How to detect/observe CPU power (thermal) throttling?

Got a shopping cart of parts that you want opinions on? Get advice from members on your planned or existing system (or upgrade).

Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee

Post Reply
quest_for_silence
Posts: 5275
Joined: Wed Jun 13, 2007 10:12 am
Location: ITALY

How to detect/observe CPU power (thermal) throttling?

Post by quest_for_silence » Thu Aug 09, 2012 2:33 am

Are AMD (and Intel, of course) CPUs supposed to scale down core their clock frequency/multipliers when any overheating occurs?

In case, is there any setting, method, technique, or software able to show those throttling occurrences in any case?

The question may look rather obvious, but seemingly I'm not able to actually detect any change in cores clock frequency (I've used AMD OverDrive 4.2.1, CPU-Z 1.57, RMClock 2.3.5 and Throttle Watch 2.02), expectedly due to thermal throttle (currently on three AMD 10h CPUs, Phenom II X4 980, Athlon II X4 605e, Athlon II X3 420e), up to 117° C reported core temps (I used two more tools, Core Temp and SpeedFan).

Das_Saunamies
*Lifetime Patron*
Posts: 2000
Joined: Tue May 15, 2007 1:39 am
Location: Finland

Re: How to detect/observe CPU power (thermal) throttling?

Post by Das_Saunamies » Sun Aug 19, 2012 2:24 pm

I'm not sure AMD throttles, to be honest, I know most if not all Intel CPUs do. If you can't see it even when the CPU gets that hot, it's probably not there. Would need to verify the feature exists.

Spoon Boy
Posts: 168
Joined: Sun Jan 23, 2005 11:44 pm
Location: London

Re: How to detect/observe CPU power (thermal) throttling?

Post by Spoon Boy » Mon Aug 20, 2012 5:32 am

Not sure about AMD, Ive only used this on i7.
But Ive used ThrottleStop in the past, there is also some good info on the ins and out of thermal throttling.

MikeC
Site Admin
Posts: 12285
Joined: Sun Aug 11, 2002 3:26 pm
Location: Vancouver, BC, Canada
Contact:

Re: How to detect/observe CPU power (thermal) throttling?

Post by MikeC » Mon Aug 20, 2012 7:15 am

afaik, throttling is featured in all modern cpus -- for many years now. It came more or less in conjunction with heatspreaders. One obvious & simple way to detect is with an AC power meter -- total power draw starts yoyoing down & up as throttling begins when a steady state stress utility like prime95 is used.

Pappnaas
Posts: 726
Joined: Mon May 16, 2011 11:23 am
Location: Germany

Re: How to detect/observe CPU power (thermal) throttling?

Post by Pappnaas » Mon Aug 20, 2012 10:08 am

Spoon Boy wrote:Not sure about AMD, Ive only used this on i7.
But Ive used ThrottleStop in the past, there is also some good info on the ins and out of thermal throttling.
"Cool & Quiet" is AMD's name for throtteling technique.

MikeC
Site Admin
Posts: 12285
Joined: Sun Aug 11, 2002 3:26 pm
Location: Vancouver, BC, Canada
Contact:

Re: How to detect/observe CPU power (thermal) throttling?

Post by MikeC » Mon Aug 20, 2012 10:46 am

Pappnaas wrote:"Cool & Quiet" is AMD's name for throtteling technique.
I don't think that's quite right. CnQ applies only in low power states, to reduce power in idle, much like Inte's Speedstep. This was adopted from Notebooks. Not sure if AMD (or Intel, for that matter) actually named the throttling-at-max-power utility.

andyb
Patron of SPCR
Posts: 3307
Joined: Wed Dec 15, 2004 12:00 pm
Location: Essex, England

Re: How to detect/observe CPU power (thermal) throttling?

Post by andyb » Mon Aug 20, 2012 12:24 pm

Not sure if AMD (or Intel, for that matter) actually named the throttling-at-max-power utility.
This thread has put doubt into my mind, I have never seen a CPU throttle for certain, i.e. I didnt have CPU-Z installed and watched the clock speeds whilst it was overheating, but I was quite certain that this feature was introduced first with the P4, and then the Athlon, and I would assume every CPU since.

Many BIOS's have a "CPU throttling option" with they typical options of 60C, 65C and 70C, I have always set it at 60C when given the choice - But I have never actually put this to the test, although I have known some laptops to slow down when their heat-sinks are clogged with dust, fluff and hair, but I still don't have any personal evidence......


Andy

Spoon Boy
Posts: 168
Joined: Sun Jan 23, 2005 11:44 pm
Location: London

Re: How to detect/observe CPU power (thermal) throttling?

Post by Spoon Boy » Mon Aug 20, 2012 2:08 pm

I found this handy info on Intel's way of controlling things when a CPU starts to cook its self, its a little technical and goes into a fair bit of detail.
Thermal-Management-Logic-and-Thermal-Monitor-Feature

The short version if you don't want to be bored by Intel documents, there are two basic ways CPU overheating is controlled,
one is PROCHOT (short for processor hot)
basically a single bit flag on each CPU die that gets tripped when the TCC (Thermal Control Circuit) limit (set by INTEL) is reached this pretty much just tells the board to cut power to the CPU.

The other methods and probably more relevant to the OP's question about lowering the frequency of the CPU are called TM and TM2, (Thermal Monitor,Thermal Monitor2)
Basicly TM changes the duty cycle of the internal processor clocks
And TM2 causes the processor to adjust its operating frequency (by dropping the bus-to-core multiplier to its minimum available value) and input voltage identification (VID) value. This combination of reduced frequency and VID results in a reduction in processor power consumption.

Erm sorry for the longish post i got a bit carried away :lol:

andyb
Patron of SPCR
Posts: 3307
Joined: Wed Dec 15, 2004 12:00 pm
Location: Essex, England

Re: How to detect/observe CPU power (thermal) throttling?

Post by andyb » Mon Aug 20, 2012 2:30 pm

The short version if you don't want to be bored by Intel documents, there are two basic ways CPU overheating is controlled,
one is PROCHOT (short for processor hot)
basically a single bit flag on each CPU die that gets tripped when the TCC (Thermal Control Circuit) limit (set by INTEL) is reached this pretty much just tells the board to cut power to the CPU.

The other methods and probably more relevant to the OP's question about lowering the frequency of the CPU are called TM and TM2, (Thermal Monitor,Thermal Monitor2)
Basicly TM changes the duty cycle of the internal processor clocks
And TM2 causes the processor to adjust its operating frequency (by dropping the bus-to-core multiplier to its minimum available value) and input voltage identification (VID) value. This combination of reduced frequency and VID results in a reduction in processor power consumption.

Erm sorry for the longish post i got a bit carried away :lol:
Don't apologise for such a short and very, very sweet post.

The TM/TM2 rings some bells from the distant past, and I would love to have a system to torture to see it in action for myself, just because :twisted:


Andy

Spoon Boy
Posts: 168
Joined: Sun Jan 23, 2005 11:44 pm
Location: London

Re: How to detect/observe CPU power (thermal) throttling?

Post by Spoon Boy » Mon Aug 20, 2012 4:27 pm

I couldn't sleep thinking over what the OP was asking about,
The question may look rather obvious, but seemingly I'm not able to actually detect any change in cores clock frequency (I've used AMD OverDrive 4.2.1, CPU-Z 1.57, RMClock 2.3.5 and Throttle Watch 2.02), expectedly due to thermal throttle (currently on three AMD 10h CPUs, Phenom II X4 980, Athlon II X4 605e, Athlon II X3 420e), up to 117° C reported core temps (I used two more tools, Core Temp and SpeedFan)
From what i can remember most software takes best guess approaches to measuring the true temperature of modern day CPU's speaking about Intel,
And i dare say it's the same for AMD, a quote from the documents of Realtemp.

Intel designed these temperature sensors to control thermal throttling and thermal shut down and for those purposes,
they tend to work excellent. They were never designed to be used to report accurate core temperatures.


So it's hard to know for sure when the TJmax (Intel) or TCase Max (AMD) is reached as TJMax (IDK about AMD) is not a clearly defined value.

If you wanted to know a best guess for when the CPU is throttling due to overheating you would be looking for the TJMax or TCase Max for that particular CPU,
AFAIK Core Temp like all other software uses a best approximation approach.

You would then have to use something like core temp to get a rough idea of what the CPU's TJMax or TCase Max is,
As well as using something like CPU-Z at the same time to view the clock multiplier, core voltage, and Frequency of the CPU.
Watch for these to drop when you get close (could be +-10c) to TJMax, TCase Max.

All the while running your favorite stress test, Personally i used IntelBurnTest works on AMD to.
But if i'm honest i would be very worried if any CPU reached 117° C :shock: like the OP said IDK if AMD run hot normally but there comes a point when
reducing the noise a PC makes can be taken to far if your going to cook you components due to lack of good air flow.

As for the other end of the scale like MikeC said AMD used to use C'n'Q I think they now call it P-States and for Intel its C-States,
but that's a hole other kettle of fish. :roll:

Post Reply