9900k with h150i pro xt, high temps than expected?

Hello,

My specs are:

9900k

Maximus XI hero

h150i pro xt (exhaust, top)

Lian Li oc11

6 ML120 fans as intake (bottom, side)

On stock CPU settings temps looks like:

Idle: 31-33c

R23: 69-72c

Prime: 70-73c

5 ghz (1.3 vcore, level 7 llc):

Idle: 35-38c

R23: 81-84c

Prime: 99c and shutdown (well, i know shutdown is instability and i should increase vcore, but with 99c showing, increasing that would not make situation better, so back to stock)

P.S Coolant temp on idle (non-oc): 27-28c, gaming load: 34-35c.

P.S 2 During OC, AVX offset was 1 (4.9ghz).

What is wrong with this cooler?

We need some coolant temperature (H150i Temp) information to assess the cooler's functionality. However, the clear difference in potential lowest coolant temp (idle temperature) to Prime load CPU temp between your stock and 5GHz settings indicates this is going to be a settings and voltage issue (+40 vs +65C). If there was a problem with the cooler, it would be present in both situations. If you start Prime 95 and 1 second later you are at 90 something plus, there is voltage/load issue.

I can see you've set LLC 7 and that should keep the voltage on a tighter rein, but actual Vcore under Prime 95 and BIOS set values are two different things.

without AVX offset you probably won't be able to run Prime95 with that voltage and LLC speaking form my own experience with the 9900k.

Edit: For reference, i'm running -3 AVX offset @5G with BIOS 1,32V/LLC6

Prime95 small FFTs hottest core 93C/coolest 88C

full custom loop with huge monoblock and pump cranked to 100%/4800rpm

It's not even 100% stable with this settings but good enough for daily use.

We need some coolant temperature (H150i Temp) information to assess the cooler's functionality. However, the clear difference in potential lowest coolant temp (idle temperature) to Prime load CPU temp between your stock and 5GHz settings indicates this is going to be a settings and voltage issue (+40 vs +65C). If there was a problem with the cooler, it would be present in both situations. If you start Prime 95 and 1 second later you are at 90 something plus, there is voltage/load issue.

 

I can see you've set LLC 7 and that should keep the voltage on a tighter rein, but actual Vcore under Prime 95 and BIOS set values are two different things.

Coolant temp is 27-28c on idle and for example after 2-3 hours of RDR2 it was around 35c (cpu temp for that game is maximum 60c, around 55-58c, sometimes less on 60fps tho)

without AVX offset you probably won't be able to run Prime95 with that voltage and LLC speaking form my own experience with the 9900k.

as far as i remember AVX offset was 1, so 4.9ghz.. but prime test were run with AVX disabled.

Right, so at "stock voltage" (which can be anything on a Asus Z390) the Prime coolant/idle to load CPU temp differential is approximately +40C (~30->70C) and when manually(?) set to 1.30v you are getting +65-70C (30->99C). Conductivity doesn't change, so this suggests the level of current/voltage has changed dramatically. +40C is about right for most default set-ups, especially with "motherboard enhancement" overclocking. However, +65-70 is out of bounds and no system can deal with that. You would need to keep the ambient temp in the 15C range to make that workable. That strongly suggests extra voltage is being applied, likely near 1.40v.

Also, running synthetic tests on the more recent motherboards can be tricky. No more out of the box and fire up Prime 95. If Prime leads to instant shutdown, but other stress tests are substantially lower, that suggests there is a configuration or unique issue with Prime. If Cinebench, AIDA, CPU-Z, OCCT, etc. also result in near maximum CPU temperature, it's surely a global BIOS setting related to voltage. If it's just Prime 95, then perhaps there is no need to chase down the issue.

Right, so at "stock voltage" (which can be anything on a Asus Z390) the Prime coolant/idle to load CPU temp differential is approximately +40C (~30->70C) and when manually(?) set to 1.30v you are getting +65-70C (30->99C). Conductivity doesn't change, so this suggests the level of current/voltage has changed dramatically. +40C is about right for most default set-ups, especially with "motherboard enhancement" overclocking. However, +65-70 is out of bounds and no system can deal with that. You would need to keep the ambient temp in the 15C range to make that workable. That strongly suggests extra voltage is being applied, likely near 1.40v.

 

Also, running synthetic tests on the more recent motherboards can be tricky. No more out of the box and fire up Prime 95. If Prime leads to instant shutdown, but other stress tests are substantially lower, that suggests there is a configuration or unique issue with Prime. If Cinebench, AIDA, CPU-Z, OCCT, etc. also result in near maximum CPU temperature, it's surely a global BIOS setting related to voltage. If it's just Prime 95, then perhaps there is no need to chase down the issue.

Is not 1.4 overkill? since 1.3 does pretty fine on r23, just quite high temps, +10c.

Btw, voltage droops are also happening, i mean when i was 5@1.3vcore hwinfo showed 1.288 even while r23, sometime it dropped, but more often i saw that number... That's why i went with llc7

i mean when i was 5@1.3vcore hwinfo showed 1.288 even while r23, sometime it dropped, but more often i saw that number... That's why i went with llc7

That's what I was fishing for. You had not indicated if you set 1.30v manually, adaptive, offset, or something else. People often report "the Vcore is 1.30v", but what they really mean is they set it to 1.30. Doesn't mean it stays there.

So if the LLC setting is keeping things tight, then the next question is what are the temps on or any other stress test besides Prime95 at the 5.0/1.30v settings?

**I see the R23 score now for 5.0. This goes back to what I talking about before. At "stock" the stress tests are essentially equal in temps. At 5.0/1.30 Prime 95 is +16C over R23. That suggests an issue with Prime vs something else. You can confirm this by running other stress tests which should bring back similar "instant on" numbers. Something is a cooler issue if you start off and CPU temps ride at 75C, but then they continually escalate. If you start a fixed load test and 1 second later it's at an unsustainable level, then there is a voltage/setting issue. However, presumably you don't actually need to run Prime95 to hunt Mersenne primes.

Is not 1.4 overkill? since 1.3 does pretty fine on r23, just quite high temps, +10c.

Btw, voltage droops are also happening, i mean when i was 5@1.3vcore hwinfo showed 1.288 even while r23, sometime it dropped, but more often i saw that number... That's why i went with llc7

What LLC does it basically shoves additional voltage into your CPU to counteract voltage drop under load. the higher LLC you set, the more voltage it shoves in. This means of that your voltage is more "stable" but at the cost of more heat generated by the CPU, since there actually is more voltage than you set in your BIOS while under load. This is what you see when your CPU temps suddenly spike into thermal limit once you apply a heavy load like Prime95. Generally I stick to LLC 4-6 and finetune with an offset Voltage if I really wanna tweak things. Getting 5G or even 4.9G stable with reasonable temps in Prime95 is kinda nearly impossible and also unnecessary if you're not benchmarking or some sort. Prime95 is only good for generating heat and maxing out your wattage if you ask me. just turn it down to 4.7-4.8G and call it a day.