Return to Level1Techs.com

Need oppinions - Threadripper 2950x concerning number reporting in windows (boost to 4.4 GHz @1.5 vcore auto with 68 degrees C)


#1

I’m new to ryzen (my last AMD machine was an Opteron more then 10 years ago) but I was really excited to be able to build a new workstation for Houdini sims with the Threadripper, so that’s exactly what I did.

System specs:

Threadripper 2950x
Asus Prime x399-A
Corsair Vengeance LPX 3000 MHz CL15 32 GB (I know it’s a bit low when considering the workload)
Enermax TR4 280 AIO
Corsair HX1000w PSU
Fractal Design R6 case

So far so good, the build was easy to put together no difficulties there. So next logical step was to to test out some basic overclocking.

Step 1. Enable the DOCP for the memory … didn’t work out of the box 'till I took down the BCLK to 100 MHz and set the memory down to 2933

Step 2. Enable performance enhancer level 3

Step 3. Set Vcore toe offset mode to a negative of -0.05v

The result was a Cinebench final score of 178 single-core and around 3475 multi-threaded score

Now the part that is worrying me:
On the single-core tests the cpu was boosting up to 4.4 GHz with a reported voltage of 1.5v in HWmonitor with an average temp of 44 degrees C and for the Multi-threaded test all cores were boosting up to 4GHz @1.35v with a temp of 64 degrees C.
When running SuperPi stress test with small FFT the temps were hitting 68 degrees C and from what I’ve read online that’s 'kinda the highest temp admissible.

Is anyone else seeing this kind of numbers?
Do you guys think it’s safe to keep these type of OC (voltage and temp) for extended periods of time ?


#2

The 68C temp is just when the boost/XFR stops and the processor will downclock to base freq. The chip can go to 95C like most other silicon ( though I wouldn’t ever get my chip that hot ). If you are manually overclocking then it will bypass SenseMI’s Power and Thermal monitoring, and won’t downclock at 68C, but it can run hotter.

WIth a Enermax TR4 360 I sit at about 55-57C under full load ( 21C ambient temps ) that’s default settings, no OC.


#3

What worries me more then the temps (used to run my intel chips @ 80 C full load) is the voltage of 1.5v that HWmonitor displays when the 4.4GHz boost kicks in. In fact I’m not yet sure if maybe the reporting software is even correct or not.

Maybe enforce an even bigger undervolt offset like -0.1v or even -0.15v though I’m not sure how the boost system will handle that?

As for manually OC-ing I don’t really want to keep the cores pined all the time at max frequency just for day to day tasks.

Also I was curious if anyone has any experience with the LLC settings on the ASUS Prime x399-A board. Can’t find much info about the offset curve online. Or actually any ASUS LLC since i think they share the same bios settings.


#4

The AMD Chips run with a bit more voltage than the Intel ones. AFAIK 1.5V for a single core under 4.4GHz turbo is fine. It’s just the one core boosting up for a second. It shouldn’t normally stay at 1.5V the whole time, it usually drops down to 1.3v or so, depending on the load, etc.

1.5V on all cores would be concerning.

I wouldn’t recommend HWMon it can be inaccurate, if you are running Windows try the Ryzen Master software it should give more accurate voltage readings.

If you don’t want to keep all cores pinned at max freq, then I would say just let SenseMi do it’s thing and don’t overclock at all.

It will automatically adjust the clocks based on power and temperature to give you the best performance, with the added benefit of SenseMi’s safety net.

Obviously it won’t be as fast as a manual overclock, but for me in day to day tasks I don’t notice a difference.


#5

My opinion is I wouldn’t want more than 1.45 going to it. If it’s just kissing 1.5 every once in a while for a very short time then a small undervolt is fine so long as it’s stable. You can safely undervolt all you want, the only downside is it becomes less stable. Push it until it locks up and then bump the offset back up a bit. .1v negative offset shouldn’t be an issue.


#6

Only had about an hour to test stuff out last night with a -0.1v offset. It was only boosting (on a single core) up to 4.32 GHz but the surprising part was that peak voltage was still hitting 1.5v sometimes.
Did a burn-in test with Prime95 small FFT of 8 hours and it was perfectly stable.

Didn’t have much time to dig around the bios but I’m hoping to find a way just to set a max voltage threshold somehow and just set it to a flat 1.45.v max value while still maintaining the ability to down clock when nothing is going on.


#7

You can always lock the voltage and maintain speedstep or whatever AMD calls it. Are you looking at vcore or vid? Those are not always the same thing.


#8

I was checking the VID since i thought that high levels are dangerous on either one.
Anyway after getting home I set the voltage on the Vcore to a flat 1.4 and this is what i got:

As you can see in the screenshot the vcore is 1.43v (probably due to the LLC) and the max VID I can see got up to 1.525v.
Let me know what you guys think.

EDIT: Oh and i forgot to mention that the asus Mb software is reporting lower vcore so HWMonitor is say 1.417v idle and the asus util is 1.384v.