What does Windows task manager CPU speed actually report?

[Tegneren]

Anyone know what the CPU speed in task manager is actually reporting?

 

I have a laptop with a i5 12500H and windows task manager never shows speed higher than about 2.5 GHz, even under load. Mostly it reports between 1.8-2.2 GHz.

HWinfo64 and MSI afterburner reports the speeds I expect based on the specs of the cpu. up to 4,5/3,3 GHz depending on the load.

 

I just want to know why there are such big difference.

On my gaming pc with a Ryzen 5800X its at least relatively close

[Kilrah]

Displaying one CPU speed value that's meaningful is pretty much impossible nowadays since you have 12 cores that can start/stop/change speed wildly hundreds of times per second each, so pretty much every program has to figure its own way of averaging things.

[Mark Kaine]

in my case its absolutely 100% correctly reporting overall frequency at any given point (win ver 1809) good luck getting that kind of accuracy with any other win ver tho.

 

(and yes i checked with hwinfo64 and ryzen master)

 

13 minutes ago, Tegneren said:

have a laptop with a i5 12500H and windows task manager never shows speed higher than about 2.5 GHz, even

even on my laptop it seems to reporting inline with intel XTU... (win ver 1803)

 

 

tbh it's not a big deal, just use hwinfo64, its mostly very accurate. 

[Tegneren]

13 minutes ago, Kilrah said:

Displaying one CPU speed value that's meaningful is pretty much impossible nowadays since you have 12 cores that can start/stop/change speed wildly hundreds of times per second each, so pretty much every program has to figure its own way of averaging things.

Sure, but it should at least show something that is representative of what its actually doing, and not show idling speed because one of the twelve cores isn't doing anything.

16 minutes ago, Mark Kaine said:

tbh it's not a big deal, just use hwinfo64, its mostly very accurate. 

Getting accurate numbers wasn't the issue, I just wanted to know what it is its actually reporting. Like, is it the average across all cores, or is it using some kind of formula or something else? 

I dont think its the average based on the speeds I see even under all core loads like cinebench.

[Mark Kaine]

10 minutes ago, Tegneren said:

I dont think its the average based on the speeds I see even under all core loads like cinebench.

yeah good question,  i think its either actually average or max... would need to check again, as said hwinfo64 ryzen master and tskmngr all report the same values for me (with slight delays, which shouldn't really matter) because i don't actually remember *what* its reporting exactly lol

[Mark Kaine]

there you go

 

 

core clocks, which is pretty much obviously an average 

 

4.12ghz <‐> 4.116ghz respectively (so as said, the *same* ... no doubt)

 

so simple to check too! : )

 

 

 

[Jinchu]

19 minutes ago, Tegneren said:

Sure, but it should at least show something that is representative of what its actually doing, and not show idling speed because one of the twelve cores isn't doing anything.

I've given up trying to get meaningful information out of the Windows Task Manager regarding the Ryzen CPUs. Both ones that I've had, had the same issue. It doesn't seem to even correlate much with the workload. I've used ryzen master to check things. hwinfo64 could be a good chose as well.

I'd also be interested in know what the Windows task manager is showing and why it seems to work somewhat differently with Intel CPUs

[Mark Kaine]

2 minutes ago, Jinchu said:

I've given up trying to get meaningful information out of the Windows Task Manager regarding the Ryzen CPUs. Both ones that I've had, had the same issue. It doesn't seem to even correlate much with the workload.

disagree,  as per my example above it does correlate precisely with the workload... you just have to know where you're looking and use the right tools.

 

admittedly i do think the win version does play a role too tho,  accuracy of tskmngr varies wildly between then so YMMV indeed 

[Jinchu]

1 minute ago, Mark Kaine said:

disagree,  as per my example above it does correlate precisely with the workload... you just have to know where you're looking and use the right tools.

 

admittedly i do think the win version does play a role too tho,  accuracy of tskmngr varies wildly between then so YMMV indeed 

Yeah. I admit that when I was trouble shooting an issue I just grabbed the next tool and didn't understand what Windows Task Manager was showing. At that point I was more interested of the clocks in a single core, for which I should have used other tools, right?

[Mark Kaine]

31 minutes ago, Tegneren said:

Sure, but it should at least show something that is representative of what its actually doing,

yes, and it is!

 

as said ymmv, but you gotta give the receipts too! "it doesn't work" isnt very convincing to me , no offense, when it's working just fine for me, since years...?  : )

[Mark Kaine]

1 minute ago, Jinchu said:

Yeah. I admit that when I was trouble shooting an issue I just grabbed the next tool and didn't understand what Windows Task Manager was showing. At that point I was more interested of the clocks in a single core, for which I should have used other tools, right?

yeah... it took me ages to learn how to properly read and interpret these things... tbh i think like 80% of all threads here are people misinterpreting the stuff they see (my cpu isnt 100% so it cant be a bottleneck!! etc)

 

im not blaming anyone,  there's lots of tools out there that don't work properly,  explanations online are usually contradictory,  you just basically have to observe and figure it out yourself. 

OR JUST USE HWINFO64 ITS MOSTLY CORRECT 

 

(but that can still be misinterpreted tbf ) 

[Mark Kaine]

6 minutes ago, Jinchu said:

At that point I was more interested of the clocks in a single core, for which I should have used other tools, right?

hwinfo64 

[Kilrah]

54 minutes ago, Tegneren said:

I just wanted to know what it is its actually reporting. Like, is it the average across all cores, or is it using some kind of formula or something else? 

Well nobody knows, that's not documented AFAIK...

 

HwInfo tells you that "core clocks" isn't meaningful on modern processors because it doesn't take sleep states into account

 

 

Effective clock is, but also doesn't match Task Manager, and since it'll almost always be super low it's not "satisfying". TM likely takes a whole bunch of metrics and does a weighted average on those, maybe giving some weight to the highest clocked core and its utilization, but weighted down if all others are idle,...

 

You can run a single core stress test, and no single core is fully used because the OS keeps switching which core it's running on. The "clocks" of all cores may be X but most of the cores are in sleep states so this clock means little, etc.

With today's complex CPUs it's at best a guessing game. Gets even worse when you have P and E cores... you basically never will get 2 tools to match.

 

 

[Tegneren]

1 hour ago, Mark Kaine said:

core clocks, which is pretty much obviously an average 

Here are two samples from my r7 5800X sowing its not matching where it should according to you.

First one its only running FAH on the GPU, so not much CPU load. Second is with all core Cinebench, and even there its 0,3GHz difference even though the average should obviously be 4,6GHz.

 

And this isn't even with the Intel cpu I based my question on with its P-core and E-core that runs on different base/boost speeds(I have to wait until after work to get actual numbers from that.

 

[Mark Kaine]

12 minutes ago, Tegneren said:

Here are two samples from my r7 5800X sowing its not matching where it should according to you.

First one its only running FAH on the GPU, so not much CPU load. Second is with all core Cinebench, and even there its 0,3GHz difference even though the average should obviously be 4,6GHz.

 

And this isn't even with the Intel cpu I based my question on with its P-core and E-core that runs on different base/boost speeds(I have to wait until after work to get actual numbers from that.

 

yup... it doesn't match... 

 

as said (and shown) i never see these kind of discrepancies (on any ryzens i had)

 

is this windows 11?  either way i think outside of installing the necessary windows version this can't be fixed by user, unfortunately.