Folding Forum

Interesting...

An 18 hour lull in GPU WUs encouraged me to look at CPU cooling again. The video suggested by TxRedneck gave me an idea that perhaps the cooler was snagging on the socket header and therefore not making good flat contact with the CPU. The stock AM4 cooler mounting arrangement doesn't give much flexibility but I pushed the cooler as far away from the header as I could, et Viola! suddenly CPU temperatures are down to the low 80s with all cores flat out. The radiator is also getting noticeably warmer, suggesting a greater heat transfer and the CPU fans are no longer screaming like a banshee.

It seems odd that the AM4 socket header would sit proud of the CPU but that appears to be what is happening. I'm letting it run for a bit to gather more data.

I've had a similar thing happen on my Noctua cooler. Somehow it is possible to install it so that it does not sit perfectly on the heatspreader. It will work but far from optimal performance. I noticed that I have to test that the cooler won't move one millimetre when the bolts are tightened to maximum, then it is correctly on top of the CPU.

Thanks, puuteknikko. I note also that moving the cooler away from the header will also have moved it closed to the CPU hot spot centre, as discussed in der8auer's video. I will probably get one of his AIO mounting brackets, which should completely fix the problem, when the product is available in the UK, since the shipping cost from Germany is nearly as expensive as the kit!

G3WGV wrote:The video suggested by TxRedneck gave me an idea that perhaps the cooler was snagging on the socket header and therefore not making good flat contact with the CPU. The stock AM4 cooler mounting arrangement doesn't give much flexibility but I pushed the cooler as far away from the header as I could, et Viola! suddenly CPU temperatures are down to the low 80s with all cores flat out. The radiator is also getting noticeably warmer, suggesting a greater heat transfer and the CPU fans are no longer screaming like a banshee.

It seems odd that the AM4 socket header would sit proud of the CPU but that appears to be what is happening. I'm letting it run for a bit to gather more data.

Interesting, indeed! I was lucky with my Ryzen 3700, then. It typically runs in the mid 70s when loaded, with occasional peaks in the low 80s. I wonder if different motherboard mfgrs have slightly different socket configs...

Nert wrote:

WussyDan wrote: I have a Noctua U12S 120mm single fan tower on the CPU

Can you add a second fan to that cooler and run it as push/pull ? I took a quick look at Amazon and it looked like it might be possible ?

Yes, the U12S can have a second fan installed on the backside of the fin stack, which I didn't know until recently. I may try that as well.

WussyDan wrote:Yes, the U12S can have a second fan installed on the backside of the fin stack, which I didn't know until recently. I may try that as well.

It's an inexpensive thing to try. I would expect it to improve things a lot .... puts more air flow right where it's needed. If that or air flow through the case isn't the problem, there's always more exotic possibilities. I've read about how Intel changed from a soldered connection to one that uses a thermal compound to wick heat from the CPU die to the transfer plate. This compound deteriorates over time and temps increase. I folded with a 4770 chip (I think that's the model number) that showed these symptoms as it aged. I had a water cooler on it, so it was still usable for folding, but the upward trend was noticeable. The 4790K used a "new and improved" compound. Don't know if that was any better than the 4770 compound or not. Here's a link that talks a little bit about it. Looks like that chip's been around for a long time. https://www.pugetsystems.com/labs/artic ... e-TIM-573/

Keep us posted if the extra fan helps. I'm guessing it will.

Nert wrote:

WussyDan wrote:Yes, the U12S can have a second fan installed on the backside of the fin stack, which I didn't know until recently. I may try that as well.

It's an inexpensive thing to try. I would expect it to improve things a lot .... puts more air flow right where it's needed. If that or air flow through the case isn't the problem, there's always more exotic possibilities. I've read about how Intel changed from a soldered connection to one that uses a thermal compound to wick heat from the CPU die to the transfer plate. This compound deteriorates over time and temps increase. I folded with a 4770 chip (I think that's the model number) that showed these symptoms as it aged. I had a water cooler on it, so it was still usable for folding, but the upward trend was noticeable. The 4790K used a "new and improved" compound. Don't know if that was any better than the 4770 compound or not. Here's a link that talks a little bit about it. Looks like that chip's been around for a long time. https://www.pugetsystems.com/labs/artic ... e-TIM-573/

Keep us posted if the extra fan helps. I'm guessing it will.

It was between the 2600k and the 3770k series that Intel went from a soldered ihs to using tim. It saved them lots of money to be sure. That said, it also gave rise to the delliding services, tools and further enthusiast markets.

At the end of the day, knowledge is King and possession of it will help us advance our cause.

Tx

Nert wrote:

WussyDan wrote:I've read about how Intel changed from a soldered connection to one that uses a thermal compound to wick heat from the CPU die to the transfer plate. This compound deteriorates over time and temps increase. I folded with a 4770 chip (I think that's the model number) that showed these symptoms as it aged. I had a water cooler on it, so it was still usable for folding, but the upward trend was noticeable.

My 4770 now gets up to 90°C when Folding (clean Noctua NH-D14 cooler), so I don't do that any more.

TxRedneck wrote:My 4770 now gets up to 90°C when Folding (clean Noctua NH-D14 cooler), so I don't do that any more.

I did some research a while back about commercial services that will delid these things and re-apply better thermal compound. Also a bunch of diy videos where people pulled them apart with razor blades and / or vices. Don't know if any of these old links that I saved are still valid or not.

https://www.pcgamer.com/meet-the-overcl ... iot-proof/

https://siliconlottery.com/products/del ... 3692666966

I never got around to pulling the trigger on trying this since I think maybe the motherboard died and I ended up swapping it out and replacing the processor to work with the newer board. It was a while back I can't remember the details.

As always individual Due Diligence is always required for anyone that stumbles on this thread.

So, for anyone who is curious, the culprit was unfortunately my poor U12S simply not being up to the task. My CPU is indeed OC'd (the lazy, sloppy way via BIOS) due to frequently playing poorly optimized CPU heavy games, which obviously isn't helping. I don't particularly feel like turning the OC off and on based on whether I'm folding or gaming, so I took a gamble and found a decent deal on the aforementioned U12A. Between that and adding a second fan to the U12S, I went with the U12A due to the wider fin stack, extra heat pipe and newer fan design. This setup has worked wonderfully after installation and a good blowout, and I'm currently folding on Full with zero CPU restrictions and sitting at a comfortable ~75C CPU average temp., even with the dirty OC enabled, and GPU staying at ~65C. I appreciate all the help and suggestions, and hopefully others find them useful as well.

For anyone who happens to be curious, here's my case (cable mgmt could be better, no doubt) with the U12A installed. It's a pretty solid medium duty cooler, and fits nicely.
https://imgur.com/gallery/HdJqH9A

reducing the cpu threads used helped me with this problem

I can't use 11/12 threads in this application