TR Forums

I’ve been pretty strongly against CLCs since I had a Corsair failed pump on one, think it was an H80. A while back when I was all about the toaster size SFFs.

You have to do CLC in those tiny cases because you’d have a PSU hanging over where the socket is.

The main gripe I have with CLCs is actually the pump noise. You can get nice noctuas for your HSF or radiator, but the pump you can never do anything about quieting down. And the pump is one more thing contributing to the noise.

These days you can get a high efficiency PSU and have that contribute very little to overall noise, and you can easily get control over other fan speeds.

One of my first-generation Corsair H80s is still going strong in a Temjin TJ08-E. The other one died due to pump failure.

I have a huge NH-D15 on my Core i7-9700K. I've been eyeing an EK-AIO 360 for my next PC.

The difficult thing about the pump noise-wise is you've got to isolate the vibration it creates. Most of the good ones don't make much noise on their own, but if the vibration travels into a drive cage or something, then yeah that can get annoying.

I've got an H7 Plus in this machine on an i7 8700K, and I'll probably stick with some other whacking huge air cooler in my next upgrade. I've been eyeballing the 3950X, or maybe one of the new Threadripper chips. So far I've been successful in convincing myself that Threadripper is overkill for my current use cases, but there are things I could do with it. I might need one..... eventually.

jmc2 wrote:

The odd bit is that if I leave the stress test on for a minute the AUTO voltage drops to 1.25v and the temps drop from low 80sC to high 70sC..

I am pretty certain that these high temps are not consistent with NH-D15's cooling capacity. It's not like that chip can put out 300W. It can't. Maybe 200 if you push it way too hard with OC.

You should re-mount that heatsink and check how well it is contacting. What thermal compound are you using? And what are your HSF RPM's?

The heat spreader is soldered on these AMD chips.

Redocbew wrote:

I've got an H7 Plus in this machine on an i7 8700K, and I'll probably stick with some other whacking huge air cooler in my next upgrade.

I'm addicted to low noise. I've been contemplating going for one of those 3 slot monster GPU air coolers, but I haven't been gaming much at all lately and it hasn't been a concern for me.

APWNH wrote:

jmc2 wrote:

The odd bit is that if I leave the stress test on for a minute the AUTO voltage drops to 1.25v and the temps drop from low 80sC to high 70sC..

I am pretty certain that these high temps are not consistent with NH-D15's cooling capacity. It's not like that chip can put out 300W. It can't. Maybe 200 if you push it way too hard with OC.
You should re-mount that heatsink and check how well it is contacting. What thermal compound are you using? And what are your HSF RPM's?
The heat spreader is soldered on these AMD chips.

Yeah, I've thought of reseting the heatsink but find that to be a high stress process for me.

I'm happy with my "mainly" 70sC temps @4Ghz and 0.0125 voltage offset, CPUZ Bench_78C/26C_ambient.
Tried a 0.0250v offset and CPUZ hits the 90sC...Not having that!

Kryonaut paste and three 3000 RPM Noctua fans.(They start ramping up @73C).
Still working on the daily evening reboot that it does.

There is the odd thing with the unchanging cpu voltage(1.09999v) on booting. I have to open RyzenMaster and click on "Reset".
Rmaster says it did reset (something unknown) and NOW the voltages will change with load. Even with the bios voltage on AUTO it still boots STUCK at 1.09999v until a Rmaster "reset". Never read anything about this before.

So close but still HAVE to run Rmaster after every boot. (I turn my pc off at night)

thanks.

jmc2 wrote:

Yeah, I've thought of reseting the heatsink but find that to be a high stress process for me.

Ya I hear ya. Probably only after doing it 5 to 10 times did I start to not dread that process and actually look forward to it...

jmc2 wrote:

I'm happy with my "mainly" 70sC temps @4Ghz and 0.0125 voltage offset, CPUZ Bench_78C/26C_ambient.
Tried a 0.0250v offset and CPUZ hits the 90sC...Not having that!

Kryonaut paste and three 3000 RPM Noctua fans.(They start ramping up @73C).

A great rule of thumb that is founded in scientific principles that I like to use is that you can get a feel for the effectiveness of a given heatsinking setup: that is to say, a measure of its ability to cool: is to try to get a good real-time read on core temperature, and then watch the temp drop behavior after you stop the load on the CPU. If you have a solid connection to the heatsink that temperature should drop like a rock and then slow down and converge toward the idle temperature. It'll more or less follow an exponential curve. So, open up task manager so u can watch the overall CPU load number, you can even break out your stopwatch and see what the core temp under load is, stop the stress test load as you start the stopwatch, and then look at how far that temperature drops at the 5 second mark. If you have a large cooler, like you do, and if you're getting like you're reporting 78 degrees C, if it is indeed making good contact it should drop very quickly, on the order of at least 20 degrees C into the 50's, after a very short amount of time. If the temperature drop is sluggish, that indicates maybe it could benefit from a redo.

Similarly, (and probably a lot easier, tbh) is look at idle temps, try to get a sample of idle temps reported by other users, etc. usually if you have good case airflow it should be 40 C or below... also idle temps should correspond well between different chips since they'll be sipping power at idle. UNLESS, you have an all core OC that is constant and you've switched off dynamic clocking. Which is, I mean some people like to do that. Actually I have done that with my 8700K, it's pegged at 5Ghz 24/7... But I keep it cool, delidded and with liquid metal on a NH-D15.

Still working on the daily evening reboot that it does.

There is the odd thing with the unchanging cpu voltage(1.09999v) on booting. I have to open RyzenMaster and click on "Reset".
Rmaster says it did reset (something unknown) and NOW the voltages will change with load. Even with the bios voltage on AUTO it still boots STUCK at 1.09999v until a Rmaster "reset". Never read anything about this before.

So close but still HAVE to run Rmaster after every boot. (I turn my pc off at night)

thanks.

Pretty interesting, that sounds quite frustrating to be honest. I think you should try to set aside a weekend day and find a good reputable AMD Zen 2 or 3950x specific bios OC guide, preferably one that uses your exact motherboard, reset your BIOS settings, and try to get it clean. I'd recommend starting from a known good state, e.g. stock settings, possibly with PBO (the thing that boosts voltages automatically? right?) off, etc. Make sure that it is stable at a baseline configuration. Maybe you have something going on with your OS install. Maybe you should run memtest to make sure your memory isn't faulty and causing the weirdness. And you can do this over a few days too, keep a notebook of your settings you changed and stuff. Good luck

I might be off-base about my suggestions since upon re-reading your issue may be something specific, so it could be something that maybe you can upgrade the BIOS for, or maybe there's some specific setting that could be causing it. Or some software problem. I would try to get to a "stock" state and start from there (for sanity).

APWNH wrote:

I am pretty certain that these high temps are not consistent with NH-D15's cooling capacity. It's not like that chip can put out 300W. It can't. Maybe 200 if you push it way too hard with OC.

It's because the chiplets are tiny and the energy density is very high. Even under a great watercooling system capable of nearly a kilowatt of heat load I can get mine into the 70 C range if I push it hard.

You know that makes me wonder if there is a benefit to mounting the heatpipes going one way instead of the other, depending on the layout of the chiplets underneath. It'd be suboptimal if only half the heatpipes can be placed directly above the core chiplets (which I presume will generate far more heat than the I/O die)

Probably not, unless there's a big pocket of impurities, or something else which causes weird heat distribution in the contact plate of the heatsink its self. If there is, then you're probably going to be having other, bigger problems anyway.

Yes it should be a totally negligible difference, the entire volume of the blocks will be diffusing the heat so the important part is just having as many heat pipes as you can get in there.

APWNH wrote:

You know that makes me wonder if there is a benefit to mounting the heatpipes going one way instead of the other, depending on the layout of the chiplets underneath. It'd be suboptimal if only half the heatpipes can be placed directly above the core chiplets (which I presume will generate far more heat than the I/O die)

It can help a little, but having more heatpipes in the proper orientation to help them phase change / wick back properly is likely more important. The heatspreaders works well to get rid of hotspots, so as long as you have the entire heatspreader in contact with your sink, you're generally getting pretty close to optimal dispersion.

APWNH wrote:

jmc2 wrote:

Similarly, (and probably a lot easier, tbh) is look at idle temps, try to get a sample of idle temps reported by other users, etc. usually if you have good case airflow it should be 40 C or below... also idle temps should correspond well between different chips since they'll be sipping power at idle. UNLESS, you have an all core OC that is constant and you've switched off dynamic clocking. Which is, I mean some people like to do that. Actually I have done that with my 8700K, it's pegged at 5Ghz 24/7... But I keep it cool, delidded and with liquid metal on a NH-D15.

Pretty interesting, that sounds quite frustrating to be honest. I think you should try to set aside a weekend day and find a good reputable AMD Zen 2 or 3950x specific bios OC guide, preferably one that uses your exact motherboard, reset your BIOS settings, and try to get it clean. I'd recommend starting from a known good state, e.g. stock settings, possibly with PBO (the thing that boosts voltages automatically? right?) off, etc. Make sure that it is stable at a baseline configuration. Maybe you have something going on with your OS install. Maybe you should run memtest to make sure your memory isn't faulty and causing the weirdness. And you can do this over a few days too, keep a notebook of your settings you changed and stuff. Good luck

----"If you have a solid connection to the heatsink that temperature should drop like a rock and then slow down and converge toward the idle temperature. It'll more or less follow an exponential curve."...

Just tested and the RMaster temp graph shows it to fall like it rises. A sharp fall from 80C to 73C then a taper to ambient.

----"UNLESS, you have an all core OC that is constant and you've switched off dynamic clocking."...

I've only turned off "Core Performace Boost" and that stops(1.5v) & all xfr or any other automatic speed boost that will take it ABOVE what is set.
So it will idle with one or two cores at 400 Mhz(yes Mhz) and the rest asleep but then boost up to the max of 4Ghz set in bios when needed.

----"reset your BIOS settings, and try to get it clean. I'd recommend starting from a known good state, e.g. stock settings, possibly with PBO (the thing that boosts voltages automatically? right?) off, etc. Make sure that it is stable at a baseline configuration."

Yes, ouch, I agree with you that a clean bios reset and careful note taking is in order. Really not looking forward to that!

Oh, PBO, have never touched that at all because I have seen 95C before when new and @ stock with a stress test. Afraid to see where PBO could take me.
Think that would be a water cooling only option.

Thanks.

APWNH wrote:

You know that makes me wonder if there is a benefit to mounting the heatpipes going one way instead of the other, depending on the layout of the chiplets underneath. It'd be suboptimal if only half the heatpipes can be placed directly above the core chiplets (which I presume will generate far more heat than the I/O die)

Here is a "Gamers Nexus" test on exactly that.

https://www.youtube.com/watch?v=O1WGlPrH4IU

And a "der8auer OC Bracket" that shifts the heatsink more directly over the cpu hot spot.(which I can't buy yet)
https://www.youtube.com/watch?v=JlJDv68fEcM

TL;DR:

Makes not a blimey bit of difference as that is what the heat spreader on the chip package is already there for...

I wonder if maybe your cooler and heat spreader aren't getting good contact. If one or the other isn't perfectly flat, there may be some bad contact that can impede cooling. JayZ has a video a couple of years ago where his CPU had a very slight bulge in the heat spreader that caused poor cooling (he was actually able to spin it like a top on a desk): after lapping the heat spreader he managed to drop ~20 degrees C off the temps.

Lapping heatspreaders is almost always beneficial to some degree.

Waco wrote:

Lapping heatspreaders is almost always beneficial to some degree.

and @K-L-Waster

I remember lapping a heatsink to make it flatter and then finding out that it was supposed to be a little convex due to the Intel heatsinks of that time being a little concave....felt a bit dumb there as I had not checked BOTH surfaces for flatness.

When I reseat my Noctua I'll do it right this time.

Buildzoid used liquid metal on a heatsink and got good results but he slowly increased the amount of metal over several applications to make sure it did not run. You do want a nickle plated surface or it will "dryout"/react with bare copper on the first application and need to be redone. Second application should be fine.
If my motherboard was horizontal I would use it but vertical, that's just too scary. Just this week or last week a hardware website showed a cpu with it running out from under the cpu heatspreader.

thanks.