What's a good temp for my 13900k?

Zorachus

[H]F Junkie
Joined
Dec 17, 2006
Messages
11,278
I just built a partially new ststem, specs in sig. I have the Kraken Z73 on my i9 13900k, and in Desktop use it hovers between 28/30c to 35c/37, and in games like 45c / 54c in heavy gaming. Is that ok?
 
I just built a partially new ststem, specs in sig. I have the Kraken Z73 on my i9 13900k, and in Desktop use it hovers between 28/30c to 35c/37, and in games like 45c / 54c in heavy gaming. Is that ok?
I'd like to get mine a bit lower. I'm seeing temps around 44 at idle and 74-76 gaming. All stock except for xmp enabled
 
Hey I actually have the same cpu and AIO, but my temps are rather high with push/pull setup.
It's about 45c idle and 60-70c gaming, but I have my cpu overclocked. In Cinebench R23 multicore test it easily goes to almost 100c.
 
I just built a partially new ststem, specs in sig. I have the Kraken Z73 on my i9 13900k, and in Desktop use it hovers between 28/30c to 35c/37, and in games like 45c / 54c in heavy gaming. Is that ok?
This is where my 13900K is usually at during gaming with a DeepCool LS720. Running at stock with XMP enabled. Others with higher temps may want to look at case airflow as well.
 
With 24 cores in Cinebench it will fire up to 100 unless you have overkill cooling. What I do do since I only game is turn off most e-cores and disable hyperthreading. That way I can decrease the voltage significantly. Just yesterday I was able to pull off 32 degrees 150 W power and -0.100v. Somewhere between 12 or 16 core is more than enough for gaming if you ask me. If you do productivity work nevermind I would leave hyperthreading and all E-cores on. I don't so I can afford to turn them off.
 
Like xDiVolatilX said, I have mine undervolted by -0.100v as well and it works great, gaming is around 50C and cinebench runs it up to 85C or so, at stock voltage I was seeing 100C on several P cores in CB23. I'm using an EK 360 AIO with low speed fans on it, but I don't render or anything just gaming so it works well with low noise.
 
With 24 cores in Cinebench it will fire up to 100 unless you have overkill cooling. What I do do since I only game is turn off most e-cores and disable hyperthreading. That way I can decrease the voltage significantly. Just yesterday I was able to pull off 32 degrees 150 W power and -0.100v. Somewhere between 12 or 16 core is more than enough for gaming if you ask me. If you do productivity work nevermind I would leave hyperthreading and all E-cores on. I don't so I can afford to turn them off.

Yah the problem is that hyperthread threads are used by thread director - you'll actually gain a bit of efficiency overall with it on
 
Yah the problem is that hyperthread threads are used by thread director - you'll actually gain a bit of efficiency overall with it on
What do you mean by efficiency? For example if you have 12 or 16 physical cores?
 
What do you mean by efficiency? For example if you have 12 or 16 physical cores?
So..

My understanding of thread director comes from a number of sources, but the info is most easily understood from here: https://www.anandtech.com/show/16881/a-deep-dive-into-intels-alder-lake-microarchitectures/2

Intel classifies the performance levels on Alder Lake in the following order:

  1. One thread per core on P-cores
  2. Only thread on E-cores
  3. SMT threads on P-cores
That means the system will load up one thread per P-core and all the E-cores before moving to the hyperthreads on the P-cores.
...it continues on


What this means is that without the SMT threads, it will mean firing up or keeping awake E cores (if the P cores are active and E cores are asleep, but SMT disabled) and/or reallocation to P cores as there is no demotion to SMT- therefore increasing power consumption - it is cheaper in terms of power to keep a P core awake and run two threads on it, rather than a P core and an E core.

I would think it'd need some sort of algorithm to go:
1. Profile thread on P core
2. If P core use of thread is low, move to e core, profile thread on ecore
3. if e core use of thread is low, and P core is active (for something else), move to SMT thread otherwise keep running on e core
 
Last edited:
So..

My understanding of thread director comes from a number of sources, but the info is most easily understood from here: https://www.anandtech.com/show/16881/a-deep-dive-into-intels-alder-lake-microarchitectures/2


...it continues on


What this means is that without the SMT threads, it will mean firing up or keeping awake E cores (if the P cores are active and E cores are asleep, but SMT disabled) and/or reallocation to P cores as there is no demotion to SMT- therefore increasing power consumption - it is cheaper in terms of power to keep a P core awake and run two threads on it, rather than a P core and an E core.

I would think it'd need some sort of algorithm to go:
1. Profile thread on P core
2. If P core use of thread is low, move to e core, profile thread on ecore
3. if e core use of thread is low, and P core is active (for something else), move to SMT thread otherwise keep running on e core
Correct me if I'm wrong, but I believe this is for battery efficiency? On a desktop with 1300W power supply I don't think this would make a difference. Efficiency for battery power and eliminating unnecessary cores for gaming might be 2 separate cases to consider. Turning off cores that you don't need for gaming might save you power overall? My testing has proven this. But I am testing Cinebench which uses all cores. Basically each cores turned off translates into less power usage. I don't think you need more than 12 cores for gaming. 24 on the 13900 seems excessive?
 
Correct me if I'm wrong, but I believe this is for battery efficiency? On a desktop with 1300W power supply I don't think this would make a difference. Efficiency for battery power and eliminating unnecessary cores for gaming might be 2 separate cases to consider. Turning off cores that you don't need for gaming might save you power overall? My testing has proven this. But I am testing Cinebench which uses all cores. Basically each cores turned off translates into less power usage. I don't think you need more than 12 cores for gaming. 24 on the 13900 seems excessive?

If that were the case there would be no efficiency cores on desktop chips.
 
Correct me if I'm wrong, but I believe this is for battery efficiency? On a desktop with 1300W power supply I don't think this would make a difference. Efficiency for battery power and eliminating unnecessary cores for gaming might be 2 separate cases to consider. Turning off cores that you don't need for gaming might save you power overall? My testing has proven this. But I am testing Cinebench which uses all cores. Basically each cores turned off translates into less power usage. I don't think you need more than 12 cores for gaming. 24 on the 13900 seems excessive?

Now that I actually have time to reply properly..

Games and software aren’t coded for “cores” they are coded for threads. Threads can be time critical(Main game loop for example) or not time critical (sound loop for example).

If you have a processor core and you run a non time critical and a time critical thread on it at the same time, then your critical thread won’t be able to run at full speed.

This is where efficiency cores come in, you can run all your non time critical stuff on them and you won’t see a performance drop (at least that is the promise). The same goes for smt threads, your non-critical stuff can run on them.

As for turning off cores you don’t need for gaming, unless you’re a dev of the game you won’t know exactly how many cores you are using. Why not let the built in architecture of the machine and the OS control if cores are on or off with core parking and hardware duty cycling?
 
Last edited:
Now that I actually have time to reply properly..

Games and software aren’t coded for “cores” they are coded for threads. Threads can be time critical(Main game loop for example) or not time critical (sound loop for example).

If you have a processor core and you run a non time critical and a time critical thread on it at the same time, then your critical thread won’t be able to run at full speed.

This is where efficiency cores come in, you can run all your non time critical stuff on them and you won’t see a performance drop (at least that is the promise). The same goes for smt threads, your non-critical stuff can run on them.

As for turning off cores you don’t need for gaming, unless you’re a dev of the game you won’t know exactly how many cores you are using. Why not let the built in architecture of the machine and the OS control if cores are on or off with core parking and hardware duty cycling?
I need to do some testing to see if 24 vs 12 cores makes a difference. I'll start with time spy extreme. Not sure, but I doubt many or if any games would use more than 12 or 16 cores max? The reason for turning off cores that games are not using is to get lower voltage along with higher clock speed keeping the max power draw lower which keeps the temperature lower. Less cores higher clocks. All just speculation though I would need to do a lot of testing.
 
Last edited:
I need to do some testing to see if 24 vs 12 cores makes a difference. I'll start with time spy extreme. Not sure, but I doubt many or if any games would use more than 12 or 16 cores max? The reason for turning off cores that games are not using is to get lower voltage along with higher clock speed keeping the max power draw lower which keeps the temperature lower. Less cores higher clocks. All just speculation though I would need to do a lot of testing.


As mentioned, it’s not cores, it’s threads. It is somewhat trivial to create 3 dozen threads that do various things.

 
Last edited:
This is where my 13900K is usually at during gaming with a DeepCool LS720. Running at stock with XMP enabled. Others with higher temps may want to look at case airflow as well.

I have 13 fans in my case... and they all have good airflow. Bottom and back intake, top and side out.
But it might be my overclock. I have set voltage to 1.390v and it's mildly overclocked to 5.5ghz all p-core and 5.8ghz 2 p-core. E-cores left at 4.3ghz.
 
So I did a little experiment, I swapped my 360 AIO for a Deepcool AK500 I have here, it is a supposed 250W air cooler. Quite a nice unit for the price I might add. I then set the motherboard to PL1 which is a 248w 300amp limit, and it actually does fantastic. I lose zero gaming performance, did drop about 3000 points on cb23 but I don't render anything so I don't care. Hottest core was 87C compared to instantly hitting 100C on all cores at PL3 with the 360 AIO. I ran a 10 minute throttle test before swapping to a single run to get a score.

PL1.png
 
I mounted that bitch like 20 times lol, used a whole tube of paste on it. Tried with and without the aftermarket bracket too.
weird. pump on max? 'cause it shouldnt be doing that poorly...
if youre happy with the air cooler, good. just odd for a 360.
 
weird. pump on max? 'cause it shouldnt be doing that poorly...
if youre happy with the air cooler, good. just odd for a 360.

Oh no, it wasn't doing poorly, PL3 is like 475W limit. It was boosting to 5.8 with a ton of voltage. I just wanted to see what the difference would be if I dialed it back and air cooled it. I originally thought something was wrong but after doing some research its the expected behavior, as long as it hasn't hit 100C it will keep trying to add more sauce.
 
Oh no, it wasn't doing poorly, PL3 is like 475W limit. It was boosting to 5.8 with a ton of voltage. I just wanted to see what the difference would be if I dialed it back and air cooled it. I originally thought something was wrong but after doing some research its the expected behavior, as long as it hasn't hit 100C it will keep trying to add more sauce.
475w!??! lol ok then, i guess it was doing ok, i just misunderstood.
 
475w!??! lol ok then, i guess it was doing ok, i just misunderstood.

Yeah sorry I explained it wrong. I'm looking at a different case so I wanted to see if I can cool it with something smaller and still get good performance.
 
So I did a little experiment, I swapped my 360 AIO for a Deepcool AK500 I have here, it is a supposed 250W air cooler. Quite a nice unit for the price I might add. I then set the motherboard to PL1 which is a 248w 300amp limit, and it actually does fantastic. I lose zero gaming performance, did drop about 3000 points on cb23 but I don't render anything so I don't care. Hottest core was 87C compared to instantly hitting 100C on all cores at PL3 with the 360 AIO. I ran a 10 minute throttle test before swapping to a single run to get a score.

View attachment 566373
Can you please post all your bio setting changes you made to get those numbers please.
 
Can you please post all your bio setting changes you made to get those numbers please.

All I did was select the cooler type, letting it go to water cooling puts the power limit at 4096W. Boxes cooler is 248W.
 
Back
Top