[SOLVED] How long does a 2700x is supposed to sustain its boost clock on a single thread task?

[O-steal]

Hello

I have a 2700x and I was wondering how long does such a CPU is supposed to sustain the boost clock on single thread tasks?

Myself, I only see sporadic spikes of less than 100ms that goes only 4,2GHz (officially 4.3)

Also, when I look at the clocks, the thread being utilised does not show any sign that he is more used than the others (all cores kinda jump up and down in frequencies, nothing is really sustained). Should one clock be significantly higher than the others?

Thanks

 

[drea.drechsler]

Hello

I have a 2700x and I was wondering how long does such a CPU is supposed to sustain the boost clock on single thread tasks?

Myself, I only see sporadic spikes of less than 100ms that goes only 4,2GHz (officially 4.3)

Also, when I look at the clocks, the thread being utilised does not show any sign that he is more used than the others (all cores kinda jump up and down in frequencies, nothing is really sustained). Should one clock be significantly higher than the others?

Thanks

It holds a boost only as long as the algorithm continues to 'see' thermal and power headroom. Usually though it's temperature that causes it to lower boosts. The highest boosts are usually pretty short, at most a few hundred mS. But you can, with careful tweaking of PBO settings, increase that quite a bit with a 2700X.

And yes, it's normal to be boosting on one core only to the highest rated boost clocks. And since it's only boosting in "light, bursty" workloads it also makes sense the core isn't really being utilized heavily.

Lastly: the scheduler should prefer certain cores over the others since those cores were designated as 'strongest'. So they should boost more often and higher than others...with some (depending on silicon lottery) possibly never boosting really high at all.

The problem is, the Windows scheduler kind of sucks in that it doesn't really recognize the strongest cores and also doesn't even recognize a full thread vs. an SMT "virtual" thread, so it schedules threads pretty badly. AMD is trying to fix it with Ryzen 3000 but it's less an issue on Ryzen 2000 because of the way they work.

 

[O-steal]

Thank you.

It holds a boost only as long as the algorithm continues to 'see' thermal and power headroom. Usually though it's temperature that causes it to lower boosts.

I am a bit surprised though, I was expecting something like a core being able to sustain 4.2GHz for a few seconds at least knowing my temps are below 70C and I had 300+ watts headroom in my tests.

Also I saw this graph from amd. It indicates performances far beyond a single burst but more like a couple of fast threads (4 Threads at 4.2GHz looks like to be expected. I do only get a single burst at that frequency (all other cores are at or below 4))

The problem is, the Windows scheduler kind of sucks in that it doesn't really recognize the strongest cores and also doesn't even recognize a full thread vs. an SMT "virtual" thread, so it schedules threads pretty badly

Indeed, the cores that see the peaks are not the one flagged with a star in Ryzen master.

 

[drea.drechsler]

...
I am a bit surprised though, I was expecting something like a core being able to sustain 4.2GHz for a few seconds at least knowing my temps are below 70C and I had 300+ watts headroom in my tests.
...

It's not just the 'watts' headroom as something called TDC and EDC along with PPT.

Package Power Tracking (“PPT”): The PPT threshold is the allowed socket power consumption permitted across the voltage rails supplying the socket. Applications with high thread counts, and/or “heavy” threads, can encounter PPT limits that can be alleviated with a raised PPT limit.

Thermal Design Current (“TDC”): The maximum current (amps) that can be delivered by a specific motherboard’s voltage regulator configuration in thermally-constrained scenarios.

Electrical Design Current (“EDC”): The maximum current (amps) that can be delivered by a specific motherboard’s voltage regulator configuration in a peak (“spike”) condition for a short period of time.

Each of these are defined in the BIOS and vary depending on the processor and motherboard VRM capability. If it starts to see any of these coming close to the limits (or certain temperature thresholds) it reduces boosting. PPT, for instance, at default will be below 300 W for any Ryzen processor. Probably about 142 watts for a 2700X (not really sure) if the motherboard has a VRM capable of delivering it.

BUT, you can effectively disable the limits by enabling PBO in BIOS and raising them to the max. Many people have managed to obtain a lot longer and higher boosting on 2700X CPU's that way.

 

[O-steal]

Thanks for the knowledge.

I am still a bit "not really satisfied" with the fact that the slideshow show many cores at 4.2 not in the context of an overclock competition, when you on the other hand, explain that boost clock can be achieved on single core providing you have a good high end motherboard and have tweaked by hand power delivery settings. If you are right, that would mean the advertised specs are more than misleading

But I think this is enough info for me, I understand a bit better how all of this works.

All 2700x possessors, feel free to add personal logs of your clocks, because today I have absolutely nothing to compare mine too and when I read reddit, I have the impression that first day users can easily reach the 4,3 out of the box since they complain about having their core on gaming "only" at 4.1 and having 4.35 when not in game