Real Temp - New temp program for Intel Core processors

[unclewebb]

Time for a tweak and a bug fix or two.
At least the GPU temps are looking good in these recent screen shots.

[loonym]

Thanks once again unc, outstanding!

[DGill]

With 2.90 rc8 only get 2 cores displayed of i7 920. 2.90 rc4 works fine, showing all 4 cores.

[shazza]

With 2.90 rc8 only get 2 cores displayed of i7 920. 2.90 rc4 works fine, showing all 4 cores.

Do you have Hyper Threading enabled or disabled? I had the same issue with HT disabled. Works fine now that I've enabled HT again.

I believe uncle is working on it.

[Wotcher]

Thanks Uncle ! - for updates & New features

just noticed ( compare RT v2.90 RC1 - to-> RT v2.90 RC8 @ the moment )

RC8 does not show ( chnanges in readouts ) when SpeedStep ( Eist / C1E )
is down-clocing CPU .. don't know was it it just designed so ?

Any way - thats no problem for my self ... just let You know .. if it makes sense ..

... snipped out image ...

Works OK for me.

... snipped out image ...

I'm in the same boat as i43 though I am wondering if it is because the Black Ops bios disables EIST when the CPU clock ratio unlock is enabled in the bios, I'll reboot to check and post back.



GPU temp display is awesome unclewebb, tyvm



EDIT:

That wasn't the problem it seems

[drminer]

Unclewebb,

I reset the latest beta to the default settings, and it still shows TJMax for the QX6800 B3 as 100C, instead of 85C, as in previous versions. It does correctly identify the CPU as QX6800, by the way. I have attached a screenprint of CPUZ and RealTemp as you requested.

Thanks!

[unclewebb]

Hang tight guys. Bug fixes are on the way. When you make too many changes at the same time, screwing up one or two or three things isn't unusual.

Maybe all the bugs were deliberate so people would come here and post some screen shots so I could see how many of you are enjoying the new GPU temps.

[Vatos_locos]

at me it cant recognize my 4850 so it gives me something like this :p

[unclewebb]

http://www.fileden.com/files/2008/3/...alTempBeta.zip

Maybe this version should be called, "the return of the missing cores." I'm hoping for that anyhow.

rge did a lot of testing for me today and noticed that his board assigns threads to cores in a very random order at times.



The theory is that this helps spread the work load to all of the cores. If RealTemp is properly finding the cores again, then it should organize them into the correct physical order.

The bug with C1E/EIST should be fixed and the code I added to handle the Q9100 processors that screwed up the detection code for the QX6800 has been fixed too.

The first person to post a screen shot of RealTemp working properly with hyper threading turned off with a Core i7 will win a prize.

Vatos_locos: Unfortunately, RealTemp only reads the Nvidia GPU temp sensor at the moment. I thought about buying a cheap Nvidia card today to save me the hassle of writing some ATI code. If I trip over some code for reading the ATI sensors, I'll probably add that to RealTemp in the future.

[shazza]

LOL ... just happened to check this thread before heading off to bed, but please ... give the prize to the second person so we can keep the excitement going

[unclewebb]

Thanks shazza for first bringing this problem to my attention and thanks for showing that RealTemp is working correctly again.

rge also sent me a PM with this info

2 cores active HT on and off gives 2 core temps.
3 cores active HT on and off gives 3 core temps.
4 cores active HT on and off gives 4 core temps.

and one of these.

If his APIC ID is still all mixed up tomorrow he might drop by and give an example of how some programs organize the temperature data based on the correct physical location and how some other programs, don't.

Edit: I'm still looking for a tester with an SLI setup. RealTemp is designed to report the highest GPU temperature if you have multiple GPUs so I'm hoping someone can post an example that shows if this new feature works or not. A comparison to a program like GPU-Z would be great.

[jcool]

I am using 2,86 I think and it's working fine on my i7. Lowest temp i7 can display is -8, so I got -8C on all cores
Unless I got enough Vcore and fire up linpack, then I can get into positive range - need to check the mount I think

[shazza]

Thanks again for this great application ... and for how quickly you fix things. Is the fact that threads are being assigned to different cores in a random manner also partially responsible for some of the temp differences between cores?

Here's a screenshot of GPU temps for my SLI setup at idle (GTX 260, 55nm) . Real Temp seems to be working fine.

[rge]

Unclewebb, had some more time to play with RC9 this am, Real temp is working perfect. It reads correctly whether 1,2,3 or 4 cores enabled and whether HT is on or off. Also everytime you reboot, even though the core order can be different, except core 0 always seems to be first, Realtemp still seems to report them in the correct order. Everest and speedfan do not.

3 pics, all with 3 cores enabled, HT off, just for example.
Everest gets confused with HT off and 3 cores enabled, etc, as it apparently starts reading just 1 core 3x or maybe 2, temps track exactly same no matter what, while RT still reads the 3 different cores correctly, and you can tell that although APIC (another unclewebbs program) is showing pic 2 has different order than pic 1 and 3, that Realtemp keeps cores in correct order all 3 pics.

[unclewebb]

Is the fact that threads are being assigned to different cores in a random manner also partially responsible for some of the temp differences between cores?

My opinion is that most of the differences in reported temperature between cores is slight differences in TJMax from one core to the next. The 100C TJ spec is a TJ Target and Intel has said that they don't always hit that target exactly when setting TJMax. It can and does vary by a few degrees from one CPU to the next as well as one core to the next. The TJMax spread for the 45nm Core 2 series is much greater than all of the other chips, combined!

There is also a slight amount of slope error in all of these sensors where a core can change temperature at a slightly different rate than what the sensor indicates. Intel says this can amount to +/- 10%.

A test I used to do with a Q6600 was to run 1 thread of Prime95 Small FFTs and then move it around from one core to the next using the Task Manager - Set Affinity... function. By doing that, it was easier to see and learn about the physical order of the cores. I also learned that Task Manager is bugged too and if your APIC ID is random, then when you ask Task Manager to run a task on a specific core, you'll likely end up with it being run somewhere else.

With the introduction of the 45nm Core 2 series, everything got closer together so it was hard to see the cores based on temperature changes. For Core i7 you might be able to run 2 threads on the same core and move the pair of them around with Set Affinity...

I followed the directions for SLI so I'm glad to see that it works. It should also handle 3 or 4 way SLI and report the highest GPU temperature. I figure if you're going to use RT to shutdown your computer based on GPU temperature, then that has to be based on the hottest GPU.

rge: Your pictures speak volumes. I know that your core0 always reports as your hottest core so it sure looks like Everest is copying that one value and repeating it 3 times. The Intel APIC documentation gives new meaning to the word confusing so it's easy to understand the misunderstandings about core order.

Anyone interested can download my APIC ID tool that I wrote for rge.
http://www.fileden.com/files/2008/3/...507/APICID.zip

When you exit RealTemp, it writes the APIC ID to the INI file so you can also have a look in there for APIC information. The core order remains constant until the next re-boot.

[Wotcher]

Speedstep display fixed in RC9, thanks unclewebb, you sure are efficient.

[unclewebb]

Speedstep display fixed in RC9, thanks unclewebb, you sure are efficient.

You're welcome.

During SpeedStep transitions, RealTemp might report slightly differently than CPU-Z, etc. Anyone that has EIST enabled but can't get RealTemp to show (6.0 / Core 2) or (12.0 / Core i7) at idle, needs to check some of their Windows settings in the Power Options.

With the correct settings, both programs should report the same at idle and full load but there might be times when RealTemp will show you an average multiplier which more accurately reflects the truth. In some situations, the multiplier can be going up and down rapidly so some software will just ignore the changes and report a steady value.

[DGill]

Thanks Unclewebb - latest version working great with i7-nice to have all cores showing again.

[rge]

Unclewebb, next time your tinkering in Realtemp, one more update,
finally got chance to test for absolute max with 4C outside temps and 1.68vcore...had posted in OCing thread
http://valid.canardpc.com/show_oc.php?id=481012

[unclewebb]

rge: I can tell from your GPU temp that you were getting serious about running an XS Bench number that you can be proud of!

I think it's time to make 2.90 Final. Maybe one more small tweak.
I'll definitely include your new numbers to give people something to shoot for.

DGill: The information rge sent me about APIC ID helped me to finally understand Core i7 and how threads are assigned to each core.
You could say, the light finally went on when I saw his numbers.

The Intel documentation makes something that is incredibly easy, almost impossible to understand.

[Vatos_locos]

i think i have a problem :p with the RC9 2.90 version
it cant remember the setings so when i close real temp i have to set them again
like i want to see all 4 cores temp in the tray and to start it minimizes
is something that i do wrong :p ? with older version it could remember the settings

[unclewebb]

i think i have a problem :p with the RC9 2.90 version it cant remember the setings so when i close real temp i have to set them again

Does anyone else have this problem because it works for me?

The only thing I can think of is that the RealTemp folder is located in a directory that you don't have the privilege of writing to. If this happens it wouldn't let you save your settings in the INI file. You can also right click on the RealTemp.ini file and do a Properties on it which might show you what's going on.

If you have that file open in a separate program while RealTemp is running then that might also block you from saving your settings. I'll try a few different things to see if I can find a problem in something I did.

Edit: Make sure you unzip the file you download. RealTemp will start from a zipped file but it might have a problem saving the settings back into a zipped folder.

[burebista]

It's fine for me.
It remember when I was at bathroom, when I opened my computer, when I update RealTemp, it's scary sometimes.

[Vatos_locos]

i have it on C:\real temp ( C is the windows drive )
i'll try to move it to an other drive to see if the same thing happens

EDIT
i just move it to the C:\Program Files x86\Real Temp
and it is working now

so i guess it a problem with my pc not with the program sorry

[unclewebb]

Thanks for testing that out Vatos_locos. I always get worried when I hear about a bug!

Windows Vista and XP can get kind of picky when you try to write to directories that you don't have the appropriate privilege to write to.

On a different note. Anyone wanting to test out the public beta version of Windows 7 should head here and get in line before Microsoft closes the door:
http://technet.microsoft.com/en-us/e.../dd353205.aspx

A 2.44GB download might tie up your internet connection for a while though.