[SOLVED] i5 7600k overclocking

[Alex1380]

Hey guys, couple of questions!



I am not a total rookie, but still could use a hand.

Currently adding to the overclock on my cpu as I have recently gotten some games which are CPU throttling,



My overclock is 5ghz currently with an AVX offset of 2.



I am currently at bios 1.355 v for my cpu boost (actual voltage about 1.360), (my last prime95 test failed at 1.350 after about 45 minutes).



Vcore according to hwm is 1.312



Now my thermals are fine, I am only hitting around 55-60 degrees when running prime95 for extended periods (and fans are a long way from being maxed) with peak tempts in the low 70s, so there is definitely room to push it a bit hotter, but I am hesitant to keep raising the vcore too much. I have seen other people push higher than this, but that was not necessarily for 24/7 overclocks.



Also when running Prime95, it is only 1 of the 4 threads which fails each time, and this thread always seems to be 1 or 2 steps ahead on the tests being run. I don't know if that is significant or means anything in particular.



So TLDR:



What is the max voltage I should be looking at?



Is it safe to increase voltage if my temps are consistently low?



Is pushing voltage too high still going to damage my PC even if temps are low?



Is it odd that 3/4 threads never fail and the same 1 does each time (although it is going longer and longer with each Vcore increase)



Thanks in advance for the help!

 

[geofelt]

How well you can oc a 7600K is determined by your luck in getting a good chip.
As of 6/19/17
What percent of samples can get an overclock
at a vcore around 1.4v.
I5-7600K
4.9 72%
5.0 52%
5.1 27%
5.2 16%
5.3 samples exist, unknown % of occurrence

I would say you are doing reasonably well.

Up to 1.4v should be ok. It is the high voltage that will damage chip, not temperature.
The chip will throttle or shut down if it detects a dangerous temperature. That is around 100c.
For stress testing, 85c should be your max.
I do not have a good stress tester that mimics what you actually do.

Cores are located on different parts of the chip so it is normal for temperatures to vary.
If the difference is very large, consider remounting your cooler, making certain it is seated evenly.

Ultimately, how much does that last multiplier buy you??
Back off a notch and do not worry.

 

[Alex1380]

Thanks a lot. I am fine with just about every stress test out there, Aida64, IBT, Real Bench, just not prime avx enabled.

If I disable AVX then prime also works fine although there is some debate whether that is a good idea or not, but as I am just playing games it is unlikely I will encounter an AVX workload from my understanding.

 

[geofelt]

AVX instructions put a heavy voltage load on the processor.
With a stress tester that uses AVX instructions, you will not be stable with a high multiplier.

As you say, for gaming, AVX presence should be minimal in games, but the possibility exists.
To counter that possibility AVX off set of 2 will lower the multiplier by 2 if such instructions are present.
Sounds to me like a good solution.

I also suggest implementing speedstep and adaptive voltage.
That will reduce the multiplier and vcore when the cpu has little to do.

 

[Alex1380]

Once I have confirmed the OC is stable I will definitely switch back to adaptive voltage.

Sorry for my ignorance but what is speedstep?

 

[geofelt]

Once I have confirmed the OC is stable I will definitely switch back to adaptive voltage.

Sorry for my ignorance but what is speedstep?

Speedstep allows software to reduce the multiplier.
In windows power management, you can give the processor a maximum of 100% under load
and a minimum of perhaps 20% when little cpu power is needed.