Passively Cooling the Intel i9-9900K

cageymaru

Fully [H]
Joined
Apr 10, 2003
Messages
22,061
Der8auer on YouTube has experimented with passively cooling an Intel i9-9900K with the ARCTIC Alpine 12; a passive CPU cooler. The ARCTIC Alpine 12 is only rated to handle 47 watts so Der8auer wasn't expecting much out of the unit. Although the passive cooler showed that it was more capable than its rating, it couldn't keep the Intel i9-9900K properly cooled at stock settings. The Intel chip was throttling, so Der8auer ended up with a stable 3.6 GHz clock speed across all cores which was more than capable of playing games on the system.

I think we could go even higher to 3.8 GHz @0.975 V. Yes, you can actually passively cool a 9900K with some kind of adjustments. You have to undervolt your CPU a little bit; underclock your CPU a little bit.
 
As an Amazon Associate, HardForum may earn from qualifying purchases.
I don't know, is it a good idea to set your pants on fire?
 
Actually I think it's a good demo of modern thermal/power management. Not to take away from Arctic, which have made a great passive HS, or Der8auer for running the experiment. But you are putting a chip in a situation where it's bound to fail, and you see the power management manages to keep it running (err... limping) without breaking.

I would like to see the same experiment run with a TR... I imagine it would do much the same - it would run, severely throttled, but it shouldn't break. I know Arctic was talking about a "semi-passive" TR HSF. No practical application here, purely for S&Gs.

Remember 15 years ago and trying this would end up with a fried CPU.
 
Actually I think it's a good demo of modern thermal/power management. Not to take away from Arctic, which have made a great passive HS, or Der8auer for running the experiment. But you are putting a chip in a situation where it's bound to fail, and you see the power management manages to keep it running (err... limping) without breaking.

I would like to see the same experiment run with a TR... I imagine it would do much the same - it would run, severely throttled, but it shouldn't break. I know Arctic was talking about a "semi-passive" TR HSF. No practical application here, purely for S&Gs.

Remember 15 years ago and trying this would end up with a fried CPU.
Personally I think it's a great example to show that Intel is seriously pushing the tech beyond where it is designed to run, and this shows where it ought to be running.
My 2 cents.

And actually, friend's Celeron 333 Slot 1... I shit you not... was a freak of nature and ran without a heatsink. He gave it to his ex wife after they divorced and she ran it for awhile, then asked me to look at it because it was slow. I brought it home and nearly shit myself when I took the side panel off... I actually still have the CPU, but lack a board to even try it in.
 
Last edited:
  • Like
Reactions: N4CR
like this
Personally I think it's a great example to show that Intel is seriously pushing the tech beyond where it is designed to run, and this shows where it ought to be running.
My 2 cents.

“Ought to be” is somewhat subjective. I would say for an enthusiast class desktop you want high clocks, TDP be damned. Which is what intel did here. They even redefined TDP so they could do so... (not that I agree with that personally)

The same silicon intended for a data center, or a HTPC, or a laptop, or a mobile - those are all going to have different priorities with respect to power draw, clock speed, core counts, etc. This chip isn’t marketed for any of those, but does illustrate that “ought to be” changes depending on application.

No pulling away from the fact that this is still Skylake and it’s still not 10nm...
 
“Ought to be” is somewhat subjective. I would say for an enthusiast class desktop you want high clocks, TDP be damned. Which is what intel did here. They even redefined TDP so they could do so... (not that I agree with that personally)

The same silicon intended for a data center, or a HTPC, or a laptop, or a mobile - those are all going to have different priorities with respect to power draw, clock speed, core counts, etc. This chip isn’t marketed for any of those, but does illustrate that “ought to be” changes depending on application.

No pulling away from the fact that this is still Skylake and it’s still not 10nm...
You have a point heh
 
Personally I think it's a great example to show that Intel is seriously pushing the tech beyond where it is designed to run, and this shows where it ought to be running.
My 2 cents.

And actually, friend's Celeron 333 Slot 1... I shit you not... was a freak of nature and ran without a heatsink. He gave it to his ex wife after they divorced and she ran it for awhile, then asked me to look at it because it was slow. I brought it home and nearly shit myself when I took the side panel off... I actually still have the CPU, but lack a board to even try it in.

It was no freak of nature, I ran several of the old 266/300/333's without heatsinks.
 
It was no freak of nature, I ran several of the old 266/300/333's without heatsinks.

And... many of those CPUs ran at a near 100% overclock. I had one. So did most of my friends.

Ahh, LAN parties and Total Annihilation. You needed the CPUs overclocked to hell just to keep a game of TA running with 4 people after an hour of base building.
 
TLDW; Probably would have done far better if it was in a case, and there was proper airflow.
Still dont see it as a good idea though.
 
And... many of those CPUs ran at a near 100% overclock. I had one. So did most of my friends.

Ahh, LAN parties and Total Annihilation. You needed the CPUs overclocked to hell just to keep a game of TA running with 4 people after an hour of base building.

Core Contingency was the expansion we would LAN up or Code Name Eagle. That shit was a buggy mess.

Was not a fan of TA Kingdoms...
 
Actually I think it's a good demo of modern thermal/power management. Not to take away from Arctic, which have made a great passive HS, or Der8auer for running the experiment. But you are putting a chip in a situation where it's bound to fail, and you see the power management manages to keep it running (err... limping) without breaking.

I would like to see the same experiment run with a TR... I imagine it would do much the same - it would run, severely throttled, but it shouldn't break. I know Arctic was talking about a "semi-passive" TR HSF. No practical application here, purely for S&Gs.

Remember 15 years ago and trying this would end up with a fried CPU.


15 years ago, Zalman had a beautiful case to passively cool every component, including a powerful GPU. The TNN 500A was pretty beastly. Here's a review from THG back in '04
 
I recon if he put the motherboard in the right case it could actually do a little better if case fans was able to make sure there was no stagnant air.
On the table he are pretty much just relying on convection / hot air rising, so what little active air outside of that should in theory better results.
 
Actually I think it's a good demo of modern thermal/power management. Not to take away from Arctic, which have made a great passive HS, or Der8auer for running the experiment. But you are putting a chip in a situation where it's bound to fail, and you see the power management manages to keep it running (err... limping) without breaking.

I would like to see the same experiment run with a TR... I imagine it would do much the same - it would run, severely throttled, but it shouldn't break. I know Arctic was talking about a "semi-passive" TR HSF. No practical application here, purely for S&Gs.

Remember 15 years ago and trying this would end up with a fried CPU.

I'd be shocked if the TR self destructed. A year or two after the notorious THG video linked elsewhere AMD also added massive throttling to prevent a burnup to their CPUs gaining feature parity with Intel.
 
I would like to see the same experiment run with a TR... I imagine it would do much the same - it would run, severely throttled, but it shouldn't break


LOL.. i can confirm, on accident, it works well.


One day on my water cooled TR1900x while gaming and running BOINC, with Task Manager up and running i was noticing it was running down at the 2Ghz range. wondering WTF?!?!?!?! I then noticed that 2 of the fans on the radiator had died, with one of them barely turning, the one i could see.. but the one on the back completely frozen up

shut the system down, replaced the fans, and then back up and running at 4Ghz on all cores, without issue


it was interesting to see how well the CPU could down throttle to keep it cool enough to still run, even under load hahahaha
 
Actually I think it's a good demo of modern thermal/power management. Not to take away from Arctic, which have made a great passive HS, or Der8auer for running the experiment. But you are putting a chip in a situation where it's bound to fail, and you see the power management manages to keep it running (err... limping) without breaking.

I would like to see the same experiment run with a TR... I imagine it would do much the same - it would run, severely throttled, but it shouldn't break. I know Arctic was talking about a "semi-passive" TR HSF. No practical application here, purely for S&Gs.

Remember 15 years ago and trying this would end up with a fried CPU.
That heatsink is nothing groundbreaking at all. It's essentially a scaled up chipset heatsink and probably costs pennies to manufacture.

As many people have pointed out, there's been plenty of decent passive cooling tech in the past, and frankly, Arctic Cooling's product is flat out pathetic.

Let's see some genuine passive coolers actually rated for high TDP's give it a go:

NoFan CR-95C

Zalman FX100

Silverstone Heligon HE02
 
Never understood the whole passive cooling thing on desktop units outside of the gimmick aspect of it. Need at least 1 fan to white noise out the coil whine. Found that out on my current set up which is water cooled and the rads/pump are about 10 ft away from the rig in the closet (ran tubing thru the wall). Had to slap one small fan on it to mask the coil whine.
 
Back
Top