Folding Forum

That first voltage number is not the Vcore number, I got that from dmidecode. I decided to reboot and enter the CMOS, the Vcore number is 1.464. The 1.1 is "CPU NB/SOC". I also enabled Game Boost to see what happens.

Turns out as soon as the computer loaded in it got a F@H job that it is running now at 4200mhz (max 4400mhz), after about ten minutes the CPU is at 82c. So 82c appears stable and is 12 to 13c below threshold.

Over one hour of running at 100% the temp is now at 87c. That's a great number on active air cooling but I'm going to reach out to Noctua to see if they offer a smaller fan I can use for the second fan on the cooler since the included fan is slightly too large to fit over the DRAM or VRM heatsink. Also it should be noted that AMD's thermal limit rating is 95c for both the 3700X and the 3900X.

I'd also look into this. https://www.youtube.com/watch?v=JlJDv68fEcM

I just sent Noctua a email enquiring if they offer a slightly smaller fan to be used as the secondary on the NH-D15 SE-AM4.

Roadpower wrote:I just sent Noctua a email enquiring if they offer a slightly smaller fan to be used as the secondary on the NH-D15 SE-AM4.

The original fans are available https://smile.amazon.com/Noctua-NF-A15- ... =0&ie=UTF8
but I'm sure you could get away with a standard 120mm fan. I've had nothing but great luck out of Cougar fans https://smile.amazon.com/Cougar-Vortex- ... 218&sr=1-5

Tx

Jon_Tesch wrote:

Juggy wrote:Also, an AIO water cooler would probably do wonders if you're running your CPU full tilt for long periods of time

I have been expecting my air cooler to perform relatively well given how large it is -- even on par with a smaller AIO, but obviously this is not the case. I am suspecting that the issue is either a problem with the power draw and voltages of my CPU or my cooler might be defective/installed incorrectly -- I doubt the latter but anything is possible. I will try to reapply thermal compound today to see if that helps. What do you mean by running my CPU full tilt?

Thanks,
Jon T.

Full tilt as in full speed 100%

TxRedneck wrote:

Roadpower wrote:I just sent Noctua a email enquiring if they offer a slightly smaller fan to be used as the secondary on the NH-D15 SE-AM4.

The original fans are available https://smile.amazon.com/Noctua-NF-A15- ... =0&ie=UTF8
but I'm sure you could get away with a standard 120mm fan. I've had nothing but great luck out of Cougar fans https://smile.amazon.com/Cougar-Vortex- ... 218&sr=1-5

Tx

This is a brand new Noctua cooler identical to what the OP has. It came with two fans, the issue is that the secondary mounting locations are very slightly restricted size wise (blocked by DRAM on one side and a VRM heatsink on the other) so I need a smaller fan to mount. I'm hoping that Noctua offers something that can readily mount on the heatsink using the same hardware or otherwise I will have to cobble something into place. In fact I happen to have one of those Courgar fans. They're quite good.

Right on, I'd just try that extra fan you have on hand, nothing to lose while you're waiting to hear back from Noctua.

Jon_Tesch wrote:Whenever I run F@H, no matter the power selection, my CPU hits ridiculously high temperatures -- 87C on the first run and 97C on the second. I stopped folding immediately when I saw these temps through CPUID HWMonitor. My GPU temps (max at around 61C) are slightly elevated but not out of reason to my knowledge.

I don't see the most important item, your OS. The Ryzen 3000 series shows temps that are 20 C too high on Linux. They have not supplied the correction factor yet. I am on Ubuntu 18.04.4 with Linux kernel 5.3.0.36, and PSensor shows Tdie to be 85 C, with an Arctic water cooler, so it is really 65 C. They have a fix, and will apply it to a subsequent kernel update.

I don't know about Windows.

Jon_Tesch wrote:...Why does specifically F@H make my CPU into a mini Chernobyl?...

Most CPU Stress tests may not use AVX instruction set as heavily as F@H. When it comes to CPU, F@H uses the code from GROMACS (https://en.wikipedia.org/wiki/GROMACS) which is extremely optimized... some parts of the code are hand-optimized in the compiler itself. Plus, the optimization process is always happening so future versions of the code would have more optimizations and use newer features to ensure that the scientific calculations are speeding up without loss of accuracy.

I have seen Linpack Xtreme (https://www.techpowerup.com/download/linpack-xtreme/) approach similar temperatures as F@H.

I have a Define S which is a longer Define C and the front cover makes a huge difference to my cooling but of course your mileage may vary

My experience with Ryzen is that the AMD UEFI BIOS sets too high a voltage. So that it can boost, the boot-up voltage is set at numbers that are only needed for that boost. I've found by setting the cpu multiplier to its stock number (3600 so set at 36 multiplier) the core voltage is only 1.1v instead of 1.4v. That drops temps by over 20C allowing the cpu cooler to be quieter but the system is still super quick.

Are you using the stock cooler?
The stock cooler on mine was pure garbage.
I had to buff the bottom to make it smooth! The bottom of the cooler had 0.5 to 1mm uneven ridges, which collected all the thermal paste.

There are 2 things you can do,
First, go into your Bios, and disable Overdrive technology. It bumps down the current under load wattage from ~135-155W to it's stock 105W value.
That should account for a good 10-15C drop in temperature.
If that doesn't do it, enable ECO mode. Your CPU will drop from stock 105W to 65W.
If you enable overdrive technology with ECO mode, your CPU will run within 80% from stock speed, at about 90-95W (that's a 10 watt drop from 105W, or an additional ~5-10C drop in CPU temperature).

Experiment with these settings, to your best value.
Then go online and buy a closed watercooling system. A decent one goes from $75-$125.
In many cases, CPU heatspreader doesn't get enough cooling, and case fans can't successfully draw out all the heat.
For a 3900x running at the 105W setting, you'll need a dual 100 or 120mm fan radiator (they usually sell with '240' in the item name).
Running a single 120mm watercooling system (going for ~$50) will overheat your CPU at 105W, but works at 65W with overdrive technology.
I run mine at 78C at 65W with O-T, but it's not recommended to use such small OC unit, unless the radiator is quite thick.


PS: If you go water cooling, you'll need a 80-120mm case fan, to cool the VRMs if you're running the CPU at it's stock wattage setting (105W + Overdrive).

MeeLee wrote:Are you using the stock cooler?
The stock cooler on mine was pure garbage.
I had to buff the bottom to make it smooth! The bottom of the cooler had 0.5 to 1mm uneven ridges, which collected all the thermal paste.

There are 2 things you can do,
First, go into your Bios, and disable Overdrive technology. It bumps down the current under load wattage from ~135-155W to it's stock 105W value.
That should account for a good 10-15C drop in temperature.
If that doesn't do it, enable ECO mode. Your CPU will drop from stock 105W to 65W.
If you enable overdrive technology with ECO mode, your CPU will run within 80% from stock speed, at about 90-95W (that's a 10 watt drop from 105W, or an additional ~5-10C drop in CPU temperature).

Experiment with these settings, to your best value.
Then go online and buy a closed watercooling system. A decent one goes from $75-$125.
In many cases, CPU heatspreader doesn't get enough cooling, and case fans can't successfully draw out all the heat.
For a 3900x running at the 105W setting, you'll need a dual 100 or 120mm fan radiator (they usually sell with '240' in the item name).
Running a single 120mm watercooling system (going for ~$50) will overheat your CPU at 105W, but works at 65W with overdrive technology.
I run mine at 78C at 65W with O-T, but it's not recommended to use such small OC unit, unless the radiator is quite thick.


PS: If you go water cooling, you'll need a 80-120mm case fan, to cool the VRMs if you're running the CPU at it's stock wattage setting (105W + Overdrive).

All of these suggestions effectively defeat the purpose and benefit of the SoC. It is a great product, but a less than optimal design, hence the creation of the cooler relocation doodad I linked above.

I only harp on it because it's legit and I want to help mitigate issues yall are having. I've no dog in this hunt beyond getting more research and results turned in.

I'm using Linux Mint 19.3 Tricia 5.6.3 and "sensors" command to watch temps. I am curious about your temps because Ryzen will start to throttle at 95C. Double check temps with HWinfo64 or Ryzen Master.

My 3950X has a NH-D15 with AM4 mounting kit. I used Ryzen Master with AutoOverclock +200 PBO to configure the BIOS for me. I flog all of the cores and my temps were below 80C in a 27C room. I had to remove one of the Noctua fans to fit an AIO cooler for my 1080TI and now my Tdie temps are ~83C.

My 3970X has a NH-U14S TR4-SP3 with 63/64 cores being used. Same overclock as above. Tdie temps are ~83C or less. Both have Fractal cases (good stuff).

You shouldn't have to undervolt or anything like that. I would recommend remounting your cooler. I prefer the pea dot paste method for the Ryzen 9. The TR gets a crazy array of pea dots. I use Arctic Silver 7. I also recommend ensuring that your CPU fan profiles are set correctly in your BIOS. Mine are set to 100% above 78C.

If all else fails for some reason, I would just AMD's Eco Mode. It works extremely well and you are left with plenty of horsepower.

I played around a bit with my 3800x today. Set voltage to 1.25 and turned on all core manual clock to 4.2 Ghz. That's running about 150Mhz all core over stock and 5C cooler (78C now).
So far it's perfectly stable.
I tried everything I could using auto and pbo, but my temps would get closer to 90C than I liked.