[SOLVED] i9 9900K Temperature Concerns - Unsure how to proceed.

[zol0s]

Hey All,

Hoping someone could help me out here as I'm a bit lost. I recently built a new PC for myself, first time I've done this and was happy to get it working.

I decided to do some benchmarking and the results were great but temperature for the CPU is a concern. Spec is as follows:

i9 9900K 8 core 3.6GHz
RTX 2080 TI
Gigabyte Z390 Gaming SLI
Corsair H100i Hydro PRO
NZXT H500 Midi-Tower Case

At idle the CPU runs at 4.7Ghz. Idle temperture is 40-45.

I ran Prime95 with Small FFTs. Within seconds of running this, the CPU hits 100 degrees, and naturally throttles repeatedly.

I've tried cranking up the H110I, by default it's on balanced. Even with putting it at Performance it has no impact on the heat.

I've removed the water block, cleaned it and the CPU of thermal paste and reapplied new paste. I double checked the water block was fixed properly.

I am looking for some suggestions on what might be the cause here and what I could do to try and address it.

Could it be the case isn't suitable? As it's a sealed front case (not detachable), I have the radiator mounted on the front inside with fans pulling on the front of the radiator - This case is designed with mounting the radiator there and as such has a grill on the front side. There is a back fan and a top fan pushing air out.

Is it because there is something wrong with the cooler? The pipes feel warm to touch and there is noise there showing activity. Corsair Link reports what appear to be accurate values and such.

Any help would be much appreciated.

Thanks.

CPU-Z Screenshot: https://imgur.com/a/DFgFRXS
Intel Extreme Tuning Settings Screenshot: https://imgur.com/a/BXb9qSW

 

[Eximo]

What is the core voltage getting up to under load? Must be pretty high to produce those temperatures.

I wouldn't be pushing it past 1.35 volts for everyday operation, possibly less given the potential power consumption of an 8-core. (I have a kabylake chip running at 1.42 volts, but it is delidded and under custom water cooling, but only a quad core)

Keep in mind that Prime95 is a very unrealistic test. Good for verifying stability, but under most circumstances the CPU will not run like that.

I like to use Cinebench as a good approximation of a heavy gaming load, though it isn't 100% accurate there either.

 

[zol0s]

Appreciate the response Eximo.

When running Prime95 with Small FFTs it shows 1.320. At idle it fluctuates between 1.188 - 1.272

CPU test in Cinebench is 1.25 to 1.3. Temperature doesn't go over 82 there.

Single core test in Cinebench is much the same but 55 or so temperature.

I have little experience with how to set voltages and such but can see in Intel's extreme utility Core Voltage is set to adaptive currently and Core Voltage, Core Voltage Offset, AVX ratio offset and Processor Core IccMax can be adjusted. Should I try to find some suitable values for those and try out setting it off adaptive to see.

 

[Eximo]

Those values seem fine, 82C in cinebench isn't bad at all.

I would use the computer normally at this point and monitor it. I don't think you will run into many scenarios where that much of the CPU is loaded at once.

By default the chip should request only what it needs to run, and it looks like it is doing that. 1.32 is okay. It is possible it would run on less, but motherboard manufacturers aim for settings that will let the most possible number of chips operate, so sometimes they ask for a little more voltage then necessary. If you wanted to go down the path of tuning it, I suggest looking at various overclocking guides. You do basically the same thing, but in reverse, attempting to lower settings until you reach an unstable point, then increasing it back up a little.

 

[CompuTronix]

Keep in mind that Prime95 is a very unrealistic test.

Eximo,

Respectfully, while this is very true for AVX versions such as 29.4, version 26.6 Small FFT's is actually ideal. However, when members read threads that say don't run Prime95 for the reason you cited, or don't offer any reason why, or read other threads that say don't run versions later than 26.6, this conflicting information further confuses the community in an already confusing and poorly understood topic.

When we generalize about Prime95 and don't emphasize that there's a distinction between versions, the community takes that at face value and tends to shy away from Prime95 altogether, without realizing there's a non-AVX version which is ideally suited to Intel's test criterion, in accordance with the Datasheets. It's well worth a few more keystrokes to point out that "Prime95 versions later than 26.6 should not be used".

It would help a great to dispel the confusion in the community if more members mentioned that the AVX issue is simply version specific. I've invested thousands of hours over the course of more than a decade researching, testing, documenting and validating various approaches to thermal testing. I take this work very seriously. If anyone has more credible, concise, comprehensive or well founded methods, then please share them with us.

Guys,

For the benefit of other members as well as Tom's readers lurking in the background, for the record, there's nothing wrong with Prime95 version 26.6. Here's why:

Not all loads are created equal. “Stress” tests vary widely and can be characterized into two categories; stability tests which are fluctuating workloads, and thermal tests which are steady workloads. Intel tests their processors at a steady 100% TDP workload to validate Thermal Specifications.

Prime95 version 26.6 Small FFT's is ideal for CPU thermal testing, because it's a steady 100% workload with steady Core temperatures that typically runs Core i variants with Hyperthreading and Core 2 processors within +/- a few % of TDP. No other utility so closely replicates Intel's test conditions.

Utilities that don't overload or underload your processor will give you a valid thermal baseline. Here’s a comparison of utilities grouped as thermal and stability tests according to % of TDP, averaged across six processor Generations at stock settings rounded to the nearest 5%:

Higher TDP tests produce higher Core temperatures. All tests will show 100% CPU Utilization in Windows Task Manager, which indicates processor resource activity, not % TDP workload. Although actual Power dissipation (Watts) varies with Core speed, Core voltage and workload, Prime95 v26.6 Small FFT’s always provides a true and steady 100% workload, whether you’re running stock or overclocked.

Shown above from left to right: Small FFT's, Blend, Linpack and IntelBurn Test.

The steady thermal signature of Small FFT's allows accurate measurements of Core temperatures. A steady 100% workload is key for thermal testing so the CPU, cooler, socket, motherboard and voltage regulators can thermally stabilize.

2nd and 3rd Generation i7, i5 and i3 CPU’s have AVX (Advanced Vector Extension) Instruction Sets. 4th through 9th Generation i9, i7, i5 and i3 CPU’s have AVX2 Instruction Sets. Prime95 versions later than 26.6 run AVX/2 code on the CPU's Floating Point Unit (FPU), which is an unrealistic workload. 2nd and 3rd Generation CPU’s are minimally affected by AVX, but 4th through 9th Generation with AVX2 may experience Core temperatures up to 20°C higher.

Many 6th through 9th Generation motherboards address the AVX problem by providing “offset” adjustments (downclock) in BIOS. -3 (300 MHz) or more may be needed to limit Core temperatures to 85°C. Since 4th and 5th Generation don’t have AVX offsets, you can create a BIOS Profile for gaming, and a downclock Profile for AVX apps such as rendering or transcoding. If you don’t use AVX apps, BIOS should still be configured for it, as certain utilities use AVX for stability testing.

AVX can be disabled in Prime95 versions later than 26.6 by inserting "CpuSupportsAVX=0" into the "local.txt" file in Prime95's folder. However, since Core temperatures will be the same as 26.6, it's easier to just use 26.6. AVX doesn't affect Core i 1st Generation, Core 2, Pentium or Celeron processors as they don't have AVX/2 Instruction Sets. As per Intel’s Datasheets, TDP and Thermal Specifications are validated “without AVX”.

CT

 

[Eximo]

I won't say I completely disagree on Prime95, just not something I like to use to validate temperature or voltage for quick tests. I think avoiding it is actually somewhat decent advice because of the version confusion, rather then attempting to educate people on something that can be confusing for first time builders. Unless I am trying to saturate a cooling loop or something, even then I might lean towards AIDA64 and toss graphics into the mix.


But I still don't think Prime95 running 100% across all cores is a realistic everyday use scenario, which is a generalization, but doesn't involve the AVX instruction sets at all. I can't think of any programs I run on my computer that will stress a CPU like that, other than benchmarks.

Were I building a 24/7 box to do constant calculations, sure it is a valid test. Or a machine that is used for some business purposes like rendering, editing, or encoding. But I've never felt it necessary to do 24 hour runs of Prime95 to validate system stability to that extent. Perhaps when troubleshooting intermittent stability.

I can't say I've installed Prime95 26.6 on a system recently, though. But that reflects my opinion of it.

 

[vMax]

I have to agree with Eximo as the temps at load are about right. I too would only use Prime95 version 26.6 and for the 9900K the blend test is more than good enough to test for stability over a long run. Small FFT's is going to put stress that will probably never be hit in the normal life of the CPU. The case setup looks fine in terms of airflow.

The only other thing that worries me is your idle temps as I would expect very low 30's C degrees.

One final point, manual overclocking at 4.7GHz with an ofset should allow you to run it at a much lower vcore and I would try that first...A bit of learning first on the settings for that motherboard but you should be able to hit 4.7 at under 1.3v with something like 1.27v

 

[CompuTronix]

I won't say I completely disagree on Prime95, just not something I like to use to validate temperature or voltage for quick tests. I think avoiding it is actually somewhat decent advice because of the version confusion, rather then attempting to educate people on something that can be confusing for first time builders. Unless I am trying to saturate a cooling loop or something, even then I might lean towards AIDA64 and toss graphics into the mix.

But I still don't think Prime95 running 100% across all cores is a realistic everyday use scenario, which is a generalization, but doesn't involve the AVX instruction sets at all. I can't think of any programs I run on my computer that will stress a CPU like that, other than benchmarks.

Were I building a 24/7 box to do constant calculations, sure it is a valid test. Or a machine that is used for some business purposes like rendering, editing, or encoding. But I've never felt it necessary to do 24 hour runs of Prime95 to validate system stability to that extent. Perhaps when troubleshooting intermittent stability.

I can't say I've installed Prime95 26.6 on a system recently, though. But that reflects my opinion of it.

For testing processor stability, Prime95 v26.6 Small FFT's is only useful for quick checks during initial overclocking efforts. However, for thermal testing, v26.6 Small FFT's is ideal for a wide variety users who, other than gaming, may potentially run AVX workloads such as rendering, transcoding, CAD, scientific number crunching, or even SETI, which can approach or reach the Core temperatures of P95 v26.6 Small FFT's.

If you look closely at the % TDP scale I posted above, you can see how AVX versions of Prime95 Small FFT's are absolutely way-overkill-brutal, which is most definitely an extremely unrealistic workload. Most users don't realize that workloads can extend well beyond 100% CPU Utilization that Windows Task Manager mis-leads everyone to think.

In the Intel Temperature Guide, I'm obligated to consider not just gamers, (which includes me), but noobs, power users, accomplished overclockers as well as all users, so that "one size fits all". Moreover, to my knowledge, I've tested every utility available, and 26.6 Small FFT's is the only test that so closely matches Intel's proprietary methodology for validating TDP and temperature. Since P95 v26.6 Small FFT's typically matches the specifications within + / - a few %, in my opinion, the test is as valid as it gets. I can't arrive at any other conclusion.

Regardless of purpose, it's more or less pointless to run P95 v26.6 Small FFT's for more than 2 hours. There are many other utilities which are much better suited for stability testing than Small FFT's. Conversely, Blend is quite good for testing memory stability, which I see as required for a new build or troubleshooting possible memory issues. RealBench is also an excellent utility for testing overall system stability, and runs a realistic AVX workload.

Concerning AIDA64, I'm not a big fan because there are 15 possible test selections which yield 15 different Core temperatures and thermal behaviors. Just as with generalizing Prime95, extremely few AIDA64 users ever bother to mention which test(s) they selected.

We live in a world where all our techno-gadgets function 24/7/365 based on engineering standards and specifications. So when discussing technologies, I believe it's absolutely critical to be very specific, because it's all about attention to detail. Otherwise, we proceed on false assumption and are easily misled, simply because a known relevant detail wasn't mentioned. It's one of those ... ahhh haaa moments ... when we say; "that would've been good to know"!

We constantly see threads that begin with "hot" and "Prime95". So 20 posts later after the thread has long since taken a wrong turn, still no one has thought to ask, and the OP hasn't bothered to mention, which version ... and let's not forget ambient temperature ... while "standard" or normal is 22°C or 72°F, we have members who start threads without mentioning that they're running their PC in their igloo at 10°C ambient ... or in the outback at 40°C ... which also would've been good to know!

So whether or not you agree with these hard-earned insights, my point is that when we write our answers, responses and explanations, the essence of the topic is in the details, so it's important to be specific.

CT

 

[zol0s]

Those values seem fine, 82C in cinebench isn't bad at all.

I would use the computer normally at this point and monitor it. I don't think you will run into many scenarios where that much of the CPU is loaded at once.

Appreciate all the help Eximo - I'm going to monitor and take a look at the overclocking guides to see about lowering the voltage.

@vMax

The only other thing that worries me is your idle temps as I would expect very low 30's C degrees.

One final point, manual overclocking at 4.7GHz with an ofset should allow you to run it at a much lower vcore and I would try that first...A bit of learning first on the settings for that motherboard but you should be able to hit 4.7 at under 1.3v with something like 1.27v

Thanks for the suggestion there - As above, when I feel confident in lowering the voltage I'll give that a shot.

I feel a bit more confident that the termatures don't appear to be wildly "out there".

@CompuTronix

and the OP hasn't bothered to mention, which version ... and let's not forget ambient temperature ... while "standard" or normal is 22°C or 72°F, we have members who start threads without mentioning that they're running their PC in their igloo at 10°C ambient ... or in the outback at 40°C ... which also would've been good to know!

Appreciate your thorough response, though much of it went over my head at this stage It's not that I hadn't bothered to mention, I'd hope the amount of information I provided in my last two responses at least demonstrated that I was making some effort to provide what I thought would be enough to help with the supposed issue. But I do apologise if that wasn't the case! One would hope that when someone emphasises inexperience in dealing with such things, a little patience and understanding can be granted by those who are generous enough to spend their own time in offering assistance.

 

[CompuTronix]

and the OP hasn't bothered to mention, which version ... and let's not forget ambient temperature ... while "standard" or normal is 22°C or 72°F, we have members who start threads without mentioning that they're running their PC in their igloo at 10°C ambient ... or in the outback at 40°C ... which also would've been good to know!

Appreciate your thorough response, though much of it went over my head at this stage It's not that I hadn't bothered to mention, I'd hope the amount of information I provided in my last two responses at least demonstrated that I was making some effort to provide what I thought would be enough to help with the supposed issue. But I do apologise if that wasn't the case! One would hope that when someone emphasises inexperience in dealing with such things, a little patience and understanding can be granted by those who are generous enough to spend their own time in offering assistance.

zol0s,

Thank you for expressing your thoughts, however, allow me to apologize for giving you the wrong impression, as I was addressing Eximo. Just to be clear, no discourtesy was intended toward anyone. To the contrary, on behalf of Tom's Moderator Team, it's customary for me to welcome you aboard!

Eximo and vMax have provided some excellent advice. There are several good overclocking guides available through Google, where you can learn how to reduce Core voltage and temperatures. We always encourage users to research the guides, especially when approaching new technology, so that potentially damaging one's hard earned hardware can be avoided.

Also, from our Forum Rules: http://www.tomshardware.com/forum/id-2083474/read-forum-rules-styling-posts.html

• Search the forum before posting. Someone else may have already answered your question or discussed the topic. Checking for duplicates can help save you time.
• Read the Stickies! Sticky threads at the top of the forum category answer a lot of questions.
• If asking a question, provide as much in the way of details as you can to help the experts answer. Include hardware specs and OS.

Part of my work here is to help users solve temperature problems, not only just one thread at a time, but also by providing a guide as a reference to educate users about processor temperatures. It's kept up to date and can be Googled, but it's also available at the top of the CPUs Forum, or by clicking on the link in my signature. If you have any other questions or concerns, please feel free to contact me by Private Message.

Oh, and if you wouldn't mind ... which version of Prime95 did you use, and what was your ambient temperature?

Once again, welcome aboard!

CT