Question 12900KS high temperatures?

Deco22

Honorable
Aug 6, 2017
24
0
10,520
Hi guys,

I've been struggling with my 12900ks so I wanted some input from people who will know better than I do.

Specs:
i9-12900KS @ stock settings
Asus Maximus Z690 Hero ( with/ Thermalright LGA 1700 Contact Frame)
Custom loop (360mm rad, corsair xc7 water block, xd5 pump/res)
32GB T-Force Delta 6200MHZ
Gigabyte RTX 3080Ti Gaming OC
Corsair HX1000i

I'm getting temperatures of 90-100c on the package at stock settings in cinebench r23 runs and also just in some general games, this seems pretty high to me since I just upgraded to a custom loop to try and solve the high temps I was getting with my previous 3690 aio, but it doesn't seem to have made much of a difference.
I'm seeing conflicting things online about normal operating temps for the 12900k and ks so I'm not entirely sure what to expect out of it, but hitting 100c in games seems abnormal to me. One thing I have seen a lot of is about mounting pressure with the 12th gen chips, but since I am using the Thermalright contact frame, I don't think this is the problem, but I'm not entirely sure.
All of my settings in BIOS are stock (except XMP) so the Intel stock power limits are enforced. I have tried a 0.025v negative offset which led to an immediate BSOD in a cinebench run, although I am by no means an experienced overclocker so I am not great at tweaking voltages and stuff, but I really wanted to try and figure out how to get the best out of this chip.

Here is a link to a screenshot of HWinfo just after playing a game: View: https://imgur.com/a/tWQjjET
the vcore seems high to me but again I'm not really sure of what the average stock vcore is for the 12900ks. As a side note, what is the difference between the 2 Package Temperature readings under CPU#0 12900KS: DTS and CPU#0 12900KS: Enhanced, and which is the more reliable reading?

Any help is greatly appreciated,

Thanks :)
 
  • Like
Reactions: KyaraM

Karadjgne

Titan
Ambassador
It's a KS. Essentially that's a best binned K on steroids with a factory undervolt/overclock tweak already added. You have 3 cores that hit 5.5GHz, the other 5 hit 5.2GHz.

With a single 360mm rad, that loop is not that much better than the AIO. Liquid cooling isn't magic, it's just higher capacity. You've moved from @ 350w to @ 400w. If the efficiency is anywhere close, temps won't move at all because of what @TerryLaze explained, the cpu will just boost a couple cores higher and raise voltages higher to make up the difference. You'd need to have closer to double the cpus possible output, where it hits the maximum amount of possible boost and stops, and then that 100°C will drop.

But unless you hit that cpu boost cutoff, or disable the boost entirely, if the cpu has the headroom to keep boosting, it will.

Doesn't help that that XD5 pump is kinda wimpy for a D5, it's got lower L/h and half the head pressure of my tiny ddc pump at double the rpm.
 
  • Like
Reactions: KyaraM

Deco22

Honorable
Aug 6, 2017
24
0
10,520
It's a KS. Essentially that's a best binned K on steroids with a factory undervolt/overclock tweak already added. You have 3 cores that hit 5.5GHz, the other 5 hit 5.2GHz.

With a single 360mm rad, that loop is not that much better than the AIO. Liquid cooling isn't magic, it's just higher capacity. You've moved from @ 350w to @ 400w. If the efficiency is anywhere close, temps won't move at all because of what @TerryLaze explained, the cpu will just boost a couple cores higher and raise voltages higher to make up the difference. You'd need to have closer to double the cpus possible output, where it hits the maximum amount of possible boost and stops, and then that 100°C will drop.

But unless you hit that cpu boost cutoff, or disable the boost entirely, if the cpu has the headroom to keep boosting, it will.

Doesn't help that that XD5 pump is kinda wimpy for a D5, it's got lower L/h and half the head pressure of my tiny ddc pump at double the rpm.
That makes sense, but surely the cpu shouldn't hit those temps straight away (and especially in games) if it was limited by the capacity of the loop? It will literally hit 100c within 5 seconds of starting the test with the liquid temp being around 28 - 30c and the peak temp of the liquid getting to around 37c. Maybe I'm misunderstanding but it seems like the issue is between the CPU IHS and the coldplate of the block? I guess delidding would help with that but I really dont want to go down that route if I dont need to. Thanks for the reply :)
 

Karadjgne

Titan
Ambassador
It's not limited by the capacity of the loop. The loop can handle upto @ 400w before its efficiency is gone. Now while that cpu is dumping 250+ watts into the loop, that wattage is not the load, it's the power used. The load is different. That's an instant temp rise created by the amount of work the core has to do.

It takes time for that heat to move from the core, through the top of the silicon, through the Tim, through the IHS, through the paste, into the coldplate and get absorbed by the coolant. What the coolant sees is seconds after what the cpu uses.

Unchecked, that cpu is fully capable of temps exceeding 200°C, the cooler moderates that by absorbing the output wattage, but can't affect the temps created by the load only the results of the load.

Add another 240mm (300w) into that loop and you'll have enough cooling capacity to overcome wattage output vs efficiency. That's when you'll see the cpu actually reach voltage limits or stability limits and stop boosting, which will stabilize the load and then the coolant can have a positive affect on temps.

As it is right now, anytime the cooler would reduce that 100°C to 99°C, the cpu says 'I don't think so...' and boosts another core higher, bringing it to 100°C again.
 

Deco22

Honorable
Aug 6, 2017
24
0
10,520
It's not limited by the capacity of the loop. The loop can handle upto @ 400w before its efficiency is gone. Now while that cpu is dumping 250+ watts into the loop, that wattage is not the load, it's the power used. The load is different. That's an instant temp rise created by the amount of work the core has to do.

It takes time for that heat to move from the core, through the top of the silicon, through the Tim, through the IHS, through the paste, into the coldplate and get absorbed by the coolant. What the coolant sees is seconds after what the cpu uses.

Unchecked, that cpu is fully capable of temps exceeding 200°C, the cooler moderates that by absorbing the output wattage, but can't affect the temps created by the load only the results of the load.

Add another 240mm (300w) into that loop and you'll have enough cooling capacity to overcome wattage output vs efficiency. That's when you'll see the cpu actually reach voltage limits or stability limits and stop boosting, which will stabilize the load and then the coolant can have a positive affect on temps.

As it is right now, anytime the cooler would reduce that 100°C to 99°C, the cpu says 'I don't think so...' and boosts another core higher, bringing it to 100°C again.
Oh ok I think I'm getting you. So the CPU at the moment is just going to keep eating up any thermal headroom it has in exchange for performance? Surely it can't be good for it to be hitting 100c all the time, even under fairly mediocre loads like gaming? Also if I'm understanding correctly that is controlled by Adaptive Boost, but I've tried turning that off and am still seeing exactly the same behaviour. Thanks :)
 

Karadjgne

Titan
Ambassador
Yep. Exactly. But be careful with what you call a mediocre load. Temp reader software will only show a single temp, the hottest core. So with a very mediocre load like CSGO that's only using 2 cores, those 2 cores will be high % usage, and 100°C, but the rest of the cpu is running 55°C. You'll see it as 100°C, but it's really not the entire cpu, just those 2 cores. For a more accurate temp, you'd need software like HWInfo64 and look at All the cores, all the core temps, all the core speeds, when running that mediocre load.

Also be aware that cores change. So in CSGO, you might see core0 and core1 running that for a few seconds, then it'll switch to core6 and core7 for a few seconds etc. So no one core takes the full brunt of the load for extended periods.

To really see exactly the affect the cpu has on cooling, check the coolant temps. Think of a decent sized pot of water on the stove. On high, that's @ 1500w, and still seems to take forever before the water temp even changes 1°C. Your cpu is putting out 250w-300w full core, so partial core use will be considerably less. Even under a seriously heavy load, for over half an hour, should be looking at coolant temps of 40° to 50°C at most. (pumps are rated for upto 50°C in general.)

If your coolant is beyond 50°C after an hour of that load, you have a loop issue, something isn't doing its job right. Which will affect cpu temps at a lower coolant temp. Could be as simple as the cpu block is upside down and the inlet coolant isn't hitting the jetplate, or you have a restriction in the rad or fittings, pipes too small, pump output weak etc.
 

Deco22

Honorable
Aug 6, 2017
24
0
10,520
Yep. Exactly. But be careful with what you call a mediocre load. Temp reader software will only show a single temp, the hottest core. So with a very mediocre load like CSGO that's only using 2 cores, those 2 cores will be high % usage, and 100°C, but the rest of the cpu is running 55°C. You'll see it as 100°C, but it's really not the entire cpu, just those 2 cores. For a more accurate temp, you'd need software like HWInfo64 and look at All the cores, all the core temps, all the core speeds, when running that mediocre load.

Also be aware that cores change. So in CSGO, you might see core0 and core1 running that for a few seconds, then it'll switch to core6 and core7 for a few seconds etc. So no one core takes the full brunt of the load for extended periods.

To really see exactly the affect the cpu has on cooling, check the coolant temps. Think of a decent sized pot of water on the stove. On high, that's @ 1500w, and still seems to take forever before the water temp even changes 1°C. Your cpu is putting out 250w-300w full core, so partial core use will be considerably less. Even under a seriously heavy load, for over half an hour, should be looking at coolant temps of 40° to 50°C at most. (pumps are rated for upto 50°C in general.)

If your coolant is beyond 50°C after an hour of that load, you have a loop issue, something isn't doing its job right. Which will affect cpu temps at a lower coolant temp. Could be as simple as the cpu block is upside down and the inlet coolant isn't hitting the jetplate, or you have a restriction in the rad or fittings, pipes too small, pump output weak etc.
Ok yeah makes sense. I still havent really been able to run a test for an extended period of time because the cpu will always hang around 100 and I'm not really too happy with keeping it there for extended periods of time. One thing I have found that really makes no sense to me is that if I run a prime95 (non avx) small fft on 8 threads the cpu will instantly hit 105, and begin to throttle, compared to a run with all 24 threads where it will take a couple minutes to hit 100c. Is there a specific reason for that? Thanks
 

Karadjgne

Titan
Ambassador
Lower core count = higher voltage due to higher boost. If you look at those 8 workers they'll all be @ 5.2GHz, after temps acclimate, but all 24 threads will be considerably lower, especially the E-cores. It's a dynamic response. If you think 250w, that's cpu max, 250w on 8 cores vs 24 cores.
 

Karadjgne

Titan
Ambassador
Only thing is add a 240mm to the loop. Spreads the energy in the coolant out on more fin area, so each fan is more effective at dissipating that energy. The coolant is a medium, it in itself is not cpu temp hot, it just carries the wattage absorbed. It's why loop order doesn't really matter, that energy is spread in the complete loop, not concentrated in one area, like before the rad.

The cpu will boost cores until something stops it. That'll be voltage or 100°C, whichever comes first. Get enough energy dissipative capacity and the cpu cannot hit 100°C before it tops out on total voltages, at which point the rads now have a chance to continually lower the temp reached as the cpu is now a static, finite wattage, not a dynamic wattage.

It's like working out in a gym, hard-core. If the gym is 72°F, you'll break a sweat and soak your shirt because your skin can't dissipate the heat fast enough. Drop the gym temp to 32°F, and you won't sweat because even with the same intensive workout, you skin is cold enough not to need to sweat no matter how hard the workout. At 52°F, you'll still sweat, but not soak unless you increase the workout. That's where Op is at now, the cpu has room for a harder workout.

In technical terms, hasn't reached critical mass yet, that tipping point where temps stop being the limiting factor to core boost, and voltages take over.
 

Deco22

Honorable
Aug 6, 2017
24
0
10,520
Sorry guys been busy the past couple of days, but thanks for the replies.

Lower core count = higher voltage due to higher boost. If you look at those 8 workers they'll all be @ 5.2GHz, after temps acclimate, but all 24 threads will be considerably lower, especially the E-cores. It's a dynamic response. If you think 250w, that's cpu max, 250w on 8 cores vs 24 cores.

Yeah I guess that makes sense, vcore is way higher when its just the 8 threads running, which is what I thought it probably was, with the current settings there is a lot of vdroop so when its not fully loaded the voltage is a lot higher then when all cores are 100%.

Nitpicking maybe, but HWInfo and others do show all cores. Unless you have the invidual core section collapsed like TC. That's my only complaint about your post.

I'm not sure why I had the core clocks expanded but not the core temps, that ones my bad :LOL:, but you're right at least for now, I'll probably just have to try and find the lowest stable vcore I can and maybe reduce the clocks, at least until I add another radiator.

Only thing is add a 240mm to the loop. Spreads the energy in the coolant out on more fin area, so each fan is more effective at dissipating that energy. The coolant is a medium, it in itself is not cpu temp hot, it just carries the wattage absorbed. It's why loop order doesn't really matter, that energy is spread in the complete loop, not concentrated in one area, like before the rad.

The cpu will boost cores until something stops it. That'll be voltage or 100°C, whichever comes first. Get enough energy dissipative capacity and the cpu cannot hit 100°C before it tops out on total voltages, at which point the rads now have a chance to continually lower the temp reached as the cpu is now a static, finite wattage, not a dynamic wattage.

It's like working out in a gym, hard-core. If the gym is 72°F, you'll break a sweat and soak your shirt because your skin can't dissipate the heat fast enough. Drop the gym temp to 32°F, and you won't sweat because even with the same intensive workout, you skin is cold enough not to need to sweat no matter how hard the workout. At 52°F, you'll still sweat, but not soak unless you increase the workout. That's where Op is at now, the cpu has room for a harder workout.

In technical terms, hasn't reached critical mass yet, that tipping point where temps stop being the limiting factor to core boost, and voltages take over.

I guess my next purchase will have to be another rad, was probably gonna have to get one eventually anyway, just a shame I'm gonna have to tear down the whole loop again so soon, but thanks for all the insight, much appreciated :)