News Core i9-13900K Cooling Guide: Testing Intel's Flagship With Budget Air and Big AIOs

[-Fran-]

You only get an heat increase of the CPU if the cooler isn't capable of transferring it away from the CPU.
If you look at a 200W power limited review and use a 200W cooler at 100% you will never get any hotter than you already are, it will be able to constantly draw away the full 200W of heat that the CPU creates.
So yes, it is.

You seem to not understand how semiconductors and heat work in relation to power. That's fine.

If a test is conducted, you'll understand.

Regards.

 

[dalauder]

I'd love to see how coolers compare on Intel vs Ryzen, unfortunately I don't have a Ryzen system to test with at the moment. ...

I feel modern Ryzen 7xxx CPUs will be more impacted than past generations by limited cooling, but given the CPU's lower overall wattage potential there shouldn't be as much potential performance lost from throttling. ...

I don't understand what you mean by PCI-e lanes here. How is this relevant to cooling?

Thanks for the investigation. You can't do a Ryzen system if you don't have one.

You might be right about the 7000 series throttling. I still want to see it, especially when the Alder Lake vs 5000 series was never investigated when cooling was limited.

For PCI-e lanes, I meant that Intel CPUs + Chipsets have consistently exposed fewer PCI-e lanes than AMD. That's one big reason I built with AMD for my last desktop. That, and I like how they don't artificially lock features as much.

I should acknowledge that I've got some skin in the game and I'm hoping AMD's stock prices go up. But I rooted for them when I didn't hold any shares.

 

[JamesJones44]

You only get an heat increase of the CPU if the cooler isn't capable of transferring it away from the CPU.
If you look at a 200W power limited review and use a 200W cooler at 100% you will never get any hotter than you already are, it will be able to constantly draw away the full 200W of heat that the CPU creates.
So yes, it is.

It is different, especially with burst speeds. If you limit a Raptor Lake CPU to 205 watts and have a crappy cooler the processor may still peak at a full 205 watts for short period of time. Same as if you leave it at 253 watts. However, if you have great cooling those little short burst before throttle can last a lot longer before being told to drop down. 205 watts with a bum cooler will look worse than 205 watts with a good cooler. The question is how far down does that scale with a good cooler? Does a good cooler make a difference when capped at 65 watts? Fiddling with the power requirements won't tell you that.

 

[TerryLaze]

It is different, especially with burst speeds. If you limit a Raptor Lake CPU to 205 watts and have a crappy cooler the processor may still peak at a full 205 watts for short period of time.

Yeah but I was talking about a cooler that can sustain the full load (that you choose) forever.
If you have a crappy cooler that can still cool 100W 24/7 then you can still run your CPU at 100w 24/7.
If you can cool 200W then you can run it at 200W 24/7.

And actually no matter what cooler you have, if you don't lock the power limit, then the CPU automatically hits the max clocks that the cooling allows within the max temp that the CPU allows, but then you always run at max temp ie 95 for ryzen and 100 for intel.

What you are talking about is taking the TDP number from the front of the box ,that doesn't mean much if you don't use the settings that make up that TDP number, and arguing that the PL2 boost of TAU sec will overpower the cooling.

However, if you have great cooling those little short burst before throttle can last a lot longer before being told to drop down.

This is relevant for PL2 boost with TAU but is irrelevant if you set a max power draw/lock PL1=Pl2. With a locked limit and during high load you will always have the same power load until the workload changes and if your cooling can handle that amount then it can handle that amount.

Locked limit means that the power limit for short bursts (Pl2) is the same as for long durations (PL1) so there is no difference for short bursts that a better cooler could help with, you hit the limit right away, IF the workload is heavy enough to reach it.

 

[Albert.Thomas]

It's the same thing as long as you know how much cooling a cooling solution can cool.
Looking up cooler reviews might be more helpful than searching for CPU reviews with certain coolers.

Not quite. Cooling 200W on a 13900K is easy. Cooling 200W on a 12900K doable, but harder than a 13900K. Cooling 200W on a i5-13600K is harder than doing so on both of the i9s.

 

[TerryLaze]

Not quite. Cooling 200W on a 13900K is easy. Cooling 200W on a 12900K doable, but harder than a 13900K. Cooling 200W on a i5-13600K is harder than doing so on both of the i9s.

Wanna give us a reason?!
What, because of the different sizes, heat densities, IHS material, tim?

That would just mean that a cooler that can cool 200W on one CPU might cool less W on a different CPU, but if the CPU can cool 200W on this CPU then it is still 200W on this CPU.

 

[Albert.Thomas]

Wanna give us a reason?!
What, because of the different sizes, heat densities, IHS material, tim?

It mainly comes down to thermal density. Running 200W through 6P+8E is harder than running 200W through 8P+8E or 8P+16E.

 

[-Fran-]

View: https://www.youtube.com/watch?v=zztbIOhJLuk


And Gordon forgot to double check there was contact and the 7950X still ran XD

Regards.

 

[Max2thous10]

Hi, I was interested in your iCue i50i Elite review; it seems to suggest the thermal paste was not well applied . Would it be better to go with some fresh TF7 thermal paste from a tube to get the best heat transfer on the first application ? Thanks for looking

 

[the_m0narch]

Thanks for an interesting piece of research.

It's great to read article with less single minded focus on the extra percent performance (which is industry driven).

A nice addition next time would be to have the case rigged for air cooling. AIO deliver the same regardless off airflow in the case, but the air cooling ones can be affected by the air flow for better or for worse.

Case fans at 30% isn't ideal for that and silent ones could be running at far greater speed without being nearly as noisy as an AIO Liquid cooler. That said, the case used are probably better than most cases in its default state with 3x120 in and 2x200 out.

A while back Linus compared air cooling with AIO Liquid and pretty much convinced me to return from AIO cooling to air on my last build. I know that the discussion isn't finished, but in regard to noise I believe air is the way to go.

Aiming for best performance/noise ratio I went with noctua NH-D15 CPU fan and rigged the whole case with noctua silent fans (noctua nf-a12x25 120 fans) and it's been working great for me. Pull air from front and front bottom and push out at the back and top. As I recall it, I have 5 120 fans that pulls and 5 that push out. Didn't find much evidence for real life benefits of "overpressure". Writing it down I agree it's a ridiculous number of fans, but buying them at sale price it was worth it for me. More fans means they run at slower speeds with less nois. Furthermore if you have a good air flow system, air cooled graphic cards will benefit from it.

Don't have any data to share, but I won't be returning to AIO Liquid cooling anytime soon.

 

[bit_user]

I believe air is the way to go.

Aiming for best performance/noise ratio I went with noctua NH-D15 CPU fan and rigged the whole case with noctua silent fans (noctua nf-a12x25 120 fans) and it's been working great for me. Pull air from front and front bottom and push out at the back and top. As I recall it, I have 5 120 fans that pulls and 5 that push out.

Thanks for sharing your experience. What CPU & GPU? Overclocked or stock?

And yes, 10 case fans is rather... 😮

Didn't find much evidence for real life benefits of "overpressure".

The main reason to use it is if you have dust filters on all of your intake fans. Then, positive-pressure will help keep dirty air from leaking in via other cracks or unfiltered holes. I haven't found much independent testing to support this (Linus' test methodology is bad, IMO), but my personal experience has shown it to be very worthwhile. Perhaps it depends a lot on the case or its filters... mine are Lian Li.

 

[DCB2020]

If you are recommending limiting the power then you had better compare to other high-end chips like the amd 7950 or the intel 13700k. At the end of the day, this is a very expensive chip targeted for the professional (e.g., studio production, rendering, etc) and if you are going to limit the capability you better to be able to show how it is economically viable.

 

[DCB2020]

Gaming is a workload that's fairly easy to cool, you could subsitute the 125w numbers for games for the most part.



I've tested using both the default ILM and the Thermalright contact frame, and on this motherboard it actually doesn't make a difference at all. See: Thermalright LGA1700-BCF Contact Frame Review : Can it tame Raptor Lake’s heat?



Like this?

I didn't include the numbers for 125w and 200w for the different coolers because there won't be much variation at the same power limits. You'll see slightly higher numbers with the better coolers, but barely so.

What is the point of limiting the capability chip that his target market and price point. The 13700k or AMD 7950x would be a better choice if you need to throttle a $600 CPU