tends to be more of an issue with AMD processors, which are known for their instability, whereas Intel chips tend to self-adjust by underclocking when overheating, even after being overclocked. This characteristic contributed to the perception in early Haswell reviews that it was a strong overclocker, but in reality, many CPUs would downclock during testing with unstable settings or when overheating. The rapid changes often went unnoticed until professional testers began focusing on high metrics and found no improvement.
Nvidia GPUs also exhibit similar behavior, particularly when overheating. I recommend using monitoring tools to track temperatures and clock speeds closely during stress tests. It seems possible that the system is throttling, suggesting a moderate undervoltage situation.
I suspect there might be an issue with the voltage settings, especially for the graphics card. The description notes that temperatures stayed below 50°C, indicating a decent cooling setup with multiple fans and a Corsair H60. However, I’m considering upgrading the GPU overclocking options, as Precision X only allows a 10% increase and I believe more power is needed. I plan to raise the CPU voltage as well.
i don't remember seeing anything suggesting the gtx960 was a strong overclocker. you might need to lower its overclock slightly. still, the most popular tool for gpu overclocking is msi afterburner.