[SOLVED] i5 6600k overclocking tempteratures with an h100i

Nov 17, 2020
2
0
10
Hello,

I am running an i5 6600k with an h100i cooler. The radiator fans are set to intake, while I have two fans set to exhaust.
The multiplier is set at 45, and the vcore set at 1.3v.
Running a 30 minute realbench stress test, I'm getting up around and slightly over 80C.
Is this in the expected temperature range for a properly functioning h100i in this this configuration?

Any insight/advice would be greatly appreciated!
 
Solution
That is not the proper way to test thermal compliance. Realbench is for testing STABILITY, not thermal compliance. This is how you test thermal compliance.

And anything above 80°C while testing thermal compliance would be considered "undesirable".

That is not the proper way to test thermal compliance. Realbench is for testing STABILITY, not thermal compliance. This is how you test thermal compliance.

And anything above 80°C while testing thermal compliance would be considered "undesirable".

 
Solution

last_intel_guy

Reputable
Nov 15, 2020
19
4
4,525
That is not the proper way to test thermal compliance. Realbench is for testing STABILITY, not thermal compliance. This is how you test thermal compliance.

And anything above 80°C while testing thermal compliance would be considered "undesirable".
platforms.3661036/[/URL]


Agree with Darkbreeze partially here. One suggestion I might note is to follow this guide I've found very helpful from a reddit post long ago.
The key is starting with an AVX test so you understand your temp limits AND stability in one starting test, then test each in depth. Also not wasting way too much time if you don't want to run an hour long semi-informative stress test :) I can't remember what cpu this was for but I would look elsewhere for reference voltage/frequency:

  1. OCCT small sets WITH avx (1.250vcore and above high temp results) - 15 to 20 minutes or until temps get too high. (MONITOR CPU ENTIRE TIME) If this passes and does not BSOD or freeze, then continue testing. This works real good for a quick test to see if you should even waste time testing this new setting. If the voltage is not enough you will typically know in under 10 minutes when your comp goes down. If it's OK, then I move on into real testing.
  2. Realbench - 1 hour (can skip to test 3 first if you want to test your AVX offset)
  3. OCCT small sets no avx - 30 to 60 minutes (depending on temps)
  4. OCCT large sets no avx - 6 to 8 hours
  5. Prime95 small sets no avx - 30 to 60 minutes (depending on temps) (where I have cleared so far currently at 4.4 ghz 1.210v)
  6. Prime95 Custom blended with no avx - 6 to 8 hours
  7. Aida64 - 6 to 8 hours
For the first test, if the program shows errors then it's unstable. If it shows overheating but no errors then the voltage is too high. Starting with low voltage and seeing what frequency you can hit at lowest possible will safely find your stress level voltage limit from temps.

Edited: per Darkbreeze's better logic regarding temperature and stress testing as independent factors. His guide gives much more context than above, so reccomend any novice start with his steps linked (RAM testing, then temp test, etc.)
 
Last edited:
  • Like
Reactions: Ballybaba
Wow thats alot of stress testing, i agree its good to test stability but you will see cpu/cache stability usually results in bluescreen in gaming (certain bsod codes help to figure out what setting to tweak) or small performance decrease / getting same performance than less frequency.

I agree longer stress testing when you overclock ram, cause i had to clean install few times when using slightly unstable ram timings.
 
  • Like
Reactions: Ballybaba

last_intel_guy

Reputable
Nov 15, 2020
19
4
4,525
Wow thats alot of stress testing, i agree its good to test stability but you will see cpu/cache stability usually results in bluescreen in gaming (certain bsod codes help to figure out what setting to tweak) or small performance decrease / getting same performance than less frequency.

It's really the only way to stress test if you want to make sure your overclock has absolutely zero lifespan impact on the CPU. If you keep it under the manageable temps in the temperature limit tests, then you know for a fact your CPU cannot get up to higher temps otherwise. Games are starting to use AVX now, especially with directx12 support, so imo if you care about something totally stable its really the only way, especially on next gen games.

While lighter stress tests are an option, this way doesn't take long since you figure it out based on step 1 usually and will 90% of the time be stable on the rest. Definitely don't recommend gaming alone as a "stress test" since gaming is only going to hit your spike and full load on rare occasion, and what are you going to do next? Tweak something and wait 100 days for the next crash to assume it's off? Basically the above tests guarantee with 24/7 uptime you will not do any damage to your CPU over the lifespan it would have had if you left it at stock settings. The golden goal of stable overclocking :) I've also found out some really important things from my own, like improper thermal paste application and some bios settings that have greatly lowered my temps.

For example, I just failed pushing my 9700k on test 5 for Prime95, so now realizing I will need to up my voltage for my 4.9 ghz overclock, and increase my AVX offset to compensate for the temps. That's all the tweaking you really need to pay attention to if you run this set. My mobo sucks so I can't hit the usual 5.0 ghz that most people get.

I agree longer stress testing when you overclock ram, cause i had to clean install few times when using slightly unstable ram timings.

Yeah, I did run my RAM at its manufacturer XMP 3200 mhz which is probably pushing my CPU a lot also :). But anyone you ask will never recommend to not at least use the auto XMP from the manufacturer, I don't do any RAM timing tweaks besides that.
 
Last edited:
  • Like
Reactions: Ballybaba

last_intel_guy

Reputable
Nov 15, 2020
19
4
4,525
Also to add, Darkbreeze's standard guide is good but may be slightly out of date. Two or three years ago much less people would care about AVX loads, so in their tests they just straight up disable it and check their temps without. So I wouldn't necessarily follow that guide completely (even though Step 5 is basically that link's test). Doing an AVX offset is more and more necessary even with great coolers, since they just fall into their own category of heat on another level.

Also -- DON'T AVX test on Prime95. I have a feeling there is a strong reason the person who posted that stress test guide avoids any AVX testing in that program. It is just the notorious one it seems, OCCT (max stress/heat) and Realbench will serve much better.
 
  • Like
Reactions: Ballybaba
That "stress test guide" ISN'T a "stress test guide". It's a thermal compliance guide. Stress testing and thermal compliance testing are ENTIRELY different things, and the sooner people begin understanding this, the better off we'll be across these communities in general. But of course, that won't happen. It's "easy" for people to believe that they are interchangeable terminology, even though they are 100% NOT interchangeable terminology or ideology.

Two. Different. Things.

Stability testing requires a wide range of conditions, loads and instructions, from low and near idle to full load, fluctuating and everything in between. Even beyond TDP spec in some cases.

Thermal compliance testing requires STEADY STATE non-fluctuating workloads, per every major data sheet, and should be conducted using a utility that adheres to steady state workload environment.

If you want to see my FULL testing regimen, not the quick and easy version, then you can feel free to look here:


If you think you have a superior or better refined process, you are of course welcome to use it at your own discretion. I happen to know that mine is appropriate for the purposes to which it is intended.
 
  • Like
Reactions: lga1156_ftw

last_intel_guy

Reputable
Nov 15, 2020
19
4
4,525
That "stress test guide" ISN'T a "stress test guide". It's a thermal compliance guide. Stress testing and thermal compliance testing are ENTIRELY different things, and the sooner people begin understanding this, the better off we'll be across these communities in general. But of course, that won't happen. It's "easy" for people to believe that they are interchangeable terminology, even though they are 100% NOT interchangeable terminology or ideology.

Apologies if my 2-sentence stress test guide without any context was curt. Your guide (now realizing the author) is actually the best context I've seen on a beginner intro overclock with my moderate research over a few years on/off so far. So many guides just say to plug in numbers and don't clarify at all what stress test/temps will actually be safe, or really why the test order matters.

Also thanks to reading your guide, I now know my issue has probably been one that nearly no guides cover: Running RAM at manufacturer XMP profile. Ironic how so many guides tell you to just enable from the start and not mention the very real possibility of instability. Been stuck at Prime95 small tests going unstable in around 15 mins and getting dump files, so I think I've found the culprit (running 16GBx2 dual channel 3200 Mhz rated RAM at 16-18-18-38). Def reccomend OP start here, especially if using a cheaper mobo/PSU!

Thanks Darkbreeze for the guide, going to switch to your test format now going forward!
 
  • Like
Reactions: Darkbreeze

last_intel_guy

Reputable
Nov 15, 2020
19
4
4,525
Sorry -- last question hijacking this thread I promise.

Darkbreeze, why are we not recommending an AVX-based test to start for temperature limits again? Noticed your guide briefly mentions AVX but definitely more relevant in 2020 and only gaining as much going forward.

If you have a recommendation for how/how not to conduct an AVX based temp test (as I will likely have some need of AVX for my programs), any advice there would be appreciated. Nervous running Prime95 without AVX as a starter to feel good about my temps. Is the idea that you stay under a certain threshold when Realbench runs since it uses AVX? Or is it because you can gauge a certain temp threshold under AVX based on the Small FFT test that your guide starts with, without unnecessarily cooking the CPU?
 
Because 90% of users won't use AVX applications and the games they DO play that use any sort of AVX will be far less taxing than even the basic AVX package in Prime, much less AVX 2 or AVX 512.

You need to know that you are thermally compliant at the baseline, first. That means no AVX. Then, if you KNOW you WILL USE AVX based applications that will be significantly more taxing you can then further tune the configuration by running additional instances of Prime95 WITH the appropriate type of AVX enabled and an offset in the BIOS to account for the difference. Find the sweet spot so to speak. There are some very good AVX specific tuning guides out there for anybody who is specifically needing that.
 
Sorry i came accross as rude, was tired. Also im used to overclocking nehalem and westmere / bloomfield, these are pre-avx so i dont need to worry about that.
Also these guides are realy good for modern cpu overclocking. Im just using ancient stuff and rarely oc much with multiplier only blck

I usually run aida64 for 15min and cinebench bench, if its stable then game for hour or two. If i get no bluescreen i leave intel burn test with 90% memory usage running for one night, if it passes then next night i use prime95 blend. Thats about it.

After delidding ( only nehalem xeon, rest soldered, also 1366 always soldered ihs) im always below 85c even up to 1.5v vcore. This is with h150i corsair i have few of these .