Extremely hot i9 9900k

[thampz]

I'm having some over heating problems with my i9 9900k. I just tossed away few years old nzxt x52 aio and replaced it with NH-D15 and I'm still experiencing high temps. Slightly less with NH-D15 than with the aio. Motherboard is Asus ROG STRIX Z390-F GAMING and bios version 1401. I'm using thermal grizzly kryonaut and I also tried with the Noctua paste also. I've re-pasted this rock multiple times with alot of paste and with very small dose without any noticeable difference.

Then the information which really matters on this situation. VCore 1.33V, AVX offset 3, All-Core sync to 5GHz. Temps are running at around 98-103'C with running just Cinebench R20. LLC is 6, dropping it to 5 is still stable and temps go down by ~3-5deg.

Vcore 1.29 with AVX 0 and 4.6G is stable and temps are ok(ish).
Vcore 1.32 with AVX 2 ends in bluescreen. Power draw is around 180W
Vcore 1.33 wtih AVX 2 ends in throttling to 4.5 all core. Power draw is around 190W and drops to 160W or something when the throttling starts.
Vcore 1.33 with AVX 3 goes through Cinebenc with 4774 score, which is awfully bad for 4.7Ghz, right? Power draw is stable at around 190-195W.'

Multicore Enhancement is disabled, all volts are manually adjusted, all power limits raised to maximum. What am I doing wrong or is my rock really that bad?

 

[trents]

A 9900k with all cores set to 5 ghz. will require custom cooling. Man, you're talking about having to cool a 16 thread CPU based on a 14nm fab.

What are your temps like with all frequency and voltage settings on default?

 

[WhitehawkEQ]

Did you by chance leave the plastic lable that come on the HS? That is the only way I know of where you could have the temps you report.
Also those Vcore voltages seem high to me.

 

[thampz]

A 9900k with all cores set to 5 ghz. will require custom cooling. Man, you're talking about having to cool a 16 thread CPU based on a 14nm fab.

What are your temps like with all frequency and voltage settings on default?

If I leave everything to auto it drops to around 4.2-4.3 allcore and vcore is ridiculously high at 1.42-1.43V

Did you by chance leave the plastic lable that come on the HS? That is the only way I know of where you could have the temps you report.
Also those Vcore voltages seem high to me.

Thank god no I think all EU models come with that large hard plastic cover which goes up to the heatpipes. Plus I used alcohol to clean the surfaces for any contaminations prior to installation. Should’ve seen if there was any excess text markings on the hs at that point.

Any ideas or configurations I should try or test?

 

[trents]

It wouldn't hurt to remove the heatsink, reclean and reapply TIM. Maybe your spread wasn't good.

Do you have the latest bios installed?

 

[thampz]

It wouldn't hurt to remove the heatsink, reclean and reapply TIM. Maybe your spread wasn't good.

Do you have the latest bios installed?

I’ve reapplied TIM multiple times with variations to quantity and material itself without any significant change in thermal readings.

Afaik 1401 bios which Im using at the moment is the latest one avalaible to my MB

 

[Johan45]

Have you tried testing with the door off the case? Could be an airflow issue

 

[thampz]

Have you tried testing with the door off the case? Could be an airflow issue

Yup, no noteworthy difference. I have 3x 140mm as intake and 2x 140mm as exhaust + psu 120mm exhaust.

I'm now playing around with AI Optimization, 4.5 allcore

https://imgur.com/DoAgZvd

with Vcore 1.165 and temps are below 70'C.. My rock just doesn't like to run at 5G

 

[trents]

What is your ambient air temp?

Is it possible your Noctua cooler is defective, as in the fluid is missing from the heat pipes? You said even at default the core voltages were too high, 1.42-1.43. That is not too high if it is only momentary. Modern CPUs, both Intel and AMD, spike high voltages when the cores turbo. But it's only momentary as the cores take turns turboing. This is normal. What you didn't say was what effect putting settings on default had on temps, which is what I asked for back in post #2. Have you actually checked the temps under default conditions?

If you set all cores to a significantly lower static frequency just so you can use a significantly lower static voltage you are likely cheating yourself from a performance standpoint, especially in games, because you will be eliminating turbo.

I'm attaching a pics of my Ryzen system at stock so you can see the voltage spikes, which momentarily go well over 1.5 when cores turbo. But if you look at the four different frequency and voltage columns (Current, Maximum, Minimum and Average, it is obvious that those spikes are momentary. One pic is at idle and the other under stress. Both AMD and Intel are designed to work this way with the latest processors.

By the way, if you aren't already doing so, I would suggest using HWInfo64 for monitoring temps, voltages and frequencies. It gives a wealth of information.

 

[thampz]

24-25'C ish. Dropping side panel from my tower only drops temps under load by 0 to 1'C per core so it hardly has any impact on the outcome

Something fishy is going on with the AI thingy. Now its doing AVX work @ 4.6GHz with barely hitting 70'C and only using Vcore 1.137V. Yesterday I had to crank Vcore to 1.33V to hit 4.7G AVX workloads.. 100 Mhz = 0.2V? Or is this the 'blackhole' people have been talking about?

 

[trents]

See my post, the one previous to your last post.

 

[trents]

24-25'C ish. Dropping side panel from my tower only drops temps under load by 0 to 1'C per core so it hardly has any impact on the outcome

Something fishy is going on with the AI thingy. Now its doing AVX work @ 4.6GHz with barely hitting 70'C and only using Vcore 1.137V. Yesterday I had to crank Vcore to 1.33V to hit 4.7G AVX workloads.. 100 Mhz = 0.2V? Or is this the 'blackhole' people have been talking about?

This is normal too. The voltage v. frequency curve is not linear when overclocking. You get to a point where to get one more .1x of overclock you have to add a lot more voltage. "The wall".

- - - Auto-Merged Double Post - - -

What is "24-25'C ish"? Is that an idle temp? When I refer to default conditions I wasn't meaning idle temps. I meant load temps with frequencies and voltages left at default in bios. Oh, okay that was probably ambient temps.

 

[thampz]

What is your ambient air temp?

Is it possible your Noctua cooler is defective, as in the fluid is missing from the heat pipes? You said even at default the core voltages were too high, 1.42-1.43. That is not too high if it is only momentary. Modern CPUs, both Intel and AMD, spike high voltages when the cores turbo. But it's only momentary as the cores take turns turboing. This is normal. What you didn't say was what effect putting settings on default had on temps, which is what I asked for back in post #2. Have you actually checked the temps under default conditions?

If you set all cores to a significantly lower static frequency just so you can use a significantly lower static voltage you are likely cheating yourself from a performance standpoint, especially in games, because you will be eliminating turbo.

I'm attaching a pics of my Ryzen system at stock so you can see the voltage spikes, which momentarily go well over 1.5 when cores turbo. But if you look at the four different frequency and voltage columns (Current, Maximum, Minimum and Average, it is obvious that those spikes are momentary. One pic is at idle and the other under stress. Both AMD and Intel are designed to work this way with the latest processors.

By the way, if you aren't already doing so, I would suggest using HWInfo64 for monitoring temps, voltages and frequencies. It gives a wealth of information.

I download HWInfo and its tons of information compared to HWM.
On default-default settings the vcore is set to 1.42-1.43 all the time, not just when its turboing. Idling at 45-50'C and ramps up to 90'C and starts to throttle up and the cpu freq drops down to 4.1GHz.
Adjusting power limits on the default settings to enable higher power draw the frequency jumps to 4.2-4.3GHz while still drawing 1.42-1.43V

If I manually set All-Core Sync to 5GHz and AVX offset to 3 while having Auto volts on Bios, it still wants to use that 1.42-1.43Vcore and with all the adjustments and using 1.42Vcore it will spike up to 100'C before bsod. That 1.42Vcore doesnt fluctuate that much while on idle and it wants to be close to that all the time. Idle temps still 45-50'C with this setup.

With previous setups I've had idle ranging from 40 to 50'C. Median being 47'Cish.

Now that I'm using this AI feature my idle temps are around 37-40'C, median being closer to 38'C than 39'C. And 4.6GHz AVX work steadies to ~70'C Only thing that bothers me is how long these BSOD will occur before the AI wants to increase Vcore to be more stable. Or when does it want to try to use 4.7GHz all-core.

Also my bios indicates 1.29V for 5G and 1.29V for 4.8G AVX. I don't know how trustworthy figures these are.

 

[trents]

Over the years, many, many people have reported that AI is problematic. Have you tried uninstalling it?

 

[thampz]

Its not a software, its done in bios. And what I've read about it people have many good things to say about it. It takes a while to learn the boundaries that it can roll within and then it will be very stable and with low temperatures.

 

[thampz]

https://imgur.com/0AfgLCp

Temps idle ~38-40'C, temps at load 80-82'C, AVX work done at 4.7G. Vcore 1.252V. Power draw during 100% load is 176W. I think I'll leave it at this. Will return to this some time in the future to see if I can go over 5G without melting my whole apartment

 

[trents]

Sounds good. Does seem like maybe you didn't win the silicon lottery with that chip, however.

For future reference, here at OC Forums we prefer you attach images to your posts rather than link them. There is a tool for this built-in. Use the Go Advanced button and then Manage Attachments.

- - - Auto-Merged Double Post - - -

Are we talking about the same thing? AI = AI Suite 3?

But your strategy sounds good. Does seem like maybe you didn't win the silicon lottery with that chip, however.

For future reference, here at OC Forums we prefer you attach images to your posts rather than link them. There is a tool for this built-in. Use the Go Advanced button and then Manage Attachments.

 

[thampz]

Sounds good. Does seem like maybe you didn't win the silicon lottery with that chip, however.

For future reference, here at OC Forums we prefer you attach images to your posts rather than link them. There is a tool for this built-in. Use the Go Advanced button and then Manage Attachments.

- - - Auto-Merged Double Post - - -

Are we talking about the same thing? AI = AI Suite 3?

But your strategy sounds good. Does seem like maybe you didn't win the silicon lottery with that chip, however.

For future reference, here at OC Forums we prefer you attach images to your posts rather than link them. There is a tool for this built-in. Use the Go Advanced button and then Manage Attachments.

Not the windows application ’Asus Ai Suite’ its a bios option. I assume links to manufacturer websites are allowed so here is a more thorough explanation of ’AI overclocking’ with Asus z390 boards. https://rog.asus.com/articles/overclocking/how-to-overclock-your-system-using-ai-overclocking/

 

[trents]

Okay, thanks for the clarification and the info. So it's an advanced bios overclocking wizard (automatic overclocking). I wonder if it's actually better than manual overclocking in the bios.

 

[Woomack]

Funny (or not) is that all are searching for an issue while these chips are just acting like that.
Recently I was testing 9700K ES with D15 and it was going up to 200W+ at auto settings, boosting up to 4.7GHz with an average clock of 4.4GHz and constant 95-100°C on all cores in Prime95 after the CPU lowered voltage to ~1.3V.

The reason why I sold my 9900K was the heat. I couldn't find any cooler that was quiet enough and could provide good enough temps for longer work. Once I stopped using it for competitive benchmarking and reviews then I just sold it.

Short version. Anything that isn't causing thermal or power throttling (can be checked with hwinfo64) and temps below a maximum of ~95°C is good. AVX is heating the CPU more than any typical application so stability tests based on that are far from reality.
Don't use software for overclocking. Doing everything in BIOS is always the best option. ASUS software will install multiple services and will run in the background lowering overall performance.