Real Temp - New temp program for Intel Core processors

[loonym]

Looks the same as yours unc but if it's useful to you I'm in
X3360

[cadaveca]

My qx9650 shows the same.

[unclewebb]

That data is very useful.

Go back and look at my data. This MSR seems to be showing that my center cores are swapped, which they are. If you look at the EAX column for my processor it shows 0-2-1-3 while your processors both are normal and show 0-1-2-3. This might be a way for me to test Quad cores for the swapped core bug. Hopefully FullSky can get this utility to show both versions since his Quad swaps back and forth more often than most.

[colt1911]

Here' s mine on a Q9450 if it helps.

[msgclb]

Here are our Q6600 CPUs side by side. I feel your pain!

[FullSky]

The MSR tool doesn't seem to be working for me. I get no data for 0x39. Have no idea why. If I type in a different MSR I do see data.

Maybe it has something to do with my QX6700. Here's the results:



My QX6700 hasn't acted up today and I haven't been able to get the swap bug to occur, which is very strange considering how this cpu has always been before.

Carey

[FullSky]

I finally got the swap bug to occur again, but the MSR tool shows no change (i.e., no data). I do get data for other MSRs I type in.



Carey

[WoZZeR999]

MSR below

[zorzyk]

My MSR - Q6600/B3
RealTemp says that bug is fixed (core 1 and 2 have been exchanged) and shows 49-49-45-44,
CoreTemp shows at the same moment 49-45-49-44.

[unclewebb]

Thanks for all the data guys. I'm trying to find a reliable way to test for the Quad bug but it looks like MSR 0x39 isn't it.

My present test and fix is working on my Q6600 and it looks like it is working on your Q6600 zorzyk. I'll keep looking FullSky to see if I can find a better way to detect this.

[MassiveOverkill]

Q9450

[HDCHOPPER]

go unclewebb exellent drive for perfection thier wtg man

[skinnee]

mine...Q9450

[zorzyk]

Q6600/B3 information addendum: after restarting my PC (setting up new video driver) situation changed:
- RealTemp does not inform about fixing core swapping
- RealTem and CoreTemp both show temperatures like 49-49-45-44
- MSR shows exactly the same as in my post above (0 0 0 0).

[unclewebb]

zorzyk: Thanks for your testing help. Looks like your Q6600 is clearly showing this bug. I think I'm getting closer to a way to detect it for all CPUs so that your temp readings will always be consistent. Some processors that have this bug swap back and forth on a regular basis so one day it looks like core1 is running cool and then the next day its running hot with no explanation as to why. I believe that users have wasted a lot of time trying to fix this by remounting heatsinks, etc. but I think the problem lies a little deeper than that. I sent you a PM with some further tests you can do when this bug returns to your Q6600.

I added a couple of more buttons to my tool which makes it easier to search for MSR data.



http://www.fileden.com/files/2008/3/3/1794507/MSR.zip

I think it's easy to see that core0/core2 in my Q6600 are on one side of the Quad while core1/core3 data is coming from the other side of the Quad. When the bug is not there, core0/core1 will line up while core2/core3 will line up as well.

The MSRs I've found so far that clearly show this are 0x2A, 0x4D, 0x4E, 0x4F, 0xC7, 0x118 and 0x19E but I haven't found documentation from Intel on most of them. I initially overlooked 0x2A but it might be the one I use for this test. Once I get some more user feedback from people that have this bug then maybe I will be able to use one of these MSRs to reliably detect the Quad bug.

[zorzyk]

unclewebb - PM sent, please check

[unclewebb]

Thumbs up for zorzyk.

Your data is excellent, well laid out and easy to read. If the Quad bug ever returns for you try running my test again to see if there are any changes in your results.

[WoZZeR999]

I tried all MSR's listed above, and Core 0 and 1 are together, and Core 2 and 3 are together.

Edit: Super-Pi with core affinity follows suit with what the MSR's say.

[unclewebb]

Thanks WoZZeR999. That's what zorzyk found as well. When the Quad bug is present, all of the MSRs I listed show core0/core2 aligned as well as core1/core3 aligned. I think FullSky occasionally has a different version of this bug where core0/core4 and core1/core2 get aligned.

Whenever the cores get out of their normal alignment, the SetAffinity... in the Task Manager and every other program that uses the SetAffinity() subroutine in C++ will not function correctly. In my case, when I ask for the temps for core1 it is presently returning the temps for core2 and vice versa. This bug helps explain the 101 Quad screen shots I've seen during the last year or so that clearly shows the temps aligned in strange patterns. This bug needs to be corrected for and I think I have enough data now to better detect when this is happening. It could be another RealTemp exclusive feature.

[jeanjean15]

Hello .

I did not read all this big topic but someone could tell me if the différence of 10°C between RealTemp and CoreTemp with my QX9650 is normal or not ?

And which temperature is right ?

[zlojack]

Hello .

I did not read all this big topic but someone could tell me if the différence of 10°C between RealTemp and CoreTemp with my QX9650 is normal or not ?

And which temperature is right ?

Yes, it's normal.

RealTemp is right.

[FullSky]

uncleweb,

I've pm'd you some results for 0x2A.

[rge]

Unclewebb, last time i suck up your space with my temp testing....really But had to show one last experiment.

IHS temp is 95C at DTS=0 on multiple E8400's, measured by multiple people.

Experiment: I took another IHS (from old cpu), put a minimal haze of mx2 on my cpu, and placed a second IHS on my E8400. So now I am going to measure temp through two IHS, and two interfaces, and compare to temp on just IHS...to illustrate gradient through tim and IHS.

If 105 tjmax is correct for E8400, a 10C gradient must exists from die to IHS at idle, undervolted state. If an additional layer of tim and another IHS is added that gradient must increase significantly by up to another 10C.

If 95 tjmax is correct for E8400, less than 1C gradient exists from die to IHS at idle, undervolted state. If an additional layer of tim and another IHS is added then additional gradient would be minimal, 1-2C.

pic 1 is realtemp on my IHS, (without second IHS added yet) IR temp reads 71.8C, realtemp 72C. (coretemp with tjmax 105 reads 82C not pictured)

pic 2 second IHS

pic 3 second IHS on cpu

pic 4 is IR reading 65.8C through a second tim and IHS, and realtemp is reading 66, 67 on cores, which proves the gradient through IHS and tim at idle, underclocked, undervolted is max 1-2C . (taped down IR on button and IR gun down...pics easier with hand out of way)

Realtemp 95 tjmax is clearly correct. No way, no how is tjmax 105 because no way no how does a 10C gradient exist through tim and IHS, in fact it is clearly less than 1-2C.

[unclewebb]

Unclewebb, last time i suck up your space with my temp testing....really

What are you talking about? Testing is where all the fun is! The more data the better. Especially when it agrees with the theory behind RealTemp and with the Intel documentation you've read and been sharing with us for many pages now.

The data is in so now it's time to write a fix for the Quad bug issue.

[unclewebb]

The Quad Bug fix is ready for testing.
http://www.fileden.com/files/2008/3/...alTempBeta.zip

Now I just have to wait for FullSky and zorzyk to give this the or

So far it works on my confused Q6600. In the Test Sensors results window it will show you if your Quad processor was found to have this bug or not.

The normal order should be 0 - 1 - 2 - 3 but here's what it shows for my confused Quad:



Thanks for everyone's help with this new feature. You never know when this bug might hit your Quad. My Q6600 was fine for 3 months before it hit.