I have similar setup and same issue, x399-a, bios 601, just memory 32G, 3500mhz with DOCP, I use 1.34V, 1.11 VSoc (I notice asus is very generous on VSoc when its set to auto and increases temp) . Cooler is Noctua. I dont remember which one but I disabled some cpu dynamic power options on bios too, I use cpupower in performance mode. My idle temps around 45 full load temp max 80 and it is survived half day run. But when I set powersave governor all cores 2.199 but temps stays same. I assume it has to do with hardcoded 1.34V or memory on bios?
Or the paste just got a bit weird. While you are at it, maybe check the fans and the radiator and the pump just in case something somewhere else is causing those temps.
I actually just got done applying new thermal paste, and ran a few more renders. I used a completely different thermal paste I recently purchased, so this is the first time it’s been used today. I’m getting similar results with temps, however oddly enough, the first render stayed between 3.10GHz - 3.50GHz despite the load temp actually going 3c higher than before (now 68c). The second render dropped to ~2.60GHz (slightly higher than the original 2.30GHz). The odd thing is that even though the performance is somewhat better, the temps aren’t that much better (other than a ~4 degree drop in idle temps). And this may be why…
I’m now using AMD Ryzen Master to monitor the CPU, I noticed that my CPU voltage is fluctuating and not staying at a constant 1.3v that I manually set it too in the BIOS. I thought my settings may have reset when I unplugged the PSU to change my thermal paste. However when I went into the BIOS, all settings remained the same, and the CPU voltage at 1.3.
TL;DR. Idle temps slightly lower, load temps slightly higher. Speed is somewhat more consistent, but still drops. 1.3v still steady (with Ryzen Master installed) until system restart after thermal paste swap - might be something to do with Ryzen Master.
I wish I could have the voltage default instead of the manual 1.3, but that will cause my PC to not stay powered on, causing it to shut on/off (my theory - it’s the mobo or PSU). However that’s a problem for another time when this clock speed reduction is solved.
Despite the higher than average temps in relation to Gamers Nexus, I don’t believe it to be the cooler (100% anyway). I checked everything and it’s fine, even under load from what I’ve been reading on this thread (and others), I feel like the current temps shouldn’t be effecting my speeds that badly. It can’t be the watt usage of the CPU, it only reached ~160 at max load. (Apparently 180+ watt usage is one of the reasons it could be under-clocked…though I’m not sure how legible that claim is, though on the contrary it sounds logical).
@wendell you replied earlier this morning. Any insight after all that has transpired up to now?
1 Like
Check your vrm temps. The fact it performed at least somewhat normally for a bit is telling. I would up the voltage to the AMD reported safe max of 1.45 or maybe 1.4 just to see what effect it has.
1 Like
I had a similiar issue while I was getting my new X399 stable, when I got it.
The thing was i noticed throttling at around ~67°C (tDie) when using fixed voltage or auto voltage + auto clocks, while running Prime95 or other constant things that would push my CPU @ 100% for longer.
As soon as I set a manual CPU clock, through multiplier, that limit vanished.
I’m now running my 1900X @ 4GHz and ~1,27V. With all that at around ~80°C when running full load long enough and everything with a Noctua cooler and 2 Be Quiet Silent Wings 3 120mm @ ~1000Rpm.
While I would get a little bit higher clock on 4 cores with XFR, which is disabled now due to the OC, it doesn’t affect anything.
1 Like
I just did two renders, and coincidentally enough the vrm temperatures hover around 68c. If I turn up the CPU voltage to 1.4 or higher, I get the problem with the PC turning on/off, as if there’s not enough power to keep it running. I’m pulling a project from a large repo, so I will have to try it again when that’s done, but I don’t have hopes it’ll work.
Hmmm, maybe I might try that later.
Yep i agree on this one.
1.3V is probablly too low, unless he goes really agressive on the LLC levels.
But then its probablly still not enough.
1.40V / 1.42V is probablly the spot you wanne be arround.
Wenn it comes to LLC levels, this really depends on how the said board works with that.
That would be just a matter of testing out and fiddling arround with.
But even without toutching any LLC levels, it still should be able to stay stable.
1 Like
@wendell @MisteryAngel I was able to change the CPU voltage to 1.4. and it made the performance even worse. Immediately after clicking render it drops to 2.20GHz. I’m going to try and set it to auto and see if that works, but it might crash as I’ve said before: …
When you applied the new thermal goop, did you also reseat the CPU? Maybe there is a contact problem on the socket and something doesn’t get power because of that? … Total shot in the dark of course.
@wendell @MisteryAngel @noenken Hmmmmm…well son of a b*tch, setting the CPU voltage to auto fixed the problem! I rendered for quite some time and it never dropped below 3.50GHz 
. The load temperatures were actually a bit lower than the 68c I was getting before.
I know it’s silly why I didn’t do this before, but as I said multiple times, when I set the CPU volts to auto, it would over power and crash my system, but it’s running stable now. Well, whatever the reason why it’s stable now I will never know 
Not sure to call this done just yet, but it’s been running flawlessly now, and steady at 65c. Here’s a video of me rendering in real time with all the stats to the left. Let me know what you think about the numbers, and if everything checks out. https://www.youtube.com/watch?v=y5tfgURPJM8, Looks fine to me though.
Thank you everyone in this thread for helping me out, and giving me clues, you’re all awesome! If I could, I’d buy you all beers…Oh, wait a sec… https://beeroverip.org/. I would keep this thread open though, in the event my good luck runs out. 
6 Likes