Post your cooling.

[Chrispy_]

I need to find me a 120mm fan to replace it with that's only 14mm thick or less, but I can't find one...

Not as cosmetically appealing but there's nothing stopping you from putting the fan on the outside of the case, and you can shroud from the radiator to the case too. Pull configurations like this love a shroud!

Towers like this with exposed heatpipes sticking out of the top are fantastic for easy and cheap shrouds, just use that hard transparent plastic packaging and then use a hole-punch for the exposed heatpipe ends which anchor it in place and typically offer a good friction fit with your regular hole-punch.

 

[freeagent]

Yes indeed..

https://www.amazon.ca/Thermalright-100700772-Fan-Duct-140/dp/B00O24R9U4

 

[FanlessOC]

Not as cosmetically appealing but there's nothing stopping you from putting the fan on the outside of the case, and you can shroud from the radiator to the case too. Pull configurations like this love a shroud!

I was thinking about putting the fan on the outside or perhaps just mount it on the other side of the cooler, and then make a 12x12cm "tunnel" where the fan currently is. Will see though, not in a hurry, because the cooler is so big that even the 92mm fan is more than enough.

 

[ShrimpBrime]

I was thinking about putting the fan on the outside or perhaps just mount it on the other side of the cooler, and then make a 12x12cm "tunnel" where the fan currently is. Will see though, not in a hurry, because the cooler is so big that even the 92mm fan is more than enough.

Haha, for sure. That cooler is monstrous! Why fix something that isn't broken!?

 

[Waldorf]

@FanlessOC
without any space between fan and case/cooler, mount it on the outside.

P12 pwm

 

[FanlessOC]

Haha, for sure. That cooler is monstrous! Why fix something that isn't broken!?

The only reason for me to go to a 120mm fan would be to make it quieter overall. But I must admit that running it completely fanless would be sweet as well

I also love the fact that I've been using this cooler for more than a decade already, moving it over from build to build... And since I made it fit on AM5 now, it will be in use for years to come.

 

[Athlonite]

 

[ShrimpBrime]

Look, I explained my word choice, I assured you that I meant nothing derogatory, and I apologised for causing confusion. What more do you want?

I want to openly apologize. I shouldn't have called you names, Im sorry for being abrasive.

And in admiration of your tweaking, I wanted to try and get my 14900K around 65w. Just to see really what kind of performance it gives.

Wound up around 2.4ghz All 32 threads and a Cinebench run offers a 68w peak load. VR reported 67.735a peak at 800mv during the run and v-core peak spike is 857mv
The Cinebench23 run (several) Score Averaged about 19634 pts. Max Peak temp with fan at 30% duty Cycle is 46c. HwInfo64 v8.10-5520.
With the Fan Removed, the Peak temp was 78c and average temp 65c. More than 10 runs but of course the heat sink is soaking and there's no "case Flow" to speak of.
At 75c, the v-core started dopping to 778mv. It did not cause any real significance on scores. The temperature over 80c did not effect performance outside margin of error.
First Screen shot is With Fan. And second screen shot is without fan. Wraith Prism HeatSink set atop 14900K no mount.

Edit, yeah. It did wind up around 77w peak. Probably 2.2ghz would do the trick. LLC is auto says lvl 3. Could stay here try Lvl 2 instead.

 

[AusWolf]

I want to openly apologize. I shouldn't have called you names, Im sorry for being abrasive.

No worries. The past is the past.

And in admiration of your tweaking, I wanted to try and get my 14900K around 65w. Just to see really what kind of performance it gives.

Wound up around 2.4ghz All 32 threads and a Cinebench run offers a 68w peak load. VR reported 67.735a peak at 800mv during the run and v-core peak spike is 857mv
The Cinebench23 run (several) Score Averaged about 19634 pts. Max Peak temp with fan at 30% duty Cycle is 46c. HwInfo64 v8.10-5520.
With the Fan Removed, the Peak temp was 78c and average temp 65c. More than 10 runs but of course the heat sink is soaking and there's no "case Flow" to speak of.
At 75c, the v-core started dopping to 778mv. It did not cause any real significance on scores. The temperature over 80c did not effect performance outside margin of error.
First Screen shot is With Fan. And second screen shot is without fan. Wraith Prism HeatSink set atop 14900K no mount.

Edit, yeah. It did wind up around 77w peak. Probably 2.2ghz would do the trick. LLC is auto says lvl 3. Could stay here try Lvl 2 instead.

View attachment 399248 View attachment 399249
View attachment 399250

First of all, how did you manage to fit a Wraith Prism on top of a 14900K?

Secondly, those are awesome results!
I haven't tested my 7800X3D in Cinebench since I applied the TDP limit, but I will - I just have to pop over into Windows, and I'm too lazy to do that right now, after work.
At factory settings, it does around 18k points while consuming roughly 80 W. It does so at around 4.8 GHz, so I have high hopes of my tuned settings, too.

Another weirdness about power limiting Ryzen is the SoC. Higher RAM speed needs a higher SoC voltage and MC clock, which results in the IO die eating more power, and leaving less for the cores. With 4800 MHz RAM and 1 VSOC, the CPU does 4.4 GHz while crunching BOINC. With 6000 MHz RAM and 1.3 VSOC, it hovers between 4 and 4.1 GHz. This will definitely affect my Cinebench score more than my RAM speed.

 

[Chrispy_]

And in admiration of your tweaking, I wanted to try and get my 14900K around 65w. Just to see really what kind of performance it gives.

It's probably quite good.

IIRC W1zzard did an Alder Lake power scaling article with the 12900K. 75W was about 65% of the performance of 241 so it's definitely pretty usable at lower power levels, I don't know if Intel's node/yields changed much between 12th gen and 14th gen, and there's only one way to find out

 

[ShrimpBrime]

No worries. The past is the past.


First of all, how did you manage to fit a Wraith Prism on top of a 14900K?

Secondly, those are awesome results!
I haven't tested my 7800X3D in Cinebench since I applied the TDP limit, but I will - I just have to pop over into Windows, and I'm too lazy to do that right now, after work.
At factory settings, it does around 18k points while consuming roughly 80 W. It does so at around 4.8 GHz, so I have high hopes of my tuned settings, too.

Another weirdness about power limiting Ryzen is the SoC. Higher RAM speed needs a higher SoC voltage and MC clock, which results in the IO die eating more power, and leaving less for the cores. With 4800 MHz RAM and 1 VSOC, the CPU does 4.4 GHz while crunching BOINC. With 6000 MHz RAM and 1.3 VSOC, it hovers between 4 and 4.1 GHz. This will definitely affect my Cinebench score more than my RAM speed.

The cooler fits, there's just no mount. Gravity mounted!

So you limited power, I limited frequency. Left all the voltage stuff auto. Intel handles it ls power delivery pretty good. But don't think it'll go lower than 800mv except for droop. If I want lower v-core, do a negative offset. But because of IO, I'm not sure it'll run. Something to test I suppose.

Maybe there's a happy place between 4800 and 6000mhz where the SOC doesn't gobble up all the power.

It's probably quite good.

IIRC W1zzard did an Alder Lake power scaling article with the 12900K. 75W was about 65% of the performance of 241 so it's definitely pretty usable at lower power levels, I don't know if Intel's node/yields changed much between 12th gen and 14th gen, and there's only one way to find out

I can very much see the single core performance hit. 2.4ghz is pretty slow lol

 

[AusWolf]

The cooler fits, there's just no mount. Gravity mounted!

Wow, that's mad!

How about the mounting pressure? Maybe you could zip tie it to the motherboard or something for better results?

So you limited power, I limited frequency. Left all the voltage stuff auto. Intel handles it ls power delivery pretty good. But don't think it'll go lower than 800mv except for droop. If I want lower v-core, do a negative offset. But because of IO, I'm not sure it'll run. Something to test I suppose.

I limited power because that way, the CPU still boosts to 5 GHz in single-threaded work, and there's no performance loss in games.

I just tested it with the 61 W power limit - roughly 16.5k points in Cinebench R23 with 6000 MHz RAM and 1.3 VSOC. Not bad compared to an 18k stock.

Although your test shows that more massive CPUs with more cores scale better with power. Maybe I'll try my luck with a dual-CCD Zen 6 next when it comes out. I'm curious.

Maybe there's a happy place between 4800 and 6000mhz where the SOC doesn't gobble up all the power.

Possibly. I know it still works with 1.2 VSOC, although that doesn't give me too much of a difference in my results. I could tune some more, but I can't be asked. Better to spend my time playing games.

I can very much see the single core performance hit. 2.4ghz is pretty slow lol

Try a power limit instead. That way, you still wouldn't lose boost in lightly threaded work.

 

[ShrimpBrime]

Wow, that's mad!

How about the mounting pressure? Maybe you could zip tie it to the motherboard or something for better results?


I limited power because that way, the CPU still boosts to 5 GHz in single-threaded work, and there's no performance loss in games.

I just tested it with the 61 W power limit - roughly 16.5k points in Cinebench R23 with 6000 MHz RAM and 1.3 VSOC. Not bad compared to an 18k stock.

Although your test shows that more massive CPUs with more cores scale better with power. Maybe I'll try my luck with a dual-CCD Zen 6 next when it comes out. I'm curious.


Possibly. I know it still works with 1.2 VSOC, although that doesn't give me too much of a difference in my results. I could tune some more, but I can't be asked. Better to spend my time playing games.


Try a power limit instead. That way, you still wouldn't lose boost in lightly threaded work.

With power limits, what you see is Not what you get under load. The effective clocks will just droop down to the power level. Say like 4.4ghz with a 180w limit even though cpu-z reports a higher clock rate.

Mounting pressure doesn't seem to make a difference. Can use 253w restricted like this.

 

[AusWolf]

With power limits, what you see is Not what you get under load. The effective clocks will just droop down to the power level. Say like 4.4ghz with a 180w limit even though cpu-z reports a higher clock rate.

Sure, but what I mean is, a power limit leaves more headroom for individual cores during lightly threaded work.

In my example, let's say that my cores need 8 W each under full load, and my IO die needs 20 W all the time. That's 8 x 8 = 64 + 20 = 84 W for the whole CPU.
With a 60 W power limit, my IO die still needs 20 W, so I have 40 W left for my cores, or 5 W per core (instead of the stock 8 W). But when only a single core is loaded, the CPU has no problem giving the full 8 W, or even more to that one core.

It's a rough example, but sort of paints the picture.

Mounting pressure doesn't seem to make a difference. Can use 253w restricted like this.

That's interesting.

Why are you using a Wraith Prism with such a high end CPU, if you don't mind me asking?

 

[Chrispy_]

Why are you using a Wraith Prism with such a high end CPU, if you don't mind me asking?

It's seen some things, too:

 

[ShrimpBrime]

Sure, but what I mean is, a power limit leaves more headroom for individual cores during lightly threaded work.

In my example, let's say that my cores need 8 W each under full load, and my IO die needs 20 W all the time. That's 8 x 8 = 64 + 20 = 84 W for the whole CPU.
With a 60 W power limit, my IO die still needs 20 W, so I have 40 W left for my cores, or 5 W per core (instead of the stock 8 W). But when only a single core is loaded, the CPU has no problem giving the full 8 W, or even more to that one core.

It's a rough example, but sort of paints the picture.


That's interesting.

Why are you using a Wraith Prism with such a high end CPU, if you don't mind me asking?

I used that to be sure the system works after DryIce runs, just plop it on there. But for whatever reason, it handles thermals pretty well. Depending on what the load is, I can do about 310w before it instant 100c gibs.

I'll have to try power restriction sometime. Usually I raise power limits lol.

It's seen some things, too:

View attachment 399361

yeah, it might be just an old opteron heatsink. Which is identicle to a wraith. Or it's an FX cooler. I think the Opteron coolers where plated at the bottom, this one is heatpipes directly to IHS plate. I had 2 handfuls of this style AMD coolers at one time. All in a pile in a drawer haha! (Yes, I was a big AMD guy till about 2017)

 

[Chrispy_]

yeah, it might be just an old opteron heatsink. Which is identicle to a wraith. Or it's an FX cooler. I think the Opteron coolers where plated at the bottom, this one is heatpipes directly to IHS plate. I had 2 handfuls of this style AMD coolers at one time. All in a pile in a drawer haha! (Yes, I was a big AMD guy till about 2017)

It's 100% a stock FX cooler, I bought FX chips for a while for ECC support when Intel were forcing you to buy expensive Xeons to get ECC.

IIRC one of them is still running as a distributed job manager for Deadline/VRay in the London office. Shit, but still hasn't caught fire and died so it's good enough to avoid needless replacement.

 

[ShrimpBrime]

It's 100% a stock FX cooler, I bought FX chips for a while for ECC support when Intel were forcing you to buy expensive Xeons to get ECC.

IIRC one of them is still running as a distributed job manager for Deadline/VRay in the London office. Shit, but still hasn't caught fire and died so it's good enough to avoid needless replacement.

Lol. I love these heatsinks though. The Wraith fans clip right on. That's a Stealth fan I think. I just yank it from a drawer and Schlaaap her on there! I've cooled a very many cpus with them. It's better than a hyper 212 all day