Hi 8Pack,
I recently purchased one of your I7 5960X CPUs binned at 4.7 GHz. The expectation from item description is that I should see 24/7 stability at 4.65 and bench stability at 4.75 Ghz with custom cooling.
Since I don't know what voltages you used when binning this CPU, I'd appreciate it if you could take a look at the numbers I generated along with the temperatures from my custom loop and comment on whether I'm hitting the performance you'd expect from this CPU at the voltages and temperatures you'd expect.
First, I assume that the synthetic "non real-world" stress tests such as AIDA64 Extreme and Prime95 Small FFT (Especially the versions with AVX instructions implemented) are not representative since at speeds in the 4.65 - 4.75 GHz range, these synthetics are going to shoot straight up to 105°C and throttling will occur (or the system will freeze/BSOD or threads will drop out). So this leaves more "real world" benchmarks like the AIDA64 Extreme benchmarks (not the stress test, but the benchmarks), Cinebench 11.5 and R15 rendering, and the RealBench H.264 Handbrake rendering benchmarks. Finally, I wasnt pushing the cache or memory to the edge so as to not complicate the stability picture and stuck to adjusting Vcore, Vinput, and LLC level and didn't try to get cute with Vccsa or other voltages/settings.
So, here's what I got:
4.65 GHz (46 x 101.1)
(Looking for 24/7 stability)
VCore = 1.325 Volts
Vinput = 1.95 Volts
LLC = Level 6
Everything else set to Auto
RealBench H.264 Handbrake
180 minute loop - stable
Max single core temperature: 79°C
4.751 GHz (47 x 101.1)
(Looking for bench stability)
Vcore = 1.45 Volts
Vinput = 2.0 Volts
LLC = Level 6
Everything else set to Auto
AIDA64 Extreme Benchmarks
Full list - stable
Max single core temperature: 104°C (ouch)
Cinebench 11.5
CPU Render 10X one after another - Stable
Max single core temperature: 94°C
Cinebench R15
CPU Render 10X one after another - Stable
Max single core temperature: 92°C
Intel XTU
CPU Stress Test
10 minutes - Pass
Max single core temperature: 98°C
RealBench H.264 Handbrake
10 minute loop - stable
Max single core temperature: 94°C
And that's it. I've got to run more benchmarks and stability tests to for the 24/7 stability at 4.65 GHz, but the 180 minutes of RealBench handbrake is a good start here. At 4.75 GHz, passing all the tests on a 10 minute loop with no failures should be sufficient to demonstrate bench stability.
Questions:
1. Is my Vcore at 4.65 GHz of 1.325 volts about what you would expect/want for a 24/7 stable system?
2. It took a 0.125 volt Vcore increase to step up from 4.65 to 4.75 GHz (1.325 volts at 4.65 GHz to 1.45 volts at 4.75 GHZ). Is this reasonable? Some of the more aggressive benchmarks and tests pushed 105°C. I wouldn't ever want to try to run like this 24/7 since cooked CPUs are costly, but for short benching sessions, does the Vcore of 1.45 volts seem acceptable/reasonable at 4.75 GHz?
3. What other benchmarks/tests would you recommend for further stability testing at 4.65 and 4.75 GHz?
Overall, a fine CPU you picked out for me!
I recently purchased one of your I7 5960X CPUs binned at 4.7 GHz. The expectation from item description is that I should see 24/7 stability at 4.65 and bench stability at 4.75 Ghz with custom cooling.
Since I don't know what voltages you used when binning this CPU, I'd appreciate it if you could take a look at the numbers I generated along with the temperatures from my custom loop and comment on whether I'm hitting the performance you'd expect from this CPU at the voltages and temperatures you'd expect.
First, I assume that the synthetic "non real-world" stress tests such as AIDA64 Extreme and Prime95 Small FFT (Especially the versions with AVX instructions implemented) are not representative since at speeds in the 4.65 - 4.75 GHz range, these synthetics are going to shoot straight up to 105°C and throttling will occur (or the system will freeze/BSOD or threads will drop out). So this leaves more "real world" benchmarks like the AIDA64 Extreme benchmarks (not the stress test, but the benchmarks), Cinebench 11.5 and R15 rendering, and the RealBench H.264 Handbrake rendering benchmarks. Finally, I wasnt pushing the cache or memory to the edge so as to not complicate the stability picture and stuck to adjusting Vcore, Vinput, and LLC level and didn't try to get cute with Vccsa or other voltages/settings.
So, here's what I got:
4.65 GHz (46 x 101.1)
(Looking for 24/7 stability)
VCore = 1.325 Volts
Vinput = 1.95 Volts
LLC = Level 6
Everything else set to Auto
RealBench H.264 Handbrake
180 minute loop - stable
Max single core temperature: 79°C
4.751 GHz (47 x 101.1)
(Looking for bench stability)
Vcore = 1.45 Volts
Vinput = 2.0 Volts
LLC = Level 6
Everything else set to Auto
AIDA64 Extreme Benchmarks
Full list - stable
Max single core temperature: 104°C (ouch)
Cinebench 11.5
CPU Render 10X one after another - Stable
Max single core temperature: 94°C
Cinebench R15
CPU Render 10X one after another - Stable
Max single core temperature: 92°C
Intel XTU
CPU Stress Test
10 minutes - Pass
Max single core temperature: 98°C
RealBench H.264 Handbrake
10 minute loop - stable
Max single core temperature: 94°C
And that's it. I've got to run more benchmarks and stability tests to for the 24/7 stability at 4.65 GHz, but the 180 minutes of RealBench handbrake is a good start here. At 4.75 GHz, passing all the tests on a 10 minute loop with no failures should be sufficient to demonstrate bench stability.
Questions:
1. Is my Vcore at 4.65 GHz of 1.325 volts about what you would expect/want for a 24/7 stable system?
2. It took a 0.125 volt Vcore increase to step up from 4.65 to 4.75 GHz (1.325 volts at 4.65 GHz to 1.45 volts at 4.75 GHZ). Is this reasonable? Some of the more aggressive benchmarks and tests pushed 105°C. I wouldn't ever want to try to run like this 24/7 since cooked CPUs are costly, but for short benching sessions, does the Vcore of 1.45 volts seem acceptable/reasonable at 4.75 GHz?
3. What other benchmarks/tests would you recommend for further stability testing at 4.65 and 4.75 GHz?
Overall, a fine CPU you picked out for me!