Is there an EPYC Trubo Boost Clock Breakdown/Safe Max Temps Anywhere?

IamSpartacus

Well-Known Member
Mar 14, 2016
2,298
553
113
I'm looking for some kind of breakdown of how AMD's boost clock works similar to how you can view that here for Intel CPU's.

So for example the E5-2680v3 has a turbo listing of 4/4/4/4/4/4/4/4/5/6/8/8 which means it will boost 400MHz once 5 cores are in use, 500MHz with 4 cores, 600MHz with 3 cores, and 800MHz with 1-2 cores. I can't find anything similar to this for AMD EPYC CPU turbo boost.

Also, finding it hard to locate any info on what the maximum safe operating temperatures are for these chips.
 
Last edited:
  • Like
Reactions: llowrey

IamSpartacus

Well-Known Member
Mar 14, 2016
2,298
553
113
In testing on my 7320p just now, I see up maxiing out 6c/12 threads it boosts to maximum (3.3GHz) while maxing to 8c/16t it goes to 3.25GHz. Pretty solid IMO.
 
  • Like
Reactions: TXAG26 and llowrey

TXAG26

Active Member
Aug 2, 2016
274
81
28
What power profile are you using and any power consumption numbers you can share? I'm still getting familiar with my system, and after burn-in with MemTest86, I'm probably going to back the DDR4 back to 2933MHz (from 3200MHz) for the better latency and slightly lower power consumption.
 

Jeff Robertson

Active Member
Oct 18, 2016
421
113
43
Chico, CA
My 7402p seems to hover real close to its max clock when all 48 threads are loaded using a noctua cooler. My understanding is that if the thermal headroom exists they can more or less get to their max speed on all cores. I have run a few benchmarks where the speed dropped by a couple hundred Mhz under full load but those were power virus tests.
 

IamSpartacus

Well-Known Member
Mar 14, 2016
2,298
553
113
My 7402p seems to hover real close to its max clock when all 48 threads are loaded using a noctua cooler. My understanding is that if the thermal headroom exists they can more or less get to their max speed on all cores. I have run a few benchmarks where the speed dropped by a couple hundred Mhz under full load but those were power virus tests.
The Noctua coolers dont blow front to back though right?
 

Jeff Robertson

Active Member
Oct 18, 2016
421
113
43
Chico, CA
I'm generally not a fan boy but I use Noctua products almost exclusively. I'm pretty sure the engineer on duty the day they designed the 90mm SP3 cooler had a crook in his neck and screwed up the math. I can't for the life of me figure out why they thought the side to side airflow design was a good idea. That being said it keeps my 7402p very cool even under 100% load and is dead silent.
 

TXAG26

Active Member
Aug 2, 2016
274
81
28
I'm generally not a fan boy but I use Noctua products almost exclusively. I'm pretty sure the engineer on duty the day they designed the 90mm SP3 cooler had a crook in his neck and screwed up the math. I can't for the life of me figure out why they thought the side to side airflow design was a good idea. That being said it keeps my 7402p very cool even under 100% load and is dead silent.
What max CPU temps are you seeing with the Noctua when the CPU is maxed out with something like Prime95 or other CPU-heavy benchmark?
 

IamSpartacus

Well-Known Member
Mar 14, 2016
2,298
553
113
I'm generally not a fan boy but I use Noctua products almost exclusively. I'm pretty sure the engineer on duty the day they designed the 90mm SP3 cooler had a crook in his neck and screwed up the math. I can't for the life of me figure out why they thought the side to side airflow design was a good idea. That being said it keeps my 7402p very cool even under 100% load and is dead silent.
The reason is it was designed for Threadripper which socket is aligned differently.
 

Jeff Robertson

Active Member
Oct 18, 2016
421
113
43
Chico, CA
What max CPU temps are you seeing with the Noctua when the CPU is maxed out with something like Prime95 or other CPU-heavy benchmark?
So with prime95 running 48 threads the temps stabilized at 59 degrees and the noctua cooler was running at 1000RPMs. CPU clocks were 3.30Ghz across all 24 cores with occasional dips to 3.28Ghz.
 

Jeff Robertson

Active Member
Oct 18, 2016
421
113
43
Chico, CA
The reason is it was designed for Threadripper which socket is aligned differently.
Good to know there is a logical reason! If AMD CPUs keep getting more popular maybe they will start giving the platform more attention and design coolers for both orientations. In my case I have a fanless PSU which the cooler blows directly at so in some ways it's a slight benefit in my specific situation.
 
  • Like
Reactions: TXAG26

TXAG26

Active Member
Aug 2, 2016
274
81
28
I’m a big Noctua heatsink fan as well but purchased the Supermicro U4 active heat pipe SP3 Epyc heatsink since it has a front to back orientation. The Supermicro SP3 comes with a 92mm fan that runs at 1,300 rpm and it’s very quiet as well. Hopefully Noctua releases a front-back airflow heatsink soon as I wouldn’t mind upgrading to one of those, although the Supermicro one is working just fine as well.
 

IamSpartacus

Well-Known Member
Mar 14, 2016
2,298
553
113
So with prime95 running 48 threads the temps stabilized at 59 degrees and the noctua cooler was running at 1000RPMs. CPU clocks were 3.30Ghz across all 24 cores with occasional dips to 3.28Ghz.
Interesting, when I stress all of my cores my CPU stays at base clock (3.0GhHz). Wonder if it's the additional TDP the 7402p has? Then again I'm testing from a Windows VM on Linux. I'll have to boot my server to Windows bare metal and see what I get.

@TXAG26 Have you tested your 7302p?
 
Last edited:

TXAG26

Active Member
Aug 2, 2016
274
81
28
Interesting, when I stress all of my cores my CPU stays at base clock (3.0GhHz). Wonder if it's the additional TDP the 7402p has? Then again I'm testing from a Windows VM on Linux. I'll have to boot my server to Windows bare metal and see what I get.

@TXAG26 Have you tested your 7302p?
@IamSpartacus Yes, I'm quite happy with it. I posted a short write-up with more info here:
https://forums.servethehome.com/ind...-nc-workstation-build-follow-up-report.27263/
 

IamSpartacus

Well-Known Member
Mar 14, 2016
2,298
553
113
Last edited:

TXAG26

Active Member
Aug 2, 2016
274
81
28
Hmmm, I'm going to have to look more into why my CPU is downscaling on full load. I checked with the scaling governor set to both On Demand and Performance. Even with Performance and all cores set to 3.3GHz, once I load up all cores the frequency drops to 3.0GHz with Prime95.
I’ll try to post specific settings this evening, but I left just about everything performance related in the bios on “auto”, and I did adjust the ctdp and some other related item to manual and I typed in “180” to correspond with the 7302p max TDP. In Windows 10 Pro, I did a custom high performance power plan, and have 0% in box for minimum CPU and 100% for max CPU and have the “active” setting enabled for fan/heatsink.
 

IamSpartacus

Well-Known Member
Mar 14, 2016
2,298
553
113
I’ll try to post specific settings this evening, but I left just about everything performance related in the bios on “auto”, and I did adjust the ctdp and some other related item to manual and I typed in “180” to correspond with the 7302p max TDP. In Windows 10 Pro, I did a custom high performance power plan, and have 0% in box for minimum CPU and 100% for max CPU and have the “active” setting enabled for fan/heatsink.
Ah, if you set cTDP to 180 that might be it. I think i left my on default which is 155. I'll have to check my BIOS settings again on next reboot.
 
  • Like
Reactions: TXAG26

IamSpartacus

Well-Known Member
Mar 14, 2016
2,298
553
113
Well setting cTDP to 180 made no difference. CPU shows running at 3.3GHz until I load it up then it drops down to 3.0+GHz.
 

TXAG26

Active Member
Aug 2, 2016
274
81
28
Well setting cTDP to 180 made no difference. CPU shows running at 3.3GHz until I load it up then it drops down to 3.0+GHz.
Which options are you running for Prime95? Mine runs as follows with 32 threads:
7302P runs at 3.18 GHz on "Smallest FFTs for high power/heat/cpu stress"
7302P runs at 3.10 GHz on "Small FFTs for max power/heat/cpu stress"
7302P runs at 3.27 GHz on "Large FFTs for stress on memory controller & ram" (brings system to its knees)
7302P runs at 3.27 GHz on "Blend All of the above" (brings the system to its knees)
 

IamSpartacus

Well-Known Member
Mar 14, 2016
2,298
553
113
Which options are you running for Prime95? Mine runs as follows with 32 threads:
7302P runs at 3.18 GHz on "Smallest FFTs for high power/heat/cpu stress"
7302P runs at 3.10 GHz on "Small FFTs for max power/heat/cpu stress"
7302P runs at 3.27 GHz on "Large FFTs for stress on memory controller & ram" (brings system to its knees)
7302P runs at 3.27 GHz on "Blend All of the above" (brings the system to its knees)
Ive only tested the default Blend test.
 

sno.cn

Active Member
Sep 23, 2016
206
73
28
36
Also don't forget if you set cTPD, set package power limit too.

For some reason changing these on 7282 didn't do anything. Works great on 7452 though, changing from default 155w to 180w pulls almost exactly 50w more from the wall on 100% load.

I set to max on my 7452 with determinism on performance. All cores are running 2823 MHz on full load. Running different number of threads predictably lets those cores boost higher.

But it's definitely a power limit, and not temp related because both CPUs are running ice cold on SM 4U heatsinks.

*Edit: One question I have is, does changing other settings that contribute to total package power, change boost behavior by allowing more cores to run at higher frequency? AMD's tuning guide has a handful of references like this: "Lowering the memory clock speed also results in power savings in the memory controller, thus allowing the rest of the SoC to consume more power potentially resulting in a performance boost elsewhere."
 
Last edited:
  • Like
Reactions: TXAG26