[SOLVED] Stress Testing bringing CPU temps to 100C instantly

[Tomascari]

Hi guys,

So I've been doing some Prime95 stress testing to try and solve another problem I've been having. This has all been in the last 3 or 4 days. Initially, the CPU was maxing out at about 75C, occasionally up to 80C. I've done about 10 CPU stress tests ranging from 5 mins to about 45 mins (PC shuts itself down randomly which is the problem I'm trying to solve).

Then a few tests ago, the CPU package and core temps jumped to 100C instantly before dropping back to 75C for the rest of the test after about 20 seconds. And since that one, it jumps to 100C instantly and just stays there. But then it instantly drops back to about 35C once the test is finished.

Can several stress tests ruin thermal paste that quickly or something? It's a brand new PC with Noctua thermal paste.

For reference, my CPU is the Intel i5 11600K (not overclocked) with a Vetroo V5 air cooler. I have also verified the data across OCCT (also briefly running a stress test here with the same temp rise), HWinfo and HWMonitor... all showing the same jump to 100C and then immediate drop after testing.

Any ideas? Thanks

 

[hotaru.hino]

As a quick sanity check if you run the stress, try pushing the heat sink against the CPU. If the temperatures start dropping, then the heat sink wasn't tightened enough.

Otherwise stress testing does not burn out quality thermal grease after a few runs.

 

[dorsai]

Stress testing won't affect or ruin thermal paste.

If you're hitting 100c the chip is simply not getting cooled correctly...since it drops back to 35c rather quickly my guess is it's likely an airflow issue. If the thermal paste wasn't done right I would tend to think it wouldn't drop so quickly back to 35c. Then again you are getting crashing so maybe pull the heatsink off and re-apply the thermal paste according to the makers directions.

What's you case airflow situation like ?

 

[Roland Of Gilead]

I agree with the other posters! ^

It's very possible it's a cooler issue. Although your cooler is rated for 150w, your CPU has a TDP of 125w (at base clocks), it draws much more than 150w at boost speeds, and whilst stressing. In fact it can draw up to 251w (PL2). It's very possible the cooler just isn't adequate enough at high clocks/load.

Which Noctua paste are you using? H1/H2 have optimum operating temps between 70-120c, so that paste is good enough.

As @hotaru.hino and @dorsai suggested, check cooler is seated correctly. You will obviously have to re-paste. Noctua's instructions suggest best method is to put a pea sized blob in centre of CPU. 2-3mm in diameter. Then seat the cooler. If the cooler has 4 screws, then tighten them in a criss-cross pattern from one corner to the opposite diagonal. This will ensure as you tighten down, you get an even spread of paste.

 

[Tomascari]

As a quick sanity check if you run the stress, try pushing the heat sink against the CPU. If the temperatures start dropping, then the heat sink wasn't tightened enough.

Otherwise stress testing does not burn out quality thermal grease after a few runs.

Thanks for the suggestion - just tried and no difference made. CPU remained at 100C while presing the heatsink down.

I agree with the other posters! ^

It's very possible it's a cooler issue. Although your cooler is rated for 150w, your CPU has a TDP of 125w (at base clocks), it draws much more than 150w at boost speeds, and whilst stressing. In fact it can draw up to 251w (PL2). It's very possible the cooler just isn't adequate enough at high clocks/load.

Which Noctua paste are you using? H1/H2 have optimum operating temps between 70-120c, so that paste is good enough.

As @hotaru.hino suggested, check cooler is seated correctly. You will obviously have to re-paste. Noctua's instructions suggest best method is to put a pea sized blob in centre of CPU. 2-3mm in diameter. Then seat the cooler. If the cooler has 4 screws, then tighten them in a criss-cross pattern from one corner to the opposite diagonal. This will ensure as you tighten down, you get an even spread of paste.

Ahhh you've cracked it, thank you! Stupidly, I didn't check the power draw from the CPU and looking through the logs now I can see the 100C corresponds perfectly to it pulling about 180W. Then it was pulling 125W when it stayed around 75C.

So, next question: is it safe enough to let it just throttle itself at 100C and continue to run stress tests to solve my main issue or would it be better to disable turbo boost?

 

[Tomascari]

Stress testing won't affect or ruin thermal paste.

If you're hitting 100c the chip is simply not getting cooled correctly...since it drops back to 35c rather quickly my guess is it's likely an airflow issue. If the thermal paste wasn't done right I would tend to think it wouldn't drop so quickly back to 35c. Then again you are getting crashing so maybe pull the heatsink off and re-apply the thermal paste according to the makers directions.

What's you case airflow situation like ?

Thanks for the response and suggestions, keith hit the nail on the head though

 

[Roland Of Gilead]

Thanks for the suggestion - just tried and no difference made. CPU remained at 100C while presing the heatsink down.

Ahhh you've cracked it, thank you! Stupidly, I didn't check the power draw from the CPU and looking through the logs now I can see the 100C corresponds perfectly to it pulling about 180W. Then it was pulling 125W when it stayed around 75C.

So, next question: is it safe enough to let it just throttle itself at 100C and continue to run stress tests to solve my main issue or would it be better to disable turbo boost?

So, that's a tough one to call in terms of the shutdowns. Your CPU is designed to throttle at about 100c, but it will also shut down if there's much more hard use than the max temps. It will do this to protect itself. With that said, shut downs like that could indicate a PSU issue. What PSU do you have? If it's not up to scratch, they can cause random reboots, but typically at load. It may be related to your problems.

edit: To answer the question of is it okay to run it with no boost, yes, of course you can. But, you have a stellar CPU. I'd recommend you get a good cooler capable of dealing with your CPU at it's max boosts (but still keep temps below 80c - which is where you always wanna be at max load). Why leave all that performance on the table because of an inadequate cooler.

 

[Tomascari]

So, that's a tough one to call in terms of the shutdowns. Your CPU is designed to throttle at about 100c, but it will also shut down if there's much more hard use than the max temps. It will do this to protect itself. With that said, shut downs like that could indicate a PSU issue. What PSU do you have? If it's not up to scratch, they can cause random reboots, but typically at load. It may be related to your problems.

edit: To answer the question of is it okay to run it with no boost, yes, of course you can. But, you have a stellar CPU. I'd recommend you get a good cooler capable of dealing with your CPU at it's max boosts (but still keep temps below 80c - which is where you always wanna be at max load). Why leave all that performance on the table because of an inadequate cooler.

Yeah the problem I'm trying to solve is causing random restarts while gaming (with CPU not highly loaded and as low as 60C) but I've found that running CPU stress tests produces the exact same restart fairly quickly every time so I'm using it as a diagnostics tool. I have a Silverstone SX700-G PSU so has plenty of juice for my system. I've made another thread for this problem here: https://forums.tomshardware.com/thr...playing-games-and-cpu-stress-testing.3702102/

I've got a strange situation for cooling which I want to change soon anyway as I have it pulling air from the rear of the case with no dust filter (cooler fan wouldn't quite fit with the RAM sticks in). I built this PC as a mini-ITX in a Cooler Master NR200P so needed a cooler that would fit and this Vetroo one is the biggest I could get that fits (and had good reviews). I could install a watercooler but decided not to (largely for aesthetics in the case admittedly). Also, the CPU is better than what I need it for - I bought it as it was available at a good price and has a decent base clock speed.

 

[CompuTronix]

... Prime95 stress testing ... package and core temps jumped to 100C instantly ...

... i5 11600K (not overclocked) with a Vetroo V5 air cooler.

... verified the data across OCCT ... HWinfo and HWMonitor... all showing the same jump to 100C ...

Latest versions of Prime95 and OCCT?

Which test?

Were they run at default settings?

If so, do you have AVX Offset values configured in BIOS?

Without AVX Offsets, if you run the default AVX workloads in Prime95 and OCCT, they will drive your processor to its maximum power consumption and Core temperatures. As per Intel's Datasheet, TDP and Thermal Specifications are validated without AVX. - 11th Generation Intel® Core™ Processor Desktop Datasheet, Volume 1 - (Refer to page 87, Section 5.1.1, Thermal Management, 1st paragraph, 2nd sentence).

To get valid results, here's the correct tests and proper settings for checking thermal performance:

Prime95
Test - Small FFTs
AVX-512 - Disabled
AVX2 - Disabled
AVX - Disabled

OCCT
Test - CPU
Data Set - Small
Mode - Normal
Load Type - Steady
Instruction Set - SSE

When properly configured as shown above, Prime95 and OCCT will provide a steady-state non-AVX workload with steady Core temperatures within a few Watts and a degree or so of one another.

CT

 

[Roland Of Gilead]

Yeah the problem I'm trying to solve is causing random restarts while gaming (with CPU not highly loaded and as low as 60C) but I've found that running CPU stress tests produces the exact same restart fairly quickly every time so I'm using it as a diagnostics tool. I have a Silverstone SX700-G PSU so has plenty of juice for my system. I've made another thread for this problem here: https://forums.tomshardware.com/thr...playing-games-and-cpu-stress-testing.3702102/

I've got a strange situation for cooling which I want to change soon anyway as I have it pulling air from the rear of the case with no dust filter (cooler fan wouldn't quite fit with the RAM sticks in). I built this PC as a mini-ITX in a Cooler Master NR200P so needed a cooler that would fit and this Vetroo one is the biggest I could get that fits (and had good reviews). I could install a watercooler but decided not to (largely for aesthetics in the case admittedly). Also, the CPU is better than what I need it for - I bought it as it was available at a good price and has a decent base clock speed.

Your PSU seems decent by all the reviews I've read. However, although it provides plenty of power and is of a generally good build, that doesn't mean it's not failing. How old is it? It does only have a 3 year warranty which is unusual for a Gold rated PSU. Typically they are 5-7 years. @Lutfij made a good suggestion to you in the other thread. Test with another PSU, or bring to local repair store and get them to swap it out to be sure.

Ah, yes, an ITX build is small. I can understand the cooler fit. Might be worth looking into a decent SFF AIO cooler to tame the wattage I just hate the idea of letting that extra performance go to waste. In your case, if you are happy with the base clocks, and to keep temps in check then disbling boost is an option.

You could also look at Throttlestop as a means of undervolting, and reducing temps, whilst still keeping some performance above base.

 

[Zerk2012]

I agree with the other posters! ^

It's very possible it's a cooler issue. Although your cooler is rated for 150w, your CPU has a TDP of 125w (at base clocks), it draws much more than 150w at boost speeds, and whilst stressing. In fact it can draw up to 251w (PL2). It's very possible the cooler just isn't adequate enough at high clocks/load.

Which Noctua paste are you using? H1/H2 have optimum operating temps between 70-120c, so that paste is good enough.

As @hotaru.hino and @dorsai suggested, check cooler is seated correctly. You will obviously have to re-paste. Noctua's instructions suggest best method is to put a pea sized blob in centre of CPU. 2-3mm in diameter. Then seat the cooler. If the cooler has 4 screws, then tighten them in a criss-cross pattern from one corner to the opposite diagonal. This will ensure as you tighten down, you get an even spread of paste.

That PL2 is off by a good bit for the 11600K.

CPU	P1 State	PL2 State	Duration
Core i9 11900K	125 Watt	250 Watt	56 seconds
Core i7 11700K	125 Watt	229 Watt	56 seconds
Core i5 11600K	125 Watt	182 Watt	56 seconds

 

[Roland Of Gilead]

That PL2 is off by a good bit for the 11600K.

CPU	P1 State	PL2 State	Duration
Core i9 11900K	125 Watt	250 Watt	56 seconds
Core i7 11700K	125 Watt	229 Watt	56 seconds
Core i5 11600K	125 Watt	182 Watt	56 seconds

Good spot. Wasn't trying to mislead. Just looked at the wrong CPU! DOH

 

[Tomascari]

Your PSU seems decent by all the reviews I've read. However, although it provides plenty of power and is of a generally good build, that doesn't mean it's not failing. How old is it? It does only have a 3 year warranty which is unusual for a Gold rated PSU. Typically they are 5-7 years. @Lutfij made a good suggestion to you in the other thread. Test with another PSU, or bring to local repair store and get them to swap it out to be sure.

Ah, yes, an ITX build is small. I can understand the cooler fit. Might be worth looking into a decent SFF AIO cooler to tame the wattage I just hate the idea of letting that extra performance go to waste. In your case, if you are happy with the base clocks, and to keep temps in check then disbling boost is an option.

You could also look at Throttlestop as a means of undervolting, and reducing temps, whilst still keeping some performance above base.

The PSU is brand new as with all the other components. I didn't look into the warranty much when I selected it as SFX PSUs are hard to come by so again, it was more of an availability purchase. But as it's brand new, there shouldn't be any problems if it needs returning. I have a very old PSU I could try but I'm wary of that based on its age and it may have been the cause of problems I was having a few years ago. Not had to take anything to a local shop for repair before so I'll look into where could be trustworthy as another option too.

The thing that seems strange to me is that it doesn't fail under GPU stress tests which pull a higher wattage from the PSU. And as it fails when gaming with GPU at 80% load and CPU well under max load (maybe 30% load) and both at perfectly operational temperatures it makes me lean towards a MB problem. But hey, I have a few more tricks to try before admitting defeat and sending parts off.

Yeah I totally agree on wasting the performance if it is there to be had so once my issues are sorted I'll look into how I can cool the CPU more effectively!

 

[Tomascari]

Latest versions of Prime95 and OCCT?

Which test?

Were they run at default settings?

If so, do you have AVX Offset values configured in BIOS?

Without AVX Offsets, if you run the default AVX workloads in Prime95 and OCCT, they will drive your processor its maximum power consumption and Core temperatures. As per Intel's Datasheet, TDP and Thermal Specifications are validated without AVX. - 11th Generation Intel® Core™ Processor Desktop Datasheet, Volume 1 - (Refer to page 87, Section 5.1.1, Thermal Management, 1st paragraph, 2nd sentence).

To get valid results, here's the correct tests and proper settings for checking thermal performance:

Prime95
Test - Small FFTs
AVX-512 - Disabled
AVX2 - Disabled
AVX - Disabled

OCCT
Test - CPU
Data Set - Small
Mode - Normal
Load Type - Steady
Instruction Set - SSE

When properly configured as shown above, Prime95 and OCCT will provide a steady-state non-AVX workload with steady Core temperatures within a few Watts and a degree or so of one another.

CT

Yep, latest versions of Prime95 and OCCT. First couple of Prime95 tests were blended then I moved to small FFTs but other than that, all settings were default.

I haven't configured AVX offsets so thank you very much for this info, I'll set this up properly when I run my next tests! I'm quite a low-level user and am just using the stress tests as a form of diagnosis for my crashing issue in this thread: https://forums.tomshardware.com/thr...playing-games-and-cpu-stress-testing.3702102/. It's crashing and restarting with no BSOD while playing some games and the CPU stress tests produce the same crash. Thanks for your help!

 

[Roland Of Gilead]

The thing that seems strange to me is that it doesn't fail under GPU stress tests which pull a higher wattage from the PSU. And as it fails when gaming with GPU at 80% load and CPU well under max load (maybe 30% load) and both at perfectly operational temperatures it makes me lean towards a MB problem. But hey, I have a few more tricks to try before admitting defeat and sending parts off.

What GPU are you running in the rig?

 

[Tomascari]

What GPU are you running in the rig?

A Gigabyte RTX 3060 Eagle OC

 

[Roland Of Gilead]

Ah right. The RTX 3xxx series can have some issues with otherwise really good PSU's. Its down to the super fast frequency switching and also instant power draw increases of up to 100w when the card is boosting. Although RTX3070/3080/3090 seem to suffer with the issue a little more. It could very well be that.

 

[Roland Of Gilead]

Prime95
Test - Small FFTs
AVX-512 - Disabled
AVX2 - Disabled
AVX - Disabled

Great advice for those testing with Prime. With AVX on you could heat your house on the heat output from the CPU

with steady Core temperatures within a few Watts and a degree or so of one another.

This is also a good guide for anyone replacing thermal paste. If your cores are more than 1 or 2 degrees apart , then you need to re-paste again. It means the paste hasn't spread fully across the CPU when placing cooler back on.

 

[Tomascari]

Ah right. The RTX 3xxx series can have some issues with otherwise really good PSU's. Its down to the super fast frequency switching and also instant power draw increases of up to 100w when the card is boosting. Although RTX3070/3080/3090 seem to suffer with the issue a little more. It could very well be that.

I had thought of the GPU initially but I've largely dismissed it as a cause as I've run Unigine Superposition and FurMark twice for an hour with no crash at all (GPU temps never get above 70C either). Regardless, I'm going to be doing some isolation testing tomorrow and removing everything I can, including the GPU, to see if I can find anything that way. Also, the retailer has given me an RMA code to send the motherboard, PSU and CPU back next week if I get nowhere this weekend.

This is also a good guide for anyone replacing thermal paste. If your cores are more than 1 or 2 degrees apart , then you need to re-paste again. It means the paste hasn't spread fully across the CPU when placing cooler back on.

Ahhh thanks for pointing that out, I hadn't joined the dots together in my head. I have been getting a bit of variation in core temps so I'll take my cooler off tomorrow as well to see how the thermal paste has spread and reapply accordingly.

 

[Roland Of Gilead]

I had thought of the GPU initially but I've largely dismissed it as a cause as I've run Unigine Superposition and FurMark twice for an hour with no crash at all (GPU temps never get above 70C either). Regardless, I'm going to be doing some isolation testing tomorrow and removing everything I can, including the GPU, to see if I can find anything that way. Also, the retailer has given me an RMA code to send the motherboard, PSU and CPU back next week if I get nowhere this weekend.


Ahhh thanks for pointing that out, I hadn't joined the dots together in my head. I have been getting a bit of variation in core temps so I'll take my cooler off tomorrow as well to see how the thermal paste has spread and reapply accordingly.

What about in reverse? Have you another GPU you can swap out? If so, and system runs fine, then you could narrow it down to RTX/PSU combo.

Yeah, that variation in core temps is indicative of bad spread. I've really noticed this playing around with different pastes on my gaming laptop. Trial and error It's almost an obsession now! Ill be a thermal paste scientist in no time!

Keep us updated on your progress. Because the thread is closed, you might PM me and let me know how your getting on. I'm interested to get to the bottom of your issue.

 

[Tomascari]

What about in reverse? Have you another GPU you can swap out? If so, and system runs fine, then you could narrow it down to RTX/PSU combo.

Yeah, that variation in core temps is indicative of bad spread. I've really noticed this playing around with different pastes on my gaming laptop. Trial and error It's almost an obsession now! Ill be a thermal paste scientist in no time!

Keep us updated on your progress. Because the thread is closed, you might PM me and let me know how your getting on. I'm interested to get to the bottom of your issue.

Yeah I have an old GPU I can try if needs be. I'll see what happens when I remove the RTX from the system entirely and go from there.

Haha to be honest, I decided to build this PC not because I needed anything in particular but because I fancied a bit of a project so now I have the means to fiddle with things I may well get obsessed with thermal paste soon too!

Yeah, I will definitely get you posted! Thank you so much for your help so far, I've really appreciated having someone to throw ideas at me!

 

[CompuTronix]

When properly configured as shown above, Prime95 and OCCT will provide a steady-state non-AVX workload with steady Core temperatures within a few Watts and a degree or so of one another.

CT

This is also a good guide for anyone replacing thermal paste. If your cores are more than 1 or 2 degrees apart , then you need to re-paste again. It means the paste hasn't spread fully across the CPU when placing cooler back on.

keith12,

It appears that you've quoted my post out of context. If you read the entire sentence, the intention was to highlight how those specific workloads are so very similar between Prime95 and OCCT; not temperature differences between individual Cores.

But since you brought it up, Intel's specification for Digital Thermal Sensor (DTS) accuracy is +/- 5°C. This means that during a steady-state 100% TDP all-Core workload, If the highest and lowest Cores are within 10°C of one another, then the processor meets the specification for thermal accuracy. However, as you've pointed out, differences between the temperatures of each Core can also be compounded by a flawed application of TIM, either above the Die (internal) or above the IHS (external), or a combination of both, which can, as well, cause individual Core temperatures to differ by more than 10°C.

CT

 

[Roland Of Gilead]

If you read the entire sentence, the intention was to highlight how those specific workloads are so very similar between Prime95 and OCCT; not temperature differences between individual Cores.

Thank you. I did read the whole sentence.

Yes, it was out of context. The point being, it lead me to my point about the thermal paste, and a good gauge of how well the pasting was done. Relevant because it was part of the OP original questioning. The wonders of conversing!

 

[Roland Of Gilead]

Yeah I have an old GPU I can try if needs be. I'll see what happens when I remove the RTX from the system entirely and go from there.

Haha to be honest, I decided to build this PC not because I needed anything in particular but because I fancied a bit of a project so now I have the means to fiddle with things I may well get obsessed with thermal paste soon too!

Yeah, I will definitely get you posted! Thank you so much for your help so far, I've really appreciated having someone to throw ideas at me!

No worries, glad to offer help. That's the best thing about TH forum. Everyone bouncing ideas around, learning lot of good stuff as we go along! Love it!