Question CPU Core and Package temperature discrepancy on a brand new Asus Rog Strix laptop ?

Status
Not open for further replies.
Nov 16, 2023
12
1
15
Hi, I have an issue on an Asus Rog Strix laptop (2023 version) with an Intel i9-13980HX CPU.

CPU package is always between 10 and 15 degrees C higher than any of the cores.

Attached image shows the values for my CPU on the left and the reference 13980hx in the same type Asus Rog Strix product on the right.

Can someone help me figure out what the issue is?

Also, the laptop is under extended warranty.

I had a discussion with an asus moderator at this link, but I was not able to get anything useful out of this person:

Thanks

Comp-1.jpg
 
package is the exterior of the CPU, so yes, its always more than core temp. Its normal. My 5800x3d cores sit on 34c at idle and package is on 46c.. Package temp fluctuates more than individual cores as it is a sum of all of them.

Max core temp isn't related to package temp apart from being a sensor inside the CPU. Its a coincidence they have same value in that shot.

Max core temp is probably a result of one of the core sensors. It = P core 0 temp on the hwinfo screenshot. the other total could be any of the E cores since a few show 58.

I don't think I ever seen package & cores at same temp.
I can get close under load
n0J82hF.jpg
 
Last edited:
Edit: your temperatures are what I expected to see on my CPU , you have maybe a degree C of delta.
I have 15 degrees difference between the hottest core and the package... 😆

Hi, so there is a hotter thermal source on the outside of the CPU? a power conduit maybe..
My biggest issue is that I am not seeing all of these hot packages on the HWinfo data of other CPUs, even other 13980hx's, Package temperature is always the hottest core. please check the attached image taken from another 13980hx.

multi-core-test.png


Thanks.
 
Last edited:
  • Like
Reactions: John Chesterfield
That appears to be a bug. Its counting the package as a core. What does GPU Z show?

try updating hwinfo, I am on 7.62.5200 and there is likely a newer one since.
Ah sorry my bad for not explaining.

On 13980hx you can disable cores you don't need and that is why its only showing 7 cores and the package. The package is the whole of the CPU.

I was disabling cores to see if the package will snap back to expected values, it didnt.
 
updating hwinfo won't hurt.

none of the cores in your listing have the same temp as the Core temps max of 96. The only temp that is 96 is the Package. It isn't a core. That is what I was getting at. The disabled cores don't add to the temp.

On my pc the core temp max is the same as one of my core temps.
CPU IA Cores is the temperature of x86 (IA) execution cores only, GT cores covers the integrated graphics part of the CPU.
CPU Package is the 256 ms averaged temperature of the hottest part of the entire CPU package (including IA, GT, uncore, etc). The second value is retrieved via a different source and should be equal, however on some systems there might be an offset.
from author of hwinfo
 
Last edited:
The Cinebench example above is not my CPU, its some other 13980hx results I found on the internet - as I wanted to check first if its normal behavior to have a large offset between the 2 temperatures (max Core temp and Package temp, as you can see from the reference they have to be the same)

Updated Hwinfo, no change:

Screenshot-2023-11-16-133822.png
 
oh sorry.

I feel like I am talking past you. Or missing the point.

I probably be better with some sleep... only slept 3 hours in last 30. Could be it.

I have asked other people to look in here, to see if they can work out what I am missing.

I will look in here again when I am awake :)
 
oh sorry.

I feel like I am talking past you. Or missing the point.

I probably be better with some sleep... only slept 3 hours in last 30. Could be it.

I have asked other people to look in here, to see if they can work out what I am missing.

I will look in here again when I am awake :)
No worries, and thank you for your efforts!

Just a quick up to speed explanation -

The package is the unit containing all of the cores so it would be logical to expect that the package temperature sensor reports the highest core temperature (doesn't matter which core has it)
I would also expect there to be a small plus minus between the 2 as there may be other elements in the package contributing to the hottest status.

However after looking at lots of HWinfo data I found that the package temperature is....(drum roll)..the highest core temperature. (duh)

Onto my CPU - the package is 15 degrees C hotter than the hottest core. Why?
 
Last edited:
Why is your reference screenshot showing E-core readings, but the screenshots of YOUR CPU are not? If you'll note, the E-core readings on the reference screenshot are hotter than any of your P-core readings, which is probably because the system is utilizing them more AND they MIGHT tend to run hotter than the P-cores if you are not pushing the P-cores at the time because system processes tend to utilize the E-cores more to leave the P-cores available for user operations. At least, they do on my system. And if what's running is in any way an intensive windows process they ARE sometimes about 8-10 degrees hotter than my P-cores on my 12700k.

So if you are not seeing the E-core temperatures there because they are not expanded, then you may NOT being SEEING the hottest core, which you are right, is generally the same as the package temp, or close to it. Although, package temp is not JUST the hottest core, it is the temperature of the ENTIRE package, and let's face it, if you have one blowtorch heating a piece of steel that piece of steel is not going to be the same temperature as if you have ten blowtorches heating the same piece of steel. Generally I am seeing about a 1-2 degree difference between my hottest core and the package temp on my 12700k platform but there HAVE been occasions, especially while running demanding applications or Prime95, that the package temp has been significantly higher than my hottest core to the tune of between 5-8 degrees. That is likely going to be MORE on your system, because you have a laptop where the CPU is in an enclosed space with only limited ability to cool while mine is in a mid tower case with extensive cooling capabilities.
 
@Vdesign83

For Intel CPUs, the package temperature data is the highest temperature found anywhere within the CPU package. It is not the exterior temperature of the CPU. Intel CPUs sample temperature data from the individual cores, the Intel GPU and there are a few more temperature sensors within the package that it gets data from. All of these individual sensors are compared and the highest temperature found is reported as the package temperature. When running a stress test like Cinebench, the package temperature should be the same as the hottest core temperature. At times there might be a tiny difference but when a CPU is equally loaded, there should not be a 13°C difference like you are seeing.

Are you seeing a 13°C difference when your 13980HX is fully loaded running Cinebench? Try running ThrottleStop 9.6 for a second opinion.


ThrottleStop shows the package temperature on its main screen. Run Cinebench and see how ThrottleStop's package temperature number compares to the individual core temperatures that ThrottleStop reports. Exit HWiNFO and enable all of your P and E cores when testing.

vxqWIIa.png


If you still have the same problem then it is possible that one of the temperature sensors within the CPU package was poorly calibrated. Intel used to say that these sensors are only accurate to +/- 5 °C. Perhaps one of your sensors is outside that limit.

You are correct that the package temperature reporting too high could result in premature thermal throttling. Post a ThrottleStop screenshot while the CPU is loaded that shows the difference you have found. Idle or light load temperature data is fairly meaningless.

Even if you can prove to yourself that the temperature sensors within your CPU are poorly calibrated, it will be next to impossible to get Asus or Intel to agree with what you have found let alone offer to replace your laptop because of this.
 
Why is your reference screenshot showing E-core readings, but the screenshots of YOUR CPU are not? If you'll note, the E-core readings on the reference screenshot are hotter than any of your P-core readings, which is probably because the system is utilizing them more AND they MIGHT tend to run hotter than the P-cores if you are not pushing the P-cores at the time because system processes tend to utilize the E-cores more to leave the P-cores available for user operations. At least, they do on my system. And if what's running is in any way an intensive windows process they ARE sometimes about 8-10 degrees hotter than my P-cores on my 12700k.

So if you are not seeing the E-core temperatures there because they are not expanded, then you may NOT being SEEING the hottest core, which you are right, is generally the same as the package temp, or close to it. Although, package temp is not JUST the hottest core, it is the temperature of the ENTIRE package, and let's face it, if you have one blowtorch heating a piece of steel that piece of steel is not going to be the same temperature as if you have ten blowtorches heating the same piece of steel. Generally I am seeing about a 1-2 degree difference between my hottest core and the package temp on my 12700k platform but there HAVE been occasions, especially while running demanding applications or Prime95, that the package temp has been significantly higher than my hottest core to the tune of between 5-8 degrees. That is likely going to be MORE on your system, because you have a laptop where the CPU is in an enclosed space with only limited ability to cool while mine is in a mid tower case with extensive cooling capabilities.
Please read what I wrote.

Only 7 P cores are on in my configuration, the rest are disabled.

Additionally: All cores being on, some being on, integrated gpu on or off doesnt make a difference at all sadly.

The CPU being loaded or not also doesn't matter, I have 15 degrees C hotter package than any of the cores.
(so if the hottest core is at 50 degrees the package will be at 65, if the hottest core is at 75 the package will be at 90) Does that clarify the situation?

thanks
 
It would be very unlikely though, since Tjunction on that CPU is 100°C and he's nowhere near that.

It's absolutely "possible" to have a poorly calibrated thermal sensor, I agree with that because I have seen it before, but if so, then it must be common on 12th and 13th gen platforms because I'm seeing disparity between the hottest core and the package temp on my own 12th Gen i7 system AND I saw it on two other systems, one a 12th Gen and the other a 13th Gen, that I build not too long ago for a client, while doing the completed build testing.

@Vdesign83 , do us all a favor. Download and run Prime95. Choose the "Small FFT" option, not "Smallest FFT", just Small FFT. Make sure to disable all of the AVX options before running the test. While that is running take full screenshots of all sensor data in HWinfo and be sure to include/expand the values for both the P-cores and E-cores.

Also, I'd recommend you download Core Temp and check the values there as well to see if there is any discrepancy between what you see there and HWinfo, which I doubt, but it's worth being sure of.
 
Please read what I wrote.

Only 7 P cores are on in my configuration, the rest are disabled.

Additionally: All cores being on, some being on, integrated gpu on or off doesnt make a difference at all sadly.

The CPU being loaded or not also doesn't matter, I have 15 degrees C hotter package than any of the cores.
(so if the hottest core is at 50 degrees the package will be at 65, if the hottest core is at 75 the package will be at 90) Does that clarify the situation?

thanks
I must have either missed this information earlier or you didn't post it, but, if you have the same issue with a 15 degree difference no matter whether there is a full load, a partial load or no load at all (Read: Sitting idle for five or more minutes with no user applications running, only background processes), then there just about absolutely has to be a problem with one of the thermal sensors HOWEVER none of it really matters AT ALL so long as neither cores nor package are coming close to Tjmax or Tjuncture.
 
Hi, @uWebb

I can see the package temperature offset no matter what the CPU is doing.

Please take a look at the screenshot from throttlestop, same behavior can be observed.

Screenshot is from CPU idling in windows, I will try to give you the loaded CPU temps as well with all cores on.

screenshot.png



Thanks.
 
It would be very unlikely though, since Tjunction on that CPU is 100°C and he's nowhere near that.

It's absolutely "possible" to have a poorly calibrated thermal sensor, I agree with that because I have seen it before, but if so, then it must be common on 12th and 13th gen platforms because I'm seeing disparity between the hottest core and the package temp on my own 12th Gen i7 system AND I saw it on two other systems, one a 12th Gen and the other a 13th Gen, that I build not too long ago for a client, while doing the completed build testing.
If you can show me an Intel I9 CPU - does not even have be a 13980 hx, with a similar package offset I would be quite happy with that! Thanks
 
I think you must have missed my last post, because at that point I'm having to agree with the probability that there is an issue with one of the thermal sensors. There is no way that at "idle" you should be seeing a 15 degree difference between the hottest core and package temp, unless something is wrong somewhere. Bad sensor or hot iGPU. Something. At this point, under warranty, I think I'd be asking for a replacement because even a 10 degree variance between cores is a warrantable condition in the eyes of Intel, so if there is still a manufacturer's warranty on the laptop itself they ought to honor it. Especially since you can show the discrepancy between cores and package under all conditions.
 
Last edited:
  • Like
Reactions: Vdesign83
Okay, additional testing is in.

I turned on all cores and set CPU to performance mode and gave it a high stress task. (Rendering)

The results were as expected - package leading the cores by 13 degrees until the moment when P Core 0 and the Package met at 97 degrees.

Meeting.png




In conclusion: It could be that there is a safety offset that decreases as the temperatures approach TJ, I just haven't seen this on any other CPU or other 13980hx's.
I would also say that if this is by design it would be a really bad idea because it would throttle the CPU in in its best temperature range (mid 70s to low 80s degrees C)

Data attached.
 
It could be that there is a safety offset that decreases as the temperatures approach TJ
There is no safety offset. Intel core temperature sensors are calibrated so they can accurately trigger thermal throttling. That is there one and only purpose. They have never been designed to give accurate temperatures, especially when a computer is lightly loaded or idle.

The amount of error in these sensors increases at lower temperatures. That is fine. Accurate idle temperatures is not their purpose. Your screenshot shows one core at a maximum temperature of 95°C and another core at a max of 97°C. The package temperature is also 97°C and your CPU is thermal throttling like it should.

it would throttle the CPU in in its best temperature range (mid 70s to low 80s degrees C)
Can you post a screenshot of that? You could use ThrottleStop to lower your P cores speed so these cores run fully loaded at about 80°C or 85°C. Does ThrottleStop show thermal throttling during this test?

The biggest problem I see is that you have a wide core to core variation in the maximum core temperatures. This can be caused by poor thermal paste application. It is also possible that the top of the CPU or the heatsink is not perfectly flat. This could be the source of the odd package temperature problem you have found. I would disassemble for a closer look. There are also issues with thermal paste pump out on mobile 13th Gen CPUs.

What is the PROCHOT throttling temperature set to? That information was cropped off the bottom of the ThrottleStop picture you posted.
 
There is no safety offset. Intel core temperature sensors are calibrated so they can accurately trigger thermal throttling. That is there one and only purpose. They have never been designed to give accurate temperatures, especially when a computer is lightly loaded or idle.

The amount of error in these sensors increases at lower temperatures. That is fine. Accurate idle temperatures is not their purpose. Your screenshot shows one core at a maximum temperature of 95°C and another core at a max of 97°C. The package temperature is also 97°C and your CPU is thermal throttling like it should.

That was my original thinking as well, however as you can see from the cinebench example posted above it is not seen on other 1980hx's. (58 degrees both values)

For me it wouldn't be fine.

The Cpu itself is only able to do like 4000 MHz under load - and only for a minute at that before de-rating.

The tests shall continue at a later date.

*PROCHOT is set at 86 degrees so in reality it is set at 70-72 degrees.

Best regards
 
You shouldn't NEED to do any of that. If it's overheating, there is a problem. It should not, by design, be reaching Tjuncture or Tjmax unless there is a thermal issue. Whether that is inadequate cooling for some reason or overclocking or a faulty sensor or calibration, it is STILL something that shouldn't be happening AND per the reference you yourself posted of another CPU with the same model, theirs does not have the behavior yours has. Being under warranty I wouldn't even mess around.

And for the record, PROCHOT is the same as Tjuncture, which are core temperatures. Tjmax, Tcase and Tcasemax are specifications. Tjuncture/PROCHOT for the 13980hx (And basically ALL 12th, 13th and 14th gen Intel core-i processors) is 100°C and IS the temperature at which it should begin throttling. THERMTRIP/TJmax/Tcasemax temp is like 130°C. BOTH of those numbers are factual based on the Intel data sheets AND have been confirmed by my good friend Computronix, author of the Intel temperature guide which you can find here:


Have in fact had conversations on this very issue in the past as noted below.


There are numerous instances where Intel contradicts their own terms. For example, Intel Extreme Tuning Utility software (XTU) and the Product Specifications website both have inconsistencies with the Datasheets, which use proper terminology.

With respect to terminology, Intel’s Product Specifications website incorrectly shows either “Tcase” or “Tjunction” as specifications. In that context, both are technically improper terms. The Datasheets, which use proper terminology, instead show “Tcase Max” and “Tj Max”. For the record, “Tcase Max” is a specification, while “Tcase” is IHS temperature. Correspondingly, “Tj Max” is a specification, while “Tjunction” is Core temperature.

Some 13 or 14 years ago when I was consulting with Arthur Liberman, who is the author of Core Temp, I brought this very point of terminology contention to his attention concerning his excellent utility. Core Temp was showing "Tjunction" which back in the day, users found confusing. I asked him if he would simply change the term to instead show "TjMax" which he implemented in the next update.

Regarding the 12900KS, those who administrate the Product Specifications website are sorely lacking in attention to detail, since the website has yet to show a value for "Tjunction", which is actually Tj Max, just as you have pointed out. However, as always, the Datasheets show the correct values for 12th Gen processors, and use proper terminology.

Go to the website for the i9-12900KS then scroll down to "Supplemental Information" and click on "view now" next to "Datasheet". Under the 12th Gen Family click on "Volume 1" which will download the correct Datasheet.

Go to Section 4, Thermal Management, then to page 101, which shows Table 25, Thermal Margin Slope. There in the 1st row you will see 8 + 8 Core, 150 Watts which is the 12900KS. Next to that it shows TCC Activation at 100°C which is THERMTRIP and is synonymous with TjMax, just like all the other 12th Gen Desktop processors of various stock coolers, Core counts and Wattages.

You mentioned that "I guess I've been mistaken all along in my belief that the Tjunction temperature listed on the specifications page for each processor IS the Tjunction max, but clearly it seems that is not the case." Actually, that IS the case. Tjuntion on the website IS TjMax in the Datasheets. Here's where your confusion may lie ...

Some 9th through 12th Generation motherboards have a "Feature" that allows the user to violate TJ Max with an "Offset" adjustment in BIOS which can increase Intel's thermal limit from 100°C, where most processors Throttle, to 115°C which can potentially quickly damage a CPU. We see this on the Forums where some knuckleheads think it's a "cool" work-around that can compensate for inadequate cooling.

These posts are very misleading and confusing in that they dupe the uninformed into thinking that TjMax is actually higher than "what they've heard". Tampering with this BIOS setting is is a serious no-no, which I adamantly discourage. The motherboard manufacturers use this "Feature" as a competitive marketing ploy, which was an extremely bad idea from concept to implementation. -Computronix


Sorry, in my above post I mistakenly wrote THERMTRIP, which is Shutdown Temperature at 130°C, while PROCHOT is 100°C or Throttle temperature, where many Generations of Intel processors can continue to run, and in many instances, the uninformed user may not notice that their unmonitored processor is actually showing throttle behaviors.

Paul isn't wrong, but for several reasons, I can't agree or condone operating a CPU at such high Core temperatures for extended periods, or as a matter of routine, without concern for longevity.

For example, Intel's 3 year warranty is now being stretched to the edge of the envelope, as well as other considerations such as nano-stress cracks developing in the top of the silicon Die near the solder TIM interface under the IHS due to extreme thermal cycling over time. Additionally, Electromigration and VT Shift remain real issues which are affected by high operating temperatures, as well as the multiple levels of Turbo Boost and TVB mechanisms which Intel has implemented in recent architectures that require Core temperatures below 70°C, and even 50°C in 12th Gen, in order to operate effectively.

For what it's worth, as I've stated in my Temp Guide, "Core temperatures below 80°C are ideal" still holds true. That's why the Datasheets show the Temperature Control Offset at 20°C, which means it begins at 80°C. Just because these processors will run just fine near or at Throttle temperature doesn't mean that it's OK to disregard the finer aspects of the Specifications, as well as common sense. Prime95 - Small FFTs -No AVX is still a good standard for a steady-state 100% workload, just as CPU-Z - Bench - Stress CPU is still a good standard for a steady-state 80% workload, which matches most heavy gaming workloads. -Computronix
 
I did a lot of testing on this thing, but without boring anybody with the details I will just leave this screenshot here.

Proof.png




I won't be buying anything ASUS in a long long time.
 
I stopped buying, and for the most part even recommending, any ASUS products when they refused to repair or replace my Hero VIII board that had a bad BIOS ROM. I fought and fought with them and then they sent the board back in the same condition I sent it to them. I purchased a damaged board off ebay, swapped ROMs, problem fixed. They could have done the same. One of the mods here had a similar issue where the onboard NIC failed and they sent him back a replacement board with a crushed integrated I/O shield next to the VRM heatsinks. Actually, it was the type where the I/O cover and VRM heatsink are all one unit, and then didn't want to replace it with one that wasn't damaged since the one he sent them wasn't damaged other than the NIC not working.

PLENTY of other stories showing that ASUS has adopted an extremely poor customer service policy since about 2015 whereas prior to that they were always very good. They still make good products, mostly, but when something is wrong, they are crap to try and work with. Which is why I created THIS a while back.

AaR98sG.jpg
 
  • Like
Reactions: Vdesign83
I won't be buying anything ASUS in a long long time.
The package temperature sensors are located inside of the Intel CPU. Why are you blaming Asus for that?

The only purpose of these sensors is to trigger thermal throttling. The previous screenshot you posted shows that the CPU core and CPU package were able to get up to 97°C during a stress test. The package sensor issue that you have does not seem to be interfering with your CPU reaching its maximum safe temperature and maximum performance. If constant throttling was taking place at 80°C or 85°C, it would be impossible for the CPU to ever reach 97°C.

Post a ThrottleStop log file that shows how much this issue is reducing performance. You can copy and paste the log file data to,

www.pastebin.com

Try playing a game or run Cinebench. Create a log file while you are actually using your computer. An HWiNFO screenshot that shows a sensor that might have glitched out for a millisecond does not prove the seriousness of the problem you have.

Your full load testing that you previously posted shows a wide variation in the maximum P core temperatures. That is usually a sign of poor thermal paste application, a heatsink that is not perfectly flat or perhaps the heatsink was not screwed down evenly. There is always room for improvement compared to how a computer is quickly put together on an assembly line.
 
The package temperature sensors are located inside of the Intel CPU. Why are you blaming Asus for that?

The only purpose of these sensors is to trigger thermal throttling. The previous screenshot you posted shows that the CPU core and CPU package were able to get up to 97°C during a stress test. The package sensor issue that you have does not seem to be interfering with your CPU reaching its maximum safe temperature and maximum performance. If constant throttling was taking place at 80°C or 85°C, it would be impossible for the CPU to ever reach 97°C.

Post a ThrottleStop log file that shows how much this issue is reducing performance. You can copy and paste the log file data to,

www.pastebin.com

Try playing a game or run Cinebench. Create a log file while you are actually using your computer. An HWiNFO screenshot that shows a sensor that might have glitched out for a millisecond does not prove the seriousness of the problem you have.

Your full load testing that you previously posted shows a wide variation in the maximum P core temperatures. That is usually a sign of poor thermal paste application, a heatsink that is not perfectly flat or perhaps the heatsink was not screwed down evenly. There is always room for improvement compared to how a computer is quickly put together on an assembly line.
If you don't understand the data on the screenshot posted above there is no benefit for either of us in continuing this debate.

Problem is the cooling system that's wholly inadequate for the task, they made a rescue device by adding 15 degrees to the running CPU temp thus throttling it sooner - It robs me the customer, of the performance I paid for. (3000 Euro for this lemon)

I am happy with this data being here as it gives me a little bit of satisfaction at least.

Best regards
 
Status
Not open for further replies.