I7-11700KF Temperatures and how to compare them with references

[MikeMS]

Hello!
I have Intel i7-11700KF on Motherboard ASUS ROG STRIX Z590-E GAMING WIFI with GamerStorm Lucifer V2 [DPGS-MCH6N-LCV2AM4] cooler on it, and also 2 intake and 1 on out fans in the case.
Overall the Processor works pretty stable even on heavy load and I'm happy with it. But hot temperatures sometime makes me worry about it's remaining lifetime...

So I want to compare my temperatures with some reference values under load, but can't find how to properly do that and in which software and where to find this "unified" temp values for the same load conditions.
Can someone help me with that and recommend something?

I recorded Stress test Bench in cpu-z with opened HWiNFO and:
In Benchmark 12.01.64 - it stays Ok, but in
Benchmark 12.01.64 AVX-512 (beta) - instantaneous thermal throttling!
Video files in the attach (download it for better quality and clearly see the values).
Yes it's synthetic tests, but keeping HWiNFO opened for a long time I sometime noted that it's going to throttle while I render in Blender if I use CPU for render and/or OpenImageDenoise - it quite heavy and often load CPU with often thermal throttling, and also in games sometime, but not much.
Is this reason to worry or this is Ok for it?
Thanks for attention!

[DeividA_Intel]

Hello MikeMS,  

Thank you for posting on the Intel® communities. I am sorry to know that you are having issues with your   

In order to better assist you, please provide the following:  

1. What is the BIOS version installed?

2. Have you made any changes to the BIOS recently?

3. Is this issue also present on idle or just when opening an app/game besides Blender?

4. Which setting did you use on the BETA version that did not use on the regular version that gives different results?

Regards,  

Deivid A. 

Intel Customer Support Technician 

[MikeMS]

Hello, Deivid A,

1. The BIOS version is 1601, but the same results was and on older version.

2. I set RAM XMP profiles, but do nothing with CPU boost, all it's settings on auto.
3. In idle temperatere is 45-50 degrees C. Which I think is Ok. I'm worried about temperatures on some CPU-heavy tasks.
4. BETA version of what? If you mean benchmark (12.01.64 AVX-512 (beta)) in cpu-z - it's just emulates heavy load on cpu... For example in Blender if I use OpenImageDenoise algorithm for Render denoise - it's going to throttling also, but after about 20-30 seconds...

but, maybe it's common behaviour for i7-11700kf...? Do you watch 2 attached vedeos?

It's would be interesting to compare how does it heat up through other users in the same test...
Thanks!

[dougaMAT]

Hello MikeMS
I'm using a translation site, so I'm sorry if it's hard to read.

This is a video of running Benchmark v19.01.64 AVX-512 (beta), thermal throttling is activated immediately after the start of the test, and then power limit throttling is activated to lower the CPU temperature. I would like to know what power limit your motherboard has set.

I have prepared a screenshot of HWiNFO displaying the information I want to know. The part surrounded by a red frame is the information you want to know.

For reference, I'll share my test results and thoughts.

I also own a Core i7-11700K and an ASUS TUF GAMING Z590-PLUS WIFI, so I ran the same test as you on an ADATA XPG LEVANTE 360 ARGB (max fan speed). The room temperature at the time of testing is 22°C.

Looking at the screenshot, you will notice that the CPU Package Power is lower than your processor. The power consumption of processors can be considered to be the same as the heat generation, so my processor generates less heat, but this is also affected by the difference in processor temperature.

Experimentally, the "high temperature test" is to stop the fan while running Benchmark v19.01.64 AVX-512 (beta) and raise the temperature. The power consumption, which was 198W at 70°C, has risen to 215W at 97°C. This is probably due to an increase in leakage current.

Anyway, unfortunately there is no unified standard for temperature comparison. The reviewers also verify the temperature according to their own standards, so if you want to refer to someone else's measurement results, you have no choice but to test under similar conditions.

But you and I have very similar processors. If you let me know what power limit your motherboard is setting, I can test accordingly.

I used to have the GamerStorm Lucifer V2 you have. It was a very reasonably priced and nicely sculpted heatsink, but it wasn't the best at cooling. If your processor is running Blender and thermal throttling kicks in after 20-30 seconds, it's a sign that your heatsink isn't good enough to handle the heat.

I'm in another thread asking ICS about the processor's upper temperature limit (TjMAX) setting. Basically, if TjMAX is set appropriately, I don't think the processor will break even if thermal throttling kicks in. I'll leave the formal opinion on this to ICS.

[MikeMS]

Hello DougaMAT! What a luck to have you here with the almost identical PC configuration!
Thank you for joining this discussion! Your information is pretty interesting for me!
Yes, your system looks much cooler than mine.
I checked the Power Limits that you mentioned - and they looks identical with yours. (I attached screenshot also)

What I noticed - that in idle most of your cores have clock rates in about 800Mhz, while mine is dancing around 4600-5000Mhz, mostly 5000Mhz. This leads to higher idle and test start temperatures (as well as test start clock rates and tdp)...
This is default (Auto) motherboard setting (while setting up memory XMP profiles it adjust bus clock also a bit, but as I understand - nothing changing with CPU rates...).

What surprises me - in AVX-512 test - that the temperatures rises to 100 C almost immediately (from ~50 to 100), at the same second. And I suppose it just can't so fast transfer the heat away from cores even to the cover of cpu.

[dougaMAT]

Hello MikeMS
Thanks for letting us know about the power limit settings. It is very interesting that power limit throttling was occurring even though the power limit was not limited. Maybe that's how it works when thermal throttling doesn't reduce power enough.

Since I had a look at your system, I added OC memory and GPU to mine, I applied XMP and set it to DDR4-3600, but the CPU core clock was lower at idle in my environment. My system is a test environment that I just built after reading your post, so the OS is clean. Your system may be affected by background applications.

If the processor temperature spikes as soon as a load is applied, I would suspect that you forgot to remove the baseplate film or did not install it correctly, but since your system can handle 170W of heat, it is probably not a problem.

The 200W+ heat generated in the AVX-512 test is difficult for air-cooled CPU coolers to handle. In my experience, to deal with 200W class heat generation, high-end air cooling or 240mm water cooling is preferable. However, I don't think applications such as Blender generate that much heat, so why not use CINEBENCH R23 (Multi Core) for load testing? Maybe the thermal throttling will be activated in about the same amount of time as when running Blender.

By the way, what kind of case are you using now? If you give me the details of your case, I can swap it out for a cooler that will fit in it and test it.

[MikeMS]

Hey, yes, interesting, I saw a huge amount of settings for CPU in bios for tweaking, but leave them on defaults (auto).
I think these CPU default settings on my MB (z590-e) slightly differentiate from your MB (z590-plus) to be a bit more aggressive to sudden loads...

The film was removed 100% )) But I am concerned about the thermal paste - when I bought Cooler I heard that the default paste that ships with it - is bad, but there were different opinions about it...
Also, a year and a half passed as the cooler was mounted. Maybe I will replace thermal paste later with some good quality, maybe.

As you recommended, I tried CinebenchR23.200 today, and was a bit surprised - it's quite soft on CPU load (compare to some other real workloads), during 2 tests - the CPU never goes to thermal throttling and the average CPU Cores temperature was around 86 C. (Screenshot at the end of test is attached and Video of the whole test)
I run Benchmark also without OBS recording and with a more preheated PC-case, but get almost the same results. Temperature just 1-2 deg higher, and test score also a bit higher - 14050. And without even HWiNFO running - 14157 points (not expected from HWiNFO such influence).

I'm using "Be Quiet Pure Base 500 Black"  case (+1 fan on intake).

Thanks!

[dougaMAT]

Thank you for verifying this. It is very, very interesting.
One thing I must say. I like to try and solve these problems. I feel no pain in trying things that may be off the mark, but not everyone does. So you don't have to force me to validate my suggestions.

Now, thanks to your recording on OBS, I have found one problem. As you may have noticed, I am experiencing the same power limit throttling as in the first video. (01:57~)

I have no idea why this power limit throttling is occurring. HWiNFO reports that the motherboard is not limiting the power limit, but at the same time the sensors are reporting that power limit throttling is occurring and the processor temperature and clock are dropping .

After much contemplation, I came up with the possibility that the motherboard's EPS12V (8-pin) + ATX12V (4-pin) are not fully connected, so I experimented with loading the CPU with only 8-pin or 4-pin connected, but the same phenomenon did not occur. (This is dangerous and should not be imitated).

One suggestion is to perform a CMOS clear and then run Load Default on the BIOS to verify the motherboard settings back to their full defaults. By doing this, you can see if the high idle clocks are due to the XMP settings or if the mysterious power limit throttling occurs even with full defaults. When being tested, displaying the Intel Extreme Tuning Utility (XTU), which is what I use, may help ICS determine the situation.

After running the test with the full default settings, when you apply XMP and save the BIOS, compare the settings that are changed with the screenshot I sent you (XMP settings). If any different items are changed, there may be a clue there.

Also, as for the high idle clock, please check the Power Option in Windows just to be sure. If it is set differently than the screenshot I sent you (Power Option), that could be the cause.

While common thermal grease will maintain some performance over time, dry grease will not do its job once it is stripped off. You may have recently installed a great video card, and it is possible that some of the thermal grease came off because it touched the CPU cooler during the installation process.

So I think the idea of trying a thermal grease upgrade is a good one. I don't know how effective it would be, but the results would help us make a decision to consider upgrading the CPU cooler, and the newly purchased thermal grease can be used the next time we replace the CPU or cooler.

Thanks for letting us know about the case you are using. With this case, I can install 240mm water cooling on the top or high-end air cooling. In the meantime, I can test the TUF GAMING LC 240 ARGB and the noctua NH-D15, which one are you interested in?

For reference, I can also send you the CINEBENCH R23 scores measured in my environment. I don't see much difference in the scores since the power limit throttling that occurred in your system is not that significant.

[MikeMS]

Hey, there is a lot to think about...) You say installing water cooling on top of the air cooling - is that possible? Wow) Never heard about that...
Thanks for your recommendations about upgrading the cooling system to a new one, I will keep it in mind, but in the near future I will try solutions without spending money, eh.))
I'm more worried about the lifetime of my CPU for a long time, and it's more question to Intel engineers - how they designed this CPU to bear these temp spikes sometime to 100 degrees in a long term of 5+ years...? I heard that more than 85 is already not good for them and shorten their life, but at other times I heard/read that modern chips are more capable to endure high temperature and working near the throttling limit - is Ok for them, and where is the truth - I don't know.

I think in a few months I will install m2 SSD - so this will be a nice reason to take off this huge radiator and change the grease, anyway I need to remove it, or the video card to be able to access to M2 slot which is almost fully covered beneath.

Nice idea with changing CPU power limits in the Windows Power Setting panel! Mine preset is - High Performance. I tried to change only "Minimum Processor state" to 5% - but, idle CPU clocks are still not changed, and that's strange. But then, I changed the whole preset to Balanced - and viola, CPU clocks in idle start to drop down to 800Mhz! I also installed Intel ExtremeTuningUtility and found the same setting there (yet not touched any other settings in IXTU). Thanks for that!)
Anyway, when I start AVX-512 test temps jump to 100C immediately with both presets (which is expected). I attach screenshots.

[dougaMAT]

The translated site will customize my opinion in an interesting way. Of course, the water-cooled cooler can be installed in the TOP Position of your case. But I have always had the idea of combining water-cooled and air-cooled coolers. If I connect the heatpipes and run water through them, I can achieve hybrid cooling!
Aside from that, I think it's good that you're trying a solution that doesn't cost a fortune. Even if it's a long shot, it's fun and I believe it will be a good experience for you to try and figure it out for yourself.

I understand that this is a question that you expect ICS to answer, and I can't answer about durability over 5 years, but I will note my understanding of the operating temperature. If I am wrong, @DeividA_Intel will correct me.

It is my understanding that modern processors have sophisticated thermal protection circuitry (TCC), and that it operates at Tjunction Max (TjMax), which is a sufficient margin from the temperature at which the CPU is damaged, to control power and heat generation in exchange for reduced CPU performance. Intel's view on these can be found at
https://www.intel.com/content/www/us/en/support/articles/000005597/processors.html

Your processor is running TCC after reaching 100°C, but it is maintaining a higher clock than the base clock and the CPU Package Power (Package TDP) is higher than the TDP of 125W, so I believe that the TCC is safely controlling the processor. It is my understanding that this is not a condition that will cause damage to the CPU in most cases, but please @DeividA_Intel for a formal answer.

If you select "High Performance" in the Power Setting of Windows, the power saving feature of the processor is disabled. So if you want to reduce the idle clock, I suggest you choose Balanced. Incidentally, the Active Power Plan you set in XTU is linked to the Windows Power Setting. Note that there are some dangerous settings in XTU, so please be careful not to change other settings until you understand what they are.

Thanks for the screenshot from the XTU test. I think the Package TDP during the test reflects the current capabilities of the CPU cooler. If you are concerned about reaching TjMAX, then limiting the power limit to around 170W may help. The power limit can be set in the BIOS or XTU, and I can show you how to do that if you want, but you may want to wait for ICS to respond.

[MikeMS]

... If I connect the heatpipes and run water through them, I can achieve hybrid cooling!

- Wow, that sounds like a very advanced approach!)

Thanks for your opinion about durability for a long time and that link - it contains the information that I'm looking for. And this sounds soothing.

Yep, I leave it now at the Balanced setting, I don't think that I noticed some performance drop.

Maybe I will try, just for interest, to lower the max power limit, just to see how it will be, but later. Also will replace the grease in a few months and will test how it helps or not, I will post when I do.

Thanks, I learned a lot of interesting and useful information.

[dougaMAT]

Thank you very much for your time here. It was a great learning experience for me to see some interesting cases.

I hope you get good results with your grease change.