RealTemp General Discussion

[unclewebb]

Tau: I think your Q9450 might have a couple of stuck sensors. Your cores are numbered 0, 1, 2, 3.

Core 2 doesn't change at all when the CPU load goes from 31.2% to Idle. It's likely that sensor is not capable of reading any value higher than a Distance to TJMax of 63. Core 1 moves 4.4 degrees over that same range. Core 0 only moves 0.6 degrees which is also a sign that core 0 might be getting stuck at a Distance to TJMax of 64. Before doing any calibration I would try to run your CPU at a cooler temperature so you can determine exactly what the sticking points are.

All of these sensors stick. When you get lucky, you hope that the sticking points are below your normal range of operation.

Between core 2 and core 3 I'm seeing an approximately 8C difference for a lot of the values. I think TJMax of these two is probably about 6C different with the rest of the difference being more slope error related.

Once your paste has had time to set up and you start to do some overclocking, let me know and then we can come up with a few more tests at different settings to do a calibration for you. It will be interesting to compare my suggestion to the calibration you're presently using.

 

[Tau]

Tau: I think your Q9450 might have a couple of stuck sensors. Your cores are numbered 0, 1, 2, 3.

Core 2 doesn't change at all when the CPU load goes from 31.2% to Idle. It's likely that sensor is not capable of reading any value higher than a Distance to TJMax of 63. Core 1 moves 4.4 degrees over that same range. Core 0 only moves 0.6 degrees which is also a sign that core 0 might be getting stuck at a Distance to TJMax of 64. Before doing any calibration I would try to run your CPU at a cooler temperature so you can determine exactly what the sticking points are.

All of these sensors stick. When you get lucky, you hope that the sticking points are below your normal range of operation.

Between core 2 and core 3 I'm seeing an approximately 8C difference for a lot of the values. I think TJMax of these two is probably about 6C different with the rest of the difference being more slope error related.

Once your paste has had time to set up and you start to do some overclocking, let me know and then we can come up with a few more tests at different settings to do a calibration for you. It will be interesting to compare my suggestion to the calibration you're presently using.

Thanks for the insight, i decided to seal up the case overnight and let it run at load, came down this morning and it was loading at ~55*C Thats much to hot for my liking, looks like I will be adjusting the cooler/fans here sooner than i anticipated.

 

[burebista]

[...] it was loading at ~55*C Thats much to hot for my liking[...]

55°C too much in load?
Peoples care too much about temperatures.

 

[Tau]

55°C too much in load?
Peoples care too much about temperatures.

Yeah, Im a performance nazi though, i probobly wont stop untill I have this thing running at 3.8Ghz+.... Meh ill see what i end up at, if all else fails ill water+pelt this sucker. Should be a cake walk for a 400Watt pelt and some water

 

[unclewebb]

55C isn't bad at all but what I learned about Core chips is that the cooler you run them, the more MHz you can reliably overclock them too. Less heat = more MHz. That's a nice simple formula to remember.

Personally, I usually sacrifice some performance and let my chips run a little hot. Burebista and I prefer computers that are nice and quiet.

 

[Tau]

55C isn't bad at all but what I learned about Core chips is that the cooler you run them, the more MHz you can reliably overclock them too. Less heat = more MHz. That's a nice simple formula to remember.

Personally, I usually sacrifice some performance and let my chips run a little hot. Burebista and I prefer computers that are nice and quiet.

Yeah, I think that i am a little over concerned, once i fix up the airflow situation i should probobly see a decrease, will just have to see were they end up wile overclocking

 

[maf]

Hi unclewebb, first I'd like to say thanks for Realtemp, I've been using the various versions for some time now.

(I've read most of the 126 page thread at xtremesystems, and was going to join there to ask you a question, but my email address didn't seem to work and I couldn't get the confirmation email. Long story etc, but I decided to give up on that idea and ask you a question here.)

I have an M0 revision E2140 and was glad when you changed the default tjmax for that chip from 85C to 90C, as it tallied well with my own findings; previously I had been using programs where I could define my own nominal value for tjmax. Version 2.88 is the one I have with 90C tjmax, and 2.6 I think with 85C. However when I downloaded the latest version, 2.9 RC12, the tjmax has gone up to 100C.

I believe you conducted a test with IR thermometer on an E2160 and heated the lid to 85C when running without a heatsink. I also believe you had concluded that the gradient from core to IHS was about 5C with no heatsink (obviously it is more when a heatsink is installed). I couldn't see any comments in your thread about the change to 100C for E2xxx chips and wondered if you could explain it? I also wondered if it may just be an oversight because you have been concentrating on support for i7 and wolfdale recently?

 

[unclewebb]

Sounds like I screwed up. Sometimes you make a minor change and don't realize that it causes more changes than what you thought. I'll have a look at the code to see what's changed.

 

[Flash]

Hi unclewebb.

I just tried to check Task Bar or Tray Info, but even tho they are checked if you press the X button the program just closes and doesn't go to Tray bar. Is this how it's suppose to be? If yes, then what is the purpose of Tray and Task bar?

Also, do you think there will be an option to startup with windows?

 

[burebista]

Put an entry in RealTemp.ini
MinimizeOnClose=1

And about startup with Windows in XP you should drag exe on Startup folder and in Vista made a Task Scheduler entry.

 

[unclewebb]

The next version of RealTemp has a Minimize on Close option in the Settings window but until I release that, burebista's way works fine.

Microsoft designed the Task Scheduler in Vista and is using it in Windows 7 because it is the best and most secure way to manage programs that start up with Windows.

The Startup folder is also available in all operating systems. I'm old school and don't like adding or removing any items from a person's registry so I make users do things manually. It's not too hard to drag a link to RealTemp into your startup folder.

In Vista, click on the blue Windows logo at the bottom left and type in Start Menu and open up the folder that appears. Open up Programs -> Startup and drag a shortcut to RealTemp in there. In XP it's even easier to find this folder by clicking on the Start button and having a look in your programs menu item.

This way, no one can ever accuse me of screwing up their registry.

Edit: It's finally done.

http://www.techpowerup.com/realtemp/

 

[troll334]

R.T. v3.0 and Q8200

Greetings All,
Just received a Dell Inspiron 530 quad core. New system.
Out of curiosity, I downloaded and ran "Real Temp v3.0".
Interestingly enough, the temps for the first
three cores show activity. The fourth core, however, displays
a constant temp even though I see work performed on it. In fact,
McAfee was almost exclusively using the fourth core during
a system scan so I thought perhaps Real Temp just couldn't
get in there for a reading. When McAfee completed and went
back to sleep, other apps started using the fourth core and
the temp was still static.
Any possibility the sensor for that core is bad?
I've left a post on one of Dell's User forums and thought I'd see
if you folks have any insight.
Thanks!
mike

 

[burebista]

It looks like a stuck sensor. Do a CPU Cool Down Test and post a screenshot here.

 

[unclewebb]

Some sensors on 45nm Core chips don't work at all below 60C but work fine above that. They were only designed and calibrated by Intel to trigger thermal throttling and thermal shutdown.

You could try running a high stress application like Prime 95 Small FFTs on all 4 cores. The way these sensors work is that when the temperature increases higher than the sticking point, the sensor should start moving. Depending on where your sticking point is, you might be able to see this sensor move at full load. It's just a limitation of the sensors that Intel uses. Not all sensors are suitable to also be used to report core temperatures.

 

[troll334]

Most Appreciated

Gents,
Thanks for your input and teachings. I appreciate it.
Went home during lunch and kicked off Prime95 then
ran to cool-down operation. I did not, however, run
seperate instances so the load was spread across all
four cores. The temp never did move. I'll rerun using
four instances and see what happens. In any event,
it does not appear that I need to worry or have Dell
ship out a new proc. Not that they'd do it anyway
I was really just concerned that the fourth core could
get into thermal runaway w/o being detected. Time
will tell eh?
Thanks again and I won't bug you further.
have a great weekend.
mike

 

[Rift]

Hey all I'm fairly new here iv just got a pretty simple question in regards to real temp.


I noticed that Intel released the TJ Max values for most of there procs. I also saw that in the release notes for Real Temp 3.0 it sais these values were updated but when I launch real temp it still shows the TJ Max for my q6600 to be 100 degrees.

According to Intel's release data http://www.tomshardware.co.uk/intel-dts-specs,news-29460.html

It's supposed to be 90 degrees if I'm not mistaken. Is this still something we have to set on our own or by updating the values was the software supposed to change it on its own? It's no biggy to change it on my own just want to make sure my numbers are right.

Rift

 

[Hayder_Master]

this version is great and best temp read imprassive work thanx

 

[unclewebb]

Good question Rift and I'm glad you brought this up.

In my opinion, Intel played a bit of a public relations game last year. The user community has been hounding them since Core processors first came out to release some information about TJMax so they finally decided to cave in, sort of.

If you look closely at their releases last year at IDF you will see that they have changed the wording. Instead of TJMax they used the term TJ Target. TJ Target is their goal and actual TJMax might be higher. They didn't document how much higher actual TJMax might be so basically their partial release of information has turned out to be mostly useless.

The other problem is that TJMax is not a fixed number like everyone has always thought that it was. Not only is it not consistent from one Q6600 G0 to another Q6600 G0, it's not even consistent from one core to the next on the same CPU. The 45nm Quads make this very obvious where the difference from core to core can approach 10C. That's just the nature of the sensors that Intel is using. Intel says it is due to production line tolerances but the problem is simply that these sensors were never designed or calibrated to be used for accurate core temperature reporting. The purpose of these sensors is to control thermal throttling and thermal shutdown and for that purpose, they work excellent. Any usability beyond that is at the users discretion.

Some people will come to the conclusion that with all of these issues, core temperature monitoring software is pointless. What I found during testing is that if you take the time to calibrate RealTemp to your CPU then, if your sensors are not sticking, you can end up with some reasonably accurate core temperatures. The sensors are not capable of 100% accurate temperature readings from idle to TJMax no matter what software you're using. All you can do is try and calibrate them to minimize the amount of error.

The calibration procedure involves running your CPU at idle at a fixed MHz and core voltage and then comparing your reported temperatures to your room temperature near your computer with the case open. The Calibration section of the documentation has more info on this:

http://www.techpowerup.com/realtemp/docs.php

This calibrates the low end and even if your actual TJMax is out by a few degrees, your reported temperatures from idle to full load are going to be fairly accurate.

I tried setting RealTemp up to use TJMax = 90C like Intel recommends and then I pointed an IR thermometer at the IHS that covers the four cores. When RealTemp reported 80C, the IHS was showing 85C as measured with the IR thermometer. It's impossible for the heat spreader covering the cores to be hotter than the source of the heat. If I use TJMax = 95C then these two values are equal but even that doesn't seem right. In order for the cores to maintain an IHS temperature of 85C, they are going to have to be hotter than the IHS surface temperature.

rge at XtremeSystems mounted a calibrated thermocouple into his IHS and did some testing and came to the conclusion that the cores would likely need to be about 5C hotter than the measured IHS temperature. That sounds reasonable to me which is why for the Q6600 G0, I have set TJMax = 100C. The authors of Core Temp and Everest have also decided to mostly ignore Intel's TJ Target numbers. There are too many 65nm processors where Intel's TJ Target spec has very little to do with the actual TJMax.

Now if someone calls Intel and says, "What is TJMax", they can tell them that information has been released and if software developers like myself have chosen not to follow those guidelines, then it's obvious that we don't know what we're doing.

If you can, run a CPU Cool Down Test and post your results. It might show another theory that I have. For many Q6600 Quads, TJMax for core2 and core3 seems to be set 5C higher than core0 and core1. The Cool Down Test will help show if your CPU follows this trend.

 

[Rift]

Unclewebb so I had some time to do the cooldown test that you linked to and I have some various results to post.

Here are the results with no modifications in RealTemp

This one is with TJmax set at 90 instead of the default 100C,

Mind you at this point I have not messed with any of the offsets because I'm still not totally sure by how much I would need to offset it and if I even need to make offset adjustments.

Then I ran the test with my OC settings at RealTemps default settings.

Like before I also ran the test with TJmax at 90C instead of 100C.

The current temperature around my case is 24C and when place about 2 feet away the temperature is 24.4C. When placed on the other side of the room the temperature hits 25C.

So these are my results. As stated above I'm still not sure about the offsets should I be adding the stated temperature offsets for High end air (true push/pull) ........6-7C. I believe my system fall under this, as I have 2 120mm Intake fans at the front of the case with 2 120mm exhaust fans at the back of the case. The CPU is being cooled by a Coolermaster V8 which has a 120mm Fan sandwiched in the middle of it.

Rift

 

[unclewebb]

If your air temperature is 24C and there is typically a 7C gradient between your air temperature near your cooler and your core temperature then I would guess that your actual core temperature is about 31C in your first screen shot. If your case is closed then even with good air flow that number might be a degree or two higher. With a GPU temp of 60C, it's easy to create a degree or two of extra heat beside your CPU heatsink. You'll have to be the judge of what you think is reasonable.

I wouldn't use TJMax = 90C. It just doesn't seem likely from what you've shown me. In my opinion, that is just a number off the top of someone's head at Intel and doesn't have anything to do with the actual TJMax for all of the Q6600 - G0 processors that I've seen so far. Other programmers seem to agree with me on that one. I don't think any software is using TJMax = 90C for a Q6600 G0, not even Intel TAT.

When you are overclocking and you take your CPU up to 80C you will notice a 2C difference in your maximum temperature between your two sets of cores. There are two theories here. Either the actual TJMax for core 2 and core 3 is 2C higher than core 0 and core 1 or perhaps those two cores are getting slightly less voltage at full load so they naturally run a couple of degrees cooler. Typically from 80C to 100C, that 2C difference will remain very consistent.

If you believe in my theory that TJMax might be a couple of degrees higher on those two cores then try running TJMax = 100, 100, 102, 102 or TJMax = 98, 98, 100, 100

The reason I recommend the second option is because that will lower your idle temperature on core0 and core1 down to 32, 32 which is probably very close to the actual temperature. These sensors are far from perfect so I'd do this and call it a day for those 2 cores.

You would then only need to adjust the idle calibration factors for core 2 and core 3. I start by trying to equalize the reported core temperatures at low MHz and low voltage across all 4 cores. After you go back to your overclocked settings, you might find you need to make some very small adjustments to these calibration factors so your reported core temperatures remain balanced at idle.

Plan B would be to say to yourself that your temperatures are close enough to reality and decide not to bother with any calibration. A lot of users choose that option. Your Quad is better than most 65nm Quads and is worlds better than any 45nm Quad I've seen.

If you decide to calibrate, I usually run Prime95 Small FFTs from idle to full load and keep an eye on how the 4 cores track each other. There might be a slight amount of lag but generally they track each other very closely on the way up and when returning back to your idle temperature. I set the logging feature to a 1 second interval and then I do a few starts and stops of Small FFTs and go back and see how it looks. Let me know if you agree / disagree and what you discover when testing.

 

[Nitro-Max]

Is this program more accurate then core temp i ask because i get differnt readings with this program?

Both coretemp and real temp report the exact same temps for each core on my Q6600 theres no differnce to me.

 

[Israar]

Here's mine, any thoughts on it guys?

--Lee

 

[unclewebb]

Your screen shot is an excellent example of why Intel does not recommend using these sensors to report accurate temperatures. In this case I agree.

Core 0 and core 1 are stuck and useless and core 3 looks like it reads way too low. Maybe core 2 might be usable if you calibrate it. Read the documentation here on TechPowerUp and check out this post by rge at XtremeSystems for more details:

http://www.xtremesystems.org/forums/showpost.php?p=3384504&postcount=2429

http://www.techpowerup.com/realtemp/docs.php

 

[Israar]

Your screen shot is an excellent example of why Intel does not recommend using these sensors to report accurate temperatures. In this case I agree.

Core 0 and core 1 are stuck and useless and core 3 looks like it reads way too low. Maybe core 2 might be usable if you calibrate it. Read the documentation here on TechPowerUp and check out this post by rge at XtremeSystems for more details:

http://www.xtremesystems.org/forums/showpost.php?p=3384504&postcount=2429

http://www.techpowerup.com/realtemp/docs.php

Thanks fella much appreciated

Will read them both tonight when I get home from work, so for now I'll put them in my favourites =)

--Lee

 

[JBravo]

Unclewebb

Thanks very much for a great app, its the only one I use to monitor temps!

I've read the calibration part of the documention, but I just want to run this by you.

I've got a E7200 45nm. I'm going to calibrate realtemp as you said by getting the ambient down to 20C thanks to an aircon,drop the speeds as low as possible with 1.1V and add 11 or 12C to get the value that I should be seeing? I'm on stock intel cooling.

Would that be fairly accurate? Coz this would be the "minimum" it would ever go right?

I would appreciate your help.

Again, thanks very much