Real Temp - New temp program for Intel Core processors

[davemelb]

Hi Guys,

If my sensor test reports 2,1,6,9. What value should I use in the calibration? 0, 0, -1, -2?

[unclewebb]

Try reading the section on Calibration in the documentation for some background info:
http://www.techpowerup.com/realtemp/docs.php

The best way to calibrate is to drop your processor down to a fixed value like 6x266MHz for 65nm or 6x333MHz for 45nm and then drop the core voltage to about 1.10 volts. Open your case and let us know a few details like what CPU you're using, what your air temperature is or water temperature if you are water cooled. A typical Quad will report temperatures about 6C above your air or water temperature during this test. Make sure your CPU is as idle as possible and that it doesn't have any background processes running that will effect your results. This is not an exact science but should get you in the ball park.

By the looks of things, I'm going to guess that you have a 45nm Quad and core0 and core1 might be very close to the sticking point. Intel says these sensors can "bottom out" at any temperature below 50C so you have to be careful when trying to calibrate. Sticking sensors can not be properly calibrated with my approach but I have a plan B if that happens. 45nm sensors can be a little on the wild and crazy side.

I was helping someone yesterday running a dual boot system with Windows 2000 and Windows XP. His load temps were the same but his idle temps were about 10C lower in XP than in 2000. I think it's because Windows 2000 wasn't letting his Core 2 Duo go into C1 at idle so it was running hotter. Anyone else heard of this before?

[davemelb]

Hi, thanx for that, my setup is:

Q9400 + air cooling
6x333 + 1.0v set in BIOS, other settings to "Auto".
ambient temp according to my thermometer = 18c (placed just outside the side case window)

I did four sensor tests and results below:

1) 0, 0, 2, 2
2) 0, 0, 1, 1
3) 0, 0, 1, 1
4) 0, 0, 2, 1

RealTemp reports core temperature as: 26, 31, 29-31, 29-31. The last two fluctuates between 29 - 31 degrees, the first two doesn't move from 26 and 31.

How should I calibrate it from here? Thanx

[unclewebb]

It looks like core0 and core1 are probably stuck. You need to have your case open and all fans as high as possible for the best accuracy during this test.

What type of CPU cooler are you using?

Intel now specifies that TjMax=100C for your Q9400. Are you using that? The latest beta of RealTemp is available here if you need to upgrade:

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

Because you have two cores that seem to be sticking it would be a good idea to run an additional test so I can see how these sensors react between idle and full load. The test is to turn the logging feature in RealTemp on and to set the interval to 1 second. Let your computer sit idle for about a minute to give me some background info and then run Prime95 small FFTs for two minutes and then back to idle for a minute. Stop RealTemp after that and send me the RealTempLog.txt file so I can have a look. You can e-mail it to me at the address in the About... box of RealTemp or PM it to me at XS. Run this test at default MHz (333x8) and core voltage. This will let me see how the sensors in core0 and core1 compare to the other two.

[rgrea]

Nothing will work with realtemp in Vista 64 and won't work on the 175 opteron AMD board either.
I've tried but there is no admisintrator account only me, in Vista. Could someone explain that part of Realtemp please.
thanks
bob

[WoodButcher]

Nothing will work with realtemp in Vista 64 and won't work on the 175 opteron AMD board either.
I've tried but there is no admisintrator account only me, in Vista. Could someone explain that part of Realtemp please.
thanks
bob

RT only works w/ Intel.

[unclewebb]

rgrea: RealTemp runs fine on Vista 64 on your Intel E8400 but AMD CPUs are not supported. If you are in a limited account you will need to right click on the RealTemp icon and run as Administrator to get the WinRing0 driver to load. RealTemp needs that driver to read the sensor in an OS friendly way.

[Brama]

rgrea: RealTemp runs fine on Vista 64 on your Intel E8400 but AMD CPUs are not supported. If you are in a limited account you will need to right click on the RealTemp icon and run as Administrator to get the WinRing0 driver to load. RealTemp needs that driver to read the sensor in an OS friendly way.

Is it not possible, like other programs that I suppose use ring0 drivers (RivaTuner, cpu-z, Everest, SppedFan, etc.) to launch the program under Vista normally and after confirm that it is ok to use the program?

As you know, when actually you launch RTemp under Vista, you get a small windows that informs you to run the program as Administrator. So you have to right click, chose "run as Administrator" and finally the program starts.

It would be simpler to launch the program with the classic double-click, answer yes when Vista is asking you if you want to start the selected program and enjoy it, like the above programs do.

Thanks,

[Brama]

rgrea: RealTemp runs fine on Vista 64 on your Intel E8400 but AMD CPUs are not supported. If you are in a limited account you will need to right click on the RealTemp icon and run as Administrator to get the WinRing0 driver to load.

This happens also if you have an Administrator account under Vista.
Vista uses a kind of "Super Admin" rights to perform critical operations, like loading drivers into ring0.

[burebista]

I'm on Vista64 SP1, administrator account and RealTemp starts fine, minimized in tray (at last ).

It would be simpler to launch the program with the classic double-click, answer yes when Vista is asking you if you want to start the selected program and enjoy it, like the above programs do.

Vista don't ask me nothing.

[unclewebb]

As you know, when actually you launch RTemp under Vista

Ummm, I hate to admit this but up until yesterday I had never run RealTemp on Vista.
Well, maybe once, but that's the extent of RealTemp Vista development so far.

I finally have access to Vista for the first time today and I'm starting to learn about some of the issues that people have been complaining about for a long time.

You can run RealTemp with a double click from an icon on the desktop as long as you are an Administrator and UAC is disabled. I know there are users that think UAC is wonderful but as a user with a brain, I had it turned off in well under 24 hours. It would have been one hour but I had to fight with Vista for a while to get it to find my wireless router.

It's silly. When I double click on an icon the last thing I need is for my nanny to ask me if I really do want to run the program that I'm trying to open. Of course I want to run it or else I wouldn't be double clicking on it! The average user is not going to waste time reading the infinite warning messages so they'll just get in the habit of clicking OK or Yes to everything. My Vista rant for the day.

If you don't want to turn off UAC then you can add RealTemp to the Vista start up using the Task Scheduler where you can assign Administrator rights to RealTemp. This works without annoying messages or having to confirm anything. Maybe I'll add a Minimize on Close option so once RealTemp has started you won't accidentally close it.

Other programs use a signed ring0 driver to get around some of the Vista issues. I'm not smart enough and didn't want to waste my time learning how to write a driver so I'm using the WinRing0 open source driver. It works, didn't cost me a nickle and saved me lots of time. I'll probably continue to use it.

I've been working lately on some Vista64 Start Minimized issues with the infinitely patient burebista. Now that I can see RealTemp, which works OK in XP, not always Start Minimized properly in Vista, maybe I can finally fix this issue.

[burebista]

[...]not always Start Minimized properly in Vista[...]

Now it always starts minimized, only Task Bar is stubborn.

[Brama]

You can run RealTemp with a double click from an icon on the desktop as long as you are an Administrator and UAC is disabled. I know there are users that think UAC is wonderful but as a user with a brain, I had it turned off in well under 24 hours. It would have been one hour but I had to fight with Vista for a while to get it to find my wireless router.

If you had a teenager son you would keep not only UAC on.......

[unclewebb]

If you had a teenager son you would keep not only UAC on.......

I'm still learning about this UAC stuff. One option is to go into the Control Panel -> User Accounts and turn it on for each user of your computer that you don't trust and then you can turn UAC off for your own account.

From a Run... box you can also type in secpol.msc to run the Local Security Policy dialog.

Click on Local Policies -> Security Options and go down to:

User Account Control: Behavior of the elevation prompt for administrators in Admin Approval Mode

and change that to

Elevate without prompting.

This will keep the nanny on for IE7 but get rid of the tiresome nag messages when you try to run a program.

[Brama]

User Account Control: Behavior of the elevation prompt for administrators in Admin Approval Mode

and change that to

Elevate without prompting.

This will keep the nanny on for IE7 but get rid of the tiresome nag messages when you try to run a program.

Thanks for suggestion!

[Captn]

Hello Uncle web,

I just installed my NEW E8400 E0 stepping, under settings on Real Temp 2.79.1 my TJMax is set to 90, should that be set to 100 for this E8400? I used this on my cpu http://www.newegg.com/Product/Produc...82E16835100012 and my idle temps with my voltage set to auto atm are 32/30 My multiplier set to x9 blah blah....hehe. Are those ok temps and also should i set my TJMax to 100? Thank you inadvance.


Does anyone know what i should set the stock voltage at?

[drminer]

Unclewebb,

I have a QX6800, B3 revision. I noticed in your latest beta that the Tjmax for this CPU has been set to 95C. Is that right? I was expecting that it would be 85C, since it shares the same TcaseMax as the QX9770 (i.e., 55C). Thanks!

[Captn]

Ok im not getting this, lol. I have my CPU Voltage set to 1.18750 and my idle temps are still 42/40 and running Orthos my temps go to 43/43 thats about it after 1 hour of running orthos. Anyone have any ideas? Or is this normal temps?


Oh i forgot to mention that i set my TJMax to 100. 42/40 @ idle bad? or should it be lower. I have a Zalman 9700 on my E8400.

[rge]

Unclewebb,

I have a QX6800, B3 revision. I noticed in your latest beta that the Tjmax for this CPU has been set to 95C. Is that right? I was expecting that it would be 85C, since it shares the same TcaseMax as the QX9770 (i.e., 55C). Thanks!

I think that is why intel did not release 65nm...to many steppings and tjmax's to keep up with. I made a chart a few posts ago and only listed GO stepping for QX6800, but you are right the B3 stepping makes much more sense to be 85C, unlike the likely 95C of GO version. Would be nice to confirm some of this with IR measurements, we need to send Unclewebb one of everything to test.

I added the B3 stepping into my earlier post...does seem to fit with 85C.
http://www.xtremesystems.org/forums/...postcount=2152

[Captn]

Well according to the newest Everest 4.6......my cpu is idle at 33 and Core 1 is at 42 & Core 2 is at 40. Under 100% load from orthos my cpu is at 41 Core 1 is at 42 & Core 2 is at 40...LOL


I forgot to mention this is at 1.1500 volts

[Captn]

Anyone?

[Captn]

Unclewebb,

I tested my temp sensor with Real Temp v2.79.1 & its telling my temp sensor is stuck, I thought Intel fixed this with the E0's ? Right now im running @ 3.6Ghz 1.20000 volts for 1 hour so far with Orthos. My idle temps stay the same in all programs Core Temp, Real Temp, Everest 4.60. 42/40 @ stock speeds, when i go to 3.6 @ 1.20000 volts Core 2 will go to 42c. Everest tell me im at Core 1 42 & Core 2 41 and it also says my cpu temp is 43 @ 3.6ghz @ 1.20000 volts. Should I RMA this cpu or just live with it?

Thank you in advance.

[unclewebb]

Vista Start Minimized with or without the Task Bar option enabled seems to finally be working 100%. When I'm more awake tomorrow I'll upload a copy to the beta section for some further testing.

drminer: Now I'm confused. As you know, Intel didn't have too much to say about TjMax for the older 65nm processors. For most of the main stream processors, 45nm or 65nm, I'm pretty confident, but the Extreme processors in 45nm are a little different and that's probably also true for the 65nm QX processors.

Here's some of my recent logic behind the TjMax guessing game. Intel states TjMax=100C for the E8400. Based on IR testing of an E8400, that results in an approximately 5C delta between IHS temperature and reported core temperature. I think that amount of difference near the Intel calibration point of about 80C to 90C should be very similar for most Core desktop processors. Based on that, to get an early B2 to report 5C over the measured IHS temperature I need to be using TjMax=90C.

The early assumption was always that the original B2 was TjMax=85C and assuming that is where I went wrong. Based on what Intel said about TjMax for the 45nm, I'm going to start assuming that TjMax=90C for the B2. These processors tended to have a Thermal Spec of about 60C.

Your QX6800 - B3 is listed as:
http://processorfinder.intel.com/det...px?sSpec=SL9UK

which has a Thermal Spec of 54.8C so you're right, I think I should be using TjMax=85C for these. That seems to be the most likely number.

The confusing part was that I originally checked the Qx6700 - B3 and found this:
http://processorfinder.intel.com/det...px?sSpec=SL9UL

which has a Thermal Spec of 65C. It almost seems that one of these numbers might be wrong. The data entry person at Intel that makes up these datasheets, etc. occasionally makes a mistake sort of like how all the new E8x00 E0 Dual Core chips are all listed with the wrong CPUID in their documentation. The QX6700 and QX6800 B3 have almost identical specs so do they have the same TjMax or a different TjMax?

rge and I have found that there is usually a strong correlation between the Intel published Thermal Spec number and TjMax but one of these processors seems to be an exception to this rule or else it might just be a data entry error.

Upon further review, I think the B3 series of QX processors should both be TjMax=85C. Have you ever tried doing my low volts / low MHz calibration test to see how your idle core temps compare to your room temperature near your computer? A few Quad owners doing that test might help to prove this.

The QX6800 - G0 has a Thermal Spec about 5C more than the B2 so 90C+5C=95C and it is about 5C less than all of the other non-extreme G0 which are 100C so 100C-5C also equals 95C. That's also similar to some of the 45nm QX processors which Intel has stated are TjMax=95C. This is the kind of game I'm left playing because Intel couldn't tell the world about TjMax for the 65nm processors.

Thanks for bringing your QX to my attention. I'll shuffle the TjMax deck one more time to reflect this logic.

Captn: Since version 2.75, an E8400 - E0 should be automatically set to TjMax=100C. Can you open up the Settings window and click on the Defaults button to see what TjMax gets set to. If you used RealTemp with a different processor then it's possible that TjMax=90C was left over in the INI file from that. I know that my E8400 - C0 Defaults to TjMax=100C so try that test out and let me know what happens.

Stuck sensors are a fact of life for 45nm. Intel has stated that at about 50C, some of these sensors can become saturated as the processor temperature continues to decrease. That's a fancy way to say they get stuck. Try my calibration test from the RealTemp documentation and give me some more info like your room temperature, etc. and I'll try to take a semi-educated guess at your official sticking point.

No temperature sensors are 100% accurate or can be fully trusted including what is reported as the single CPU sensor reading by Everest and SpeedFan. That sensor on my board at low temperatures reads at least 7C too high in my opinion.

One thing I've learned about the core temperature game is to assume as little as possible so avoid making any comparisons to other questionable data that may be just as inaccurate as your core temps.

its telling my temp sensor is stuck, I thought Intel fixed this with the E0's ?

I think that rumor got started so users could justify to their spouses why they needed a new E0 processor. The Intel master plan seems to be for them to start spending a few more pennies for better sensors when Core i7 is released. That upgrade is going to be much more expensive overall so we're all going to need a compelling reason to upgrade. Marketing department probably told them, "Let's save the good temp sensors for then." A tiny bone for the enthusiast community. More like a milk bone if you ask me.

[drminer]

Unclewebb,
I would like to be a test case for you, but I can't. I have a cheap water cooling system that does not indicate water temp.

I would like to share a thought I have had about the vexed issue of temps. It occurred to me to combine some insights from Computronix and from rge.

TCase = TJunction - 5C (Computronix posits a 5C delta at full load)

TCaseMax = TJMax - 30C (rge has posited this relationship in a previous post)

So, TCase < TCaseMax

is achieved if

Distance to TJMax (= TJMax - TJunction) > 25C.

The upshot then is that to keep your TCase within Intel spec means that the Distance to TJMax must be kept over 25C. This is in line with what many overclockers are currently doing. What do you think?

[hardtarget]

Not sure if theres a problem with my CPU temps as I get a difference of nearly 10c between the two cores I do have it OC, but as you can see from the pics its not really pushing it by any means.