"Stable" overclock, issues
"Stable" overclock, issues
azca :
FPSUsername :
azca :
Consider setting it to 4Ghz, it should boost to 4.4Ghz.
Should I increase all cores to 4.4, or adjust it as follows: 4.3 on core 2 and 4.2 on core 3?
Try starting with a base of 4.0 and turbo up to 4.4 for all cores.
CPU: OCCT usually crashes immediately, especially on core 3 when in turbo mode (auto turbo).
I noticed in CPU-Z it draws around 1.38V at some point, which suggests OCCT might be drawing more voltage than permitted.
I previously configured Ring voltage to auto, cache to auto (35), and cpu offset voltage to +0.175.
It seems stable at 4.4GHz with 1.275V should work well.
Reference: http://imgur.com/a/NlBPp
CPU Max temperatures during that test (using core temp):
Core 0: 70
Core 1: 68
Core 2: 69
Core 3: 65
It seems the chip isn't performing well, so consider reverting to stock clocks and begin testing stability and a 100MHz step-up from there. I understand it didn't run Windows at 4.5GHz with 1.2V as initially suggested for a dirty overclock. Should I attempt:
Stock speeds, auto voltage,
Benchmark with aida64 and verify current voltage.
Manually adjust voltage
increase by 100MHz each time
Benchmark
If unsuccessful -> +0.025V
Retry
I also have the same Mobo and an i5 4670k, which is quite similar to the 4690k. Don't forget to skip the offset voltage and assume it's the maximum CPU voltage. For instance, if I set a normal vcore of 1.1 plus .11, it would result in around 1.21. Under stress it jumps to 1.25, which is quite high.
Start with manual vcore without offset and test your maximum GHz under stress for hours.
Evil Joker shares his experience with the same Mobo and an i5 4670k, which closely matches the 4690k. He suggests avoiding offset voltage and assuming the highest CPU voltage, noting that running a normal vcore of 1.1 plus .11 gives around 1.21, which rises to 1.25 under stress. He recommends using manual vcore without offset and conducting a stress test for hours. He provides a link to a diagram and plans to try adjusting to offset later for power-saving features.
FPSUsername :
Evil Joker :
I own the same Mobo and an i5 4670k, which is quite close to the 4690k. Just avoid using offset voltage and assume the highest CPU voltage will be used. For instance, running a normal vcore of 1.1 plus .11 gives me about 1.21, which is quite high under stress—around 1.25. That’s a significant jump.
Start with manual vcore instead of offset and test your maximum GHz under stress for hours...
http://i.imgur.com/Cdrj2pW.png
Once you’ve identified your best manual setting, adjust it to offset if you need power efficiency.
Make sure turbo mode is turned on.
Begin with manual 1.2V using the CPU, then enable turbo frequency to 42 and begin the stress test. I’ll share a screenshot later.
I’ll try 4GHz @1.2V with turbo disabled and Windows power settings in performance mode, then benchmark tomorrow.
It appears stable at 4.2GHz with 1.2v (1.10h test); even 4.3 was tested for 10 minutes. At 4.4GHz, I adjusted the voltage to 1.225V and OCCT showed 1.25V. A 10-minute test was also conducted (should be longer). 4.5GHz at 1.275V worked without issues, but CSGO crashed after 10 minutes. I’ll return to 4.2 tomorrow and test 4.3 then.
It seems you resolved the problem by changing the ram boot mode to auto.
CPU: 4.4GHz @1.25v
Ram: 2133MHz @1.56v, cl 10-9-10-24