Opinions on CPU temps

Hello everyone!

So today, I was doing some temp readings and I would like a second opinion on them.

Information:

Specs:
Gigabyte z97X gaming 5
4690K h100i

Overclocking Status
1.2 V
4.4 Ghz

Temps
AMBIENT TEMP: 23-24C

NOTE: temperatures are averages based on what I mainly saw.

Aida 64: 65C
Prime 95 (Small, max heat setting): 82C
OCCT: 60C
Intel BurnL 71C
Surfing web: 33C
I can take a gaming temp if needed.

To me, this seems a little high. I feel like I should be expecting better from the H100i.

Any opinions, criticisms and general inquiries are greatly appreciated.

Thanks!
 
Solution
Synthetic benchmarks are not recommended for Haswell and essentially serve no purpose other than accelerating setting your TIM. The older version of P95 doesn't include AVX and other extensions so it only tests part of your CPU's stability , the newer versions do but since AVX is present and the way it tests is via a constant hammering of AVX puts an unrealistic load on the CPU....AVX raise CPU voltage by 0.10 to 0.13 under adaptive settings .... and while applications that use AVX do this in short bursts with no harm whatsoever, P95's constant hammering at that boost voltage can be bad news for your CPU..... Same is true to a lesser extent w./ AIDA 64

Intel XTU is safe and realistic but as it hits the CPU in a serious of stress -...
Looks actually like the temps went down. I forgot to mention that I had just re-applied thermal paste (because I originally had higher temps).

I think the paste is being "broken in" if that is a thing.

Aida: 60
OCCT: 55
Intel burn: 65
Same web surf temp

These I am liking a bit better.
 
Synthetic benchmarks are not recommended for Haswell and essentially serve no purpose other than accelerating setting your TIM. The older version of P95 doesn't include AVX and other extensions so it only tests part of your CPU's stability , the newer versions do but since AVX is present and the way it tests is via a constant hammering of AVX puts an unrealistic load on the CPU....AVX raise CPU voltage by 0.10 to 0.13 under adaptive settings .... and while applications that use AVX do this in short bursts with no harm whatsoever, P95's constant hammering at that boost voltage can be bad news for your CPU..... Same is true to a lesser extent w./ AIDA 64

Intel XTU is safe and realistic but as it hits the CPU in a serious of stress - rest - stress - rest segments, it to gives inaccurate results. Unless you built your PC, to get your name on web overclocking leaderboards, P95, AIDA performance is not really of significance. What you want to be concerned about is how hot will it get running real applications. The best tool I have found so far for this is RoGs Real Bench which is a set of applications known for being the largest load on a CPU. It includes AVX but only in the manner used in real appli9cations, not artificial loading akin to testing a car's mpg by towing a 7,500 pound trailer up steep hills.

So what I would worry about is not how hot the CPU gets when doing imaginary simulated synthetic loads but how hot it might get in running any *real* application you might actually use. More importantly, you can be Aida, Prime 95, OCCT, Intel Burn, Intel XTU stable and still crash under RoG Bench. This is because the way the synthetics work is they throw a constant heavy bar5age of one thing for x minutes, then something else for y minutes.... RB throws several different applications at the CPU at the same time, each using various extensions / instruction sets in a multitasking environment.

So, as far as what to worry about, I wouldn't set my OC based upon "imaginary" loads. I'd test under RB and crank it up to 4.5 or 4.6 if you could stay under 75C in RB


As for "is that normal ?", at first glance the temps don't look that bad.... but it does seem kinda high fro 4.4 Ghz.... after all the 4790k safely does 4.4 Ghz with the Intel stick cooler and a 4790k w/ HT turned off is essentially a 4690k. If you keep the H100i down to tolerable noise levels (< 45dba), it's performance is not that impressive

b4.jpg


You didn't say if you were letting it go to max or not. Those are tests with AIDA 64 (Stress Test CPU & FPU) ...at full speed (68dbA), the H100i knocks off about 3C (73C CPU temp)..... so your 4.4 Ghz @ 65C compares well with theirs but:

1) I don't know whether you ran the same CPU & FPU test.
2) They used Haswell which runs significantly hotter than the Devils Canyon CPUs
3) 8 core i7s run hotter than 4 core i5's

at 4.4 Ghz,(39 cache, 1600 RAM speed) my 4770k cores averaged 57.75 in real bench.... ambient was 26C ... with 44 multiplier, 44 cache and 2400 RAM speed they averaged 59C

In short while a little higher than I'd expect, I wouldn't worry about the temps unless you're breaking 75C in an application based stress test.....with the AVX voltage boost of 0.13 volts, you don't want to overstress your CPU with synthetics, nor be overly conservative on your OC for fear of temps your system will never gain see.
 
Solution