Need help with OC

[manofsteel]

Hi

 

So i am currently OC'ig my CPU and i run prime 95 small FFT test on my CPU i5 6600k OC'ed to 4,1 Ghz with core votlage set to 1,250 V and i see that my temps are high like 82 C on hottest core after some time . Is it cuz of prime95 load? because when i run AIDA 64 temps arent that high and the core voltage could even less (it passed on 1.220 V).

 

Current system:

CPU i5 6600K

CPU COOLER Be Quiet Dark Rock 3

Paste NT-H1 from noctua

MOBO MSI Z170 Gaming M3

PSU EVGA FULL modular 750 W gold +

RAM Corsair Vengence lpx 2800 16 gigs in 4x4 kit XMP enabled

 

I gonna add a quick thing here i accidently OC'ed Ring ratio at the begin (to 4.2) is it a bad thing and if i use lower voltage now Prime does error on 1 of my cores and stops the test so .

 

Thanks for help

[runit3]

P95 is going to run any chip hotter than nearly any other test besides IBT. There's plenty of good alternatives out there (XTU, AIDA, ASUS Realbench). IMO P95 is completely unnecessary for testing stability, and you will never encounter those temps in any real world application aside from fully leveraged F@H.

[Peakfire]

It's because of the small FFTs test. Do not run that, as it's really unnecessary. Use Blend instead or another stress test tool.

I suppose the ring ratio is the cache. If so, get it to 4GHz and increase the VCCIO to ~1.08v

 

[manofsteel]

4 hours ago, Peakfire said:

It's because of the small FFTs test. Do not run that, as it's really unnecessary. Use Blend instead or another stress test tool.

I suppose the ring ratio is the cache. If so, get it to 4GHz and increase the VCCIO to ~1.08v

 

well with asus realbench i got to 4.4 GHz OC now on 1.320 V per core and VCCIO set to 1.1 V ( on msi its I/O Voltage and VT voltage i think) and doing now extended test for long term stability for 3h or 4h (i hope its enough) and temps are 70 C at most so its good right?

[manofsteel]

5 hours ago, runit3 said:

P95 is going to run any chip hotter than nearly any other test besides IBT. There's plenty of good alternatives out there (XTU, AIDA, ASUS Realbench). IMO P95 is completely unnecessary for testing stability, and you will never encounter those temps in any real world application aside from fully leveraged F@H.

bump

 

[runit3]

1 hour ago, manofsteel said:

well with asus realbench i got to 4.4 GHz OC now on 1.320 V per core and VCCIO set to 1.1 V ( on msi its I/O Voltage and VT voltage i think) and doing now extended test for long term stability for 3h or 4h (i hope its enough) and temps are 70 C at most so its good right?

Those values are fine. You have another 10-15C of thermal overhead. I'd give it 1.4v and try 4.7, then notch down. DRP3 should be able to keep up.

[manofsteel]

Just now, runit3 said:

Those values are fine. You have another 10-15C of thermal overhead. I'd give it 1.4v and try 4.7, then notch down. DRP3 should be able to keep up.

what i read and heard and watch the 1.4 V should be avoided if possible but i propably could do 4.5 with ease but my goal was 4.4 sine previous chip on turbo boost had 3.4 GHz so overall its near 1/3 of improved in frequency or even more . still its a good chip what with air cooling it can handle 4.4 Ghz and go more  i guess. And does 3 to 4 h stability test enough to call it a day?

[runit3]

Just now, manofsteel said:

what i read and heard and watch the 1.4 V should be avoided if possible but i propably could do 4.5 with ease but my goal was 4.4 sine previous chip on turbo boost had 3.4 GHz so overall its near 1/3 of improved in frequency or even more . still its a good chip what with air cooling it can handle 4.4 Ghz and go more  i guess. And does 3 to 4 h stability test enough to call it a day?

No, not at all. One test for 4 hours is rarely sufficient. That being said, you're at a pretty standard voltage for that low of a clock.

 

If you wanted to be thorough XTU or AIDA overnight (~8hr), 5-10 passes of Realbench, run some games for a few hours, watch youtube/Netflix/Hulu, and do some normal day to day tasks. You find stability with variety. By testing with all of your common applications, and the extreme ones, you're dialing in CPU stability for nearly all compute loads that it would ever encounter.

[manofsteel]

1 minute ago, runit3 said:

No, not at all. One test for 4 hours is rarely sufficient. That being said, you're at a pretty standard voltage for that low of a clock.

 

If you wanted to be thorough XTU or AIDA overnight (~8hr), 5-10 passes of Realbench, run some games for a few hours, watch youtube/Netflix/Hulu, and do some normal day to day tasks. You find stability with variety. By testing with all of your common applications, and the extreme ones, you're dialing in CPU stability for nearly all compute loads that it would ever encounter.

in other words i should just wait for end of test play some games while watching yt for example do some shit on background and leave the computer overnight on aida stress test right? i can do this for most of the compute i propably wont do much on CPU department (thats why it is i5 since gaming and some basis like watchin movies or smth like that) but on GPU will in the future i think. Back to values i got crash on extended realbench stress test with 8GB ram option on when CV was 1.300 so yea

[runit3]

Just now, manofsteel said:

in other words i should just wait for end of test play some games while watching yt for example do some shit on background and leave the computer overnight on aida stress test right? i can do this for most of the compute i propably wont do much on CPU department (thats why it is i5 since gaming and some basis like watchin movies or smth like that) but on GPU will in the future i think. Back to values i got crash on extended realbench stress test with 8GB ram option on when CV was 1.300 so yea

Realbench can be a realbitch, it's an excellent test for real world productivity testing. I've also been surprised by how many OC's of mine have gotten kicked by Hulu. These are after 6hr runs in XTU (I don't sleep much).

 

Skylake is not Haswell, you should be able to run up to 1.4v with no issue as long as you have decent cooling (which you do).

[manofsteel]

1 hour ago, runit3 said:

Realbench can be a realbitch, it's an excellent test for real world productivity testing. I've also been surprised by how many OC's of mine have gotten kicked by Hulu. These are after 6hr runs in XTU (I don't sleep much).

 

Skylake is not Haswell, you should be able to run up to 1.4v with no issue as long as you have decent cooling (which you do).

meh 4,4 GHz propably will be sufficient and at mosti  feel like i will get 4.6 for not long that beign said i thank you for advice and about temps i got 73 C on hottest core at the end . I just cant wait to see the action on game like B&S and how my CPU will react

[manofsteel]

3 hours ago, runit3 said:

Realbench can be a realbitch, it's an excellent test for real world productivity testing. I've also been surprised by how many OC's of mine have gotten kicked by Hulu. These are after 6hr runs in XTU (I don't sleep much).

 

Skylake is not Haswell, you should be able to run up to 1.4v with no issue as long as you have decent cooling (which you do).

Okay now there is another question that revealed after watchin Jaytwocents and it is temps of MB cuz my MB is msi z170 gaming m3 which is medium and there is smth like TMPIN3 and its max temp is 62 C degree it is good?

[runit3]

9 hours ago, manofsteel said:

Okay now there is another question that revealed after watchin Jaytwocents and it is temps of MB cuz my MB is msi z170 gaming m3 which is medium and there is smth like TMPIN3 and its max temp is 62 C degree it is good?

Well within spec if it's the VRM's or PCH

[manofsteel]

3 hours ago, runit3 said:

Well within spec if it's the VRM's or PCH

but is it good?

[runit3]

6 hours ago, manofsteel said:

but is it good?

Without know exactly what TMPIN3 is reading on your board there's no way to tell. Mine, for example, is a bogus sensor as "TMPIN3" within HWMonitor reads exactly 1:1 with the CPU temp readout. Most of the time TMPIN3 is VRM's or PCH, in either case sub 70C is perfectly fine, ~50-60C under load is optimal.

You have nothing to worry about.