• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

Stuck cores move location in other temp progs

baldrick1001

New Member
Joined
Jun 15, 2010
Messages
5 (0.00/day)
A bit of a cheekly question really, I have had no problem with realtemp moniotring my core temps, I have two stuck sensors, one that sits at 39 and one that sits at 43 - but they go up when the temperature rises - which is fine by me.

But what I have noticed is that coretemp will randomly, on reboots, change the location of the stuck cores. Generally, i.e. 75% of the time the cores are:

Core #0 = 32c (not a stuck sensor)
Core #1 = 43c (stuck sensor)
Core #2 = 28 (not a stuck sensor)
Core #3 = 39c (stuck sensor)

but sometimes they could look like this:

Core #0 = 32c (not a stuck sensor)
Core #1 = 43c (stuck sensor)
Core #2 = 39c (stuck sensor)
Core #3 = 28 (not a stuck sensor)

i.e. the sensor stuck at 39c has moved to from core 3 to core 2.

Why would this prog move the cores around when real temp keeps them static (aside from realtemp being the platinum product). Could this be a probelm with my motherboard, cpu pins or the programme? Is there, for instance, a cpu pin per sensor??


My system specs are:

Q9650
Gigabyte EP45-UD3R
4 gig RAM
80 GIG SSD
Other stuff.
 
Joined
Sep 8, 2005
Messages
836 (0.12/day)
Location
Romania
System Name Rocket
Processor Ryzen 3600X
Motherboard ASRock B450 PRO4
Cooling Noctua NH-D15
Memory HyperX Predator Black 16GB DDR4 3200MHz CL16
Video Card(s) MSI GeForce GTX 1060 GAMING X 6GB
Storage ADATA SX8200 PRO 512GB + Intel 535 Series 120GB + WD 6400AAKS
Display(s) Benq EW2420
Case Antec P182
Power Supply Antec Signature 650
Software Win 10
This is another nice feature of RealTemp, using correct core order for APIC ID.
When CoreTemp shuffle your cores take a look at APIC ID in RealTemp. ;)
RealTemp sorts the temperature data into the correct physical order while Core Temp does not.

apic_id.png

Take a look at a Xeon with HT. :D

apic_id_xeon.png
 

baldrick1001

New Member
Joined
Jun 15, 2010
Messages
5 (0.00/day)
wow- now you've lost me.. What is APIC and why do the two pics that you have attached show different APIC ID's? Sorry if I'm being a little thick.

Can I take it that the coretemp core sensor shuffle, as I will call from this day hence, is not a motherboard/cpu pin problem but a coretemp software problem?


Thanks for help..
 
Joined
Jan 14, 2009
Messages
2,644 (0.48/day)
Location
...
System Name MRCOMP!
Processor 5800X3D
Motherboard MSI Gaming Plus
Cooling Corsair 280 AIO
Memory 64GB 3600mhz
Video Card(s) GTX3060
Storage 1TB SSD
Display(s) Samsung Neo
Case No Case... just sitting on cardboard :D
Power Supply Antec 650w
Real temp sorts out witch core is witch, and correctly assigns core 1 as core 1 and core 2 as core 2

core temp just randomly selects core 1 - 4 and puts them in any order.
core 1 is core 3.
core 2 is core 1.
random




etleast thats what burbista said. (what i interperted he said.)
 
Joined
Sep 8, 2005
Messages
836 (0.12/day)
Location
Romania
System Name Rocket
Processor Ryzen 3600X
Motherboard ASRock B450 PRO4
Cooling Noctua NH-D15
Memory HyperX Predator Black 16GB DDR4 3200MHz CL16
Video Card(s) MSI GeForce GTX 1060 GAMING X 6GB
Storage ADATA SX8200 PRO 512GB + Intel 535 Series 120GB + WD 6400AAKS
Display(s) Benq EW2420
Case Antec P182
Power Supply Antec Signature 650
Software Win 10
Correct slyfox2151. :toast:
But it's not entirely CoreTemp fault. Windows at every boot (can) shuffle cores so CoreTemp pick whatever Windows shows.
 

unclewebb

ThrottleStop & RealTemp Author
Joined
Jun 1, 2008
Messages
7,249 (1.26/day)
Most software, including Windows and the task manager, assumes that the cores are ordered sequentially so a Quad Core would look like this to Windows:

0123

Monitoring software that asks for the temperature of core 0 gets data from core 0. If it asks for core 1 data, it gets core 1 data, etc., etc.

The problem is that sometimes when you boot up, this order is not sequential but Windows and Core Temp and most every other program still assumes it is.

What's happening in your case is the two end cores are getting swapped. If 0123 is the normal order for your computer then it has changed to 0132.

After this happens, when software asks for the temperature of Core #2, Windows puts that request on the wrong core and returns the temperature data from Core #3. When software asks for the temperature of Core #3, Windows returns the information for Core #2.

Core 0 always seems to be in the correct position but after that, the cores can fall in any order like this:

0123
0132
0213
0231
0312
0321

Monitoring software that makes assumptions about how the cores are ordered is going to be in for a surprise when you boot up and the order has changed to an order that is not 0123. The Core Temp reported core temperatures will not be in the correct physical order.

APIC ID is sort of like the serial number for each core. Since you can't depend on Windows to give you the correct order, you have to check out the ID number of what core your temperature request was actually running on. That's what RealTemp does and no other software bothers to do.

When you look in the RealTemp Settings window, it will tell you how your cores are seen by Windows and what order they are in. This is reported in the APIC ID box. If that box is presently showing 0132, it is very likely that a day or a week or a month from now it will switch back to 0123 and you will notice in Core Temp that your cores have swapped again. The center cores of my Q6600 used to do this once in a while for no apparent reason and about a month later it would just swap back for no apparent reason.

No worries. Just one of those weird things that can happen. Your cores really aren't moving around inside your CPU. :)
 

unclewebb

ThrottleStop & RealTemp Author
Joined
Jun 1, 2008
Messages
7,249 (1.26/day)
Core ordering is a bigger problem for the newer Core i Quad CPUs that support hyper threading. Instead of 4 threads now there are 8 threads and after thread 0, the order can be all over the place. Thread 0/1 belong to core 0, thread 2/3 belong to core 1, thread 4/5 belong to core 2 and thread 6/7 belong to core 3. The two typical core orders are this:

02461357
01234567

In the first case, the first 4 threads represent the 4 different cores so if you ignore APIC ID and read the first 4 threads looking for temperature information, software will get lucky and report temperature data from each of the 4 cores and it will be in the correct physical order.

If the threads are ordered sequentially such as the second example and software reads the first 4 threads, it will report the core temperature from thread 0 and thread 1 individually even though the data for those is both coming from the same core 0. It will do the same thing with thread 2 and thread 3 so it ends up reporting the temperature for core 0 twice and reporting the temperature for core 1 twice. It reports 4 temperatures but that only represents data from the first 2 cores and core 2 and core 3 are not reported at all. This has been going on in some commercial monitoring software since the Core i7 was first introduced.

Before RealTemp reads a single temperature sensor, it goes through the threads one by one and organizes them so the reported data is coming from the correct physical core. Other monitoring software doesn't bother.

Users then make comparisons to other programs and when 5 programs report one thing and RealTemp reports something different then they assume that RealTemp must be wrong. With some software reporting core 1 data as core 3, it's impossible to make a fair comparison.

Here's a good example that rge at XtremeSystems sent me a long time ago when his beta bios was having a bad day. The thread order was all over the place and any software that ignored this would not report temperature data from the correct physical core.



The 4 cores are in one order and the 4 hyper threads follow and they are in a completely different order with core 0 data being on both the first and the last thread.
Any simple algorithm that doesn't check the APIC ID of each thread would fail to be accurate.
 

baldrick1001

New Member
Joined
Jun 15, 2010
Messages
5 (0.00/day)
Thanks for your help with this - you really have helped me to understand the finer points of core temperature management. Seeing as you've added a picture, I thought I'd do the same - just to show how spurious the core sensor management can be - when using, let’s say, programmes that are not as refined as Real Temp. As I noted before, if I were to reboot my PC five times in succession I would expect the core temps to be reallocated at least once - seems strange that windows would do that, but I now feel comfortable in the knowledge that it’s not a problem specifically with my PC, it's just one of those windows things.

Thanks to all for your time with this andespecially for explaining the how's and why's. It really is very much appreciated.


[/IMG]
 

unclewebb

ThrottleStop & RealTemp Author
Joined
Jun 1, 2008
Messages
7,249 (1.26/day)
Next time this happens just have a look at what APIC ID RealTemp is reporting in the Settings window. That funny number in the box should make some sense now. :)

The first time my Q6600 reported that the center two cores had swapped positions I was scratching my head because all software including the Task Manager kept telling me the same thing. That's when I decided to look into this further and was able to come up with a solution for RealTemp. At least one program reports this correctly now. The hard part is trying to convince others that one is right and the rest are wrong.
 
Top