Real Temp - New temp program for Intel Core processors

[unclewebb]

I'll think about it. The RealTemp / RivaTuner plugin works good for this. You can use RivaTuner instead of RealTemp and the RealTemp plugin will show your Intel temps on a G15. I was doing that for a while so I wasn't sure if G15 support was really needed. I'll put it on the things to maybe do list.

I was doing some beta testing tonight and one user with a newer 5870 card was having problems with RealTemp crashing. If you have any sort of ATI card or cards in your system, could you post a screen shot of the RealTemp GPU window and let me know if this feature works. I haven't seen any problems when testing on Nvidia hardware. I'm just trying to find out how many users this problem is happening to. So far it gets one thumbs up from a 4670 owner.

http://www.sendspace.com/file/5100no

[ekerazha]

v 3.40 bug: RealTemp shows 0.00 Mhz as CPU freq, CPU is Atom 330, v 3.00 worked fine. It also say "133.86 x 0.0", so there's a bug when reading the multiplier.

[ExBrat72]

If you have any sort of ATI card or cards in your system, could you post a screen shot of the RealTemp GPU window and let me know if this feature works.

Is this what you need Kevin?

I have a Sapphire ATI 5850 and it all seems to work OK.
Between version 3.46 and v3.5 I have lost the ability to display the GPU temp in the Tray Info. The CPU displays OK but the GPU temp will not stick with the 'Show icon and notifications' setting. (it was working OK with 3.46 before I went to 3.5) Whatever I select it reverts to 'Only show notifications'.
(Win 764 Pro)

[eva2000]

http://www.intel.com/pressroom/archi...202comp_sm.htm

SANTA CLARA, Calif., Dec. 2, 2009 – Researchers from Intel Labs demonstrated an experimental, 48-core Intel processor, or "single-chip cloud computer," that rethinks many of the approaches used in today's designs for laptops, PCs and servers. This futuristic chip boasts about 10 to 20 times the processing engines inside today's most popular Intel® Core™-branded processors.
----------------------------------------------------------------------------
The system tray isn't big enough to monitor all of them.
I don't even want to think about two of these on the same board.

haha that's what ATI Eyeinfinity is for 3x2560 = 7680 horizontal resolution

[JohnZS]

Hi Kevin
Once again fantastic software.
Just noticed that a little issue with RealTemp 3.50. With my Quad Core Extreme Processor, if I adjust the multiplier to 9.5 or something the %load value does not read as 100% when the processor is loaded. For example a multiplier of 9.5 gives a % load of 97.4% even though the processor is @ 100% load.
The clock speed is also read incorrectly too as RealTemp says that 9.5 multi = 10 multi.
Thanks
John

[unclewebb]

ExBrat72: Thanks for the screen shot. One user had a pair of 5870 cards in Crossfire and was having problems with RealTemp starting up. RealTemp seems to work fine with single ATI cards or with x2 cards. He's going to do a W7 reinstall to see if that makes any difference since he was having a few other issues besides RealTemp.

Thanks JohnZS. Those Extreme processors always give me headaches. I'll send you a test version to see if we can finally fix this issue for you.

ekerazha: Recent versions of RealTemp have been switched to using the internal timers to determine the multiplier. This is a more accurate method except when a CPU like your Atom doesn't have any of these timers. Try adding this to your RealTemp.ini file:

MSRMulti=1

Hopefully that will use the original method. I'll try to make this automatic in future versions so if RealTemp doesn't find any working timers, it will automatically switch to this method.

Any problem solved is a new problem made.

I kind of like that quote John.

Between version 3.46 and v3.5 I have lost the ability to display the GPU temp in the Tray Info.

I can't remember changing anything that would cause this but I'll look into it. Does this option still exist in the RealTemp Settings window? Maybe I can blame this on Windows 7.

[JohnZS]

Thanks Kevin
Ah yes those Extreme Editions can be a pain, if it isn't the extreme sensors it is the manic multipliers!
Intel certainly like to keep you on your toes
That quote "Any problem solved is a new problem made" rings true for pretty much everything
Thanks once again for your efforts and hardwork
John

[Razer(x)]

Maybe I can blame this on Windows 7.

Nope, it works perfectly for me

[ExBrat72]

I can't remember changing anything that would cause this but I'll look into it. Does this option still exist in the RealTemp Settings window? Maybe I can blame this on Windows 7.

Sorted that one Kevin, put it down to me or win7...
I unchecked the GPU 'show icon' in settings, rebooted the PC and rechecked 'show icon' ... all working now - Thanks
John

[mejobloggs]

I'm new to this, and 150pages is a bit much for me to read...

HWMonitor shows my load temps as 52°C, Real Temp shows 42°C. Which one do I believe?

Using RealTemp version either 3.00 or 3.40, can't remember. Will check when I get home

[unclewebb]

Post a screen shot of RealTemp and tell me some details about your processor in case the screen shot isn't obvious enough and I will tell you why different programs are reporting your core temperature differently. Intel didn't do a great job documenting how to get accurate core temperatures out of their CPUs so different programmers have made different assumptions about the specifications of your CPU.

[endlesszeal]

Hi Uncle. I am having the same "problem" as mejobloggers. Its on a laptop with a Pentium T5200. It definitely doesnt feel like 74C because when my other laptop with RealTemp hits that high, I can feel heat emitting from the exhaust fan and keyboard.



This was taken with it sitting on an Antec Laptop cooler with 2x60mm fans spinning on "high."

[unclewebb]

Here is the Intel Spec for a T5200.
http://processorfinder.intel.com/det...px?sSpec=SL9VP

Thermal Specification: 100°C

Another name for thermal specification on the mobile processors is TJMax or the maximum junction temperature. Intel Core processors don't have a typical thermometer in them that software can read. They have a reverse thermometer that counts down to zero as the CPU gets hotter. On these early CPUs, there is no way to read TJMax from the CPU. All you can do is look it up on the Intel site. It's possible the Intel information I posted is wrong but for the early laptops with a CPUID = 6F6, TJMax = 100C is usually right.

Here's the formula used:

Reported Temperature = TJMax - Digital Thermal Sensor Reading

The raw data coming from the digital thermal sensor is reported in RealTemp in the Distance to TJMax boxes.
In your picture the formula would be:

Reported Temperature = 100 - 28
Reported Temperature = 72C

It's obvious that HWMonitor is using TJMax = 85C in this calculation:

Reported Temperature = 85 - 28
Reported Temperature = 57C

You have to remember that this temperature is the temperature at the hottest spot on the core and dissipates rapidly from that point. 72C on the core is not going to translate to 72C on the bottom of the laptop. Many laptops have very marginal cooling systems. They don't have the room for a large heatsink and fan so they get hot. The heatsink may not be attached snugly to the CPU in some laptops. If this is the case then the CPU can get very hot within seconds of pushing the power button.

I pulled my laptop apart and shoved a penny in there to take up some of the slack so the heatsink could make better contact with the CPU. If you pull your laptop apart and it is easy to wiggle the heatsink back and forth then that could be part of the problem with the high core temperatures you are seeing.

My wife had a Dell T7200 laptop and it used TJMax = 100C and the reported temperatures didn't look too unusual at all for a laptop.

If you are using Windows 7 or Vista, there is a setting in the Control Panel -> Power Options called Minimum processor state. That needs to be set to a low number like 5% to get your CPU to idle down properly. The CPU might be getting more core voltage than it should be at idle if this is not set correctly. In XP set the Power Schemes to Mobile / laptop processor to control this.

My opinion is that RealTemp and the Intel documentation is correct for your CPU. If you think TJMax should be set to 85 then you can manually do that in the Settings window of RealTemp but I think your reported core temperature will be less accurate if you do.

[mejobloggs]

Post a screen shot of RealTemp and tell me some details about your processor....

I have a Q6600 B3, I'll post a screenshot later

[unclewebb]

For the Q6600 B3, RealTemp by default should be using TJMax = 90C. If some other software decides to use TJMax = 100C for that CPU then it will report a core temperature 10C higher than what RealTemp reports.

I went with 90C for the B3 because the B3 consists internally of two E6600 B2 cores. The E6600 CPU when tested has a TJMax = 90C. No other software to the best of my knowledge agrees with that.

Intel says that they increased TJMax by 10C when they went from the Q6600 B3 to the G0. Most Q6600 G0 owners would agree that TJMax=100C for their CPUs so 90C for the B3 sounds reasonable to me.

TJMax is full of misinformation. You'll have to decide for your self what value is most credible. I've seen some Q6600 CPUs where core 2 and core 3 seem to be set 5C higher than core 0 and core 1 so your CPU might actually be close to TJMax = 90, 90, 95, 95

Post a RealTemp CPU Cool Down Test and I'll have a look. Include your room temperature and let me know if your case is open or closed.

[mejobloggs]

I can't find the "RealTemp CPU Cool Down Test"

But here are the idle/load temps. Side of case is off

Room temp: 22 deg C
Cooler: Megahalem
Settings: stock, everything on auto

[burebista]

I can't find the "RealTemp CPU Cool Down Test"

Press that Sensor Test button.

[mejobloggs]

Thanks

[unclewebb]

Your sensors look fine. These 65nm sensors didn't have the sticking issues that the newer 45nm sensors have. You would have to heat your CPU up to about 70C before I could be absolutely sure but it looks like TJMax is fairly consistent across all 4 cores. Actual TJMax might be 2C higher on core 2 and maybe 1C higher on core 3. Intel agrees that due to manufacturing variances, TJMax is not 100% consistent from one CPU or core to the next but have never released any information about an error specification like +/- 5C, etc.

A Quad core with a good cooler like a Megahalem should idle about 7C or 8C above room temperature which in your case would be about 30C. Core 0 tends to be the most accurate and in your screen shot of RealTemp at 1600 MHz and 1.10 volts, it is showing 28C which looks reasonable to me. All of these sensors have some slope error and if they read a couple of degrees too low at idle, that's not unusual at all. They weren't designed to report 100% accurate idle temperatures. They are much more accurate at 70C and beyond.

HWMonitor has decided to assume that TJMax=100C.

Reported Temperature = TJMax - Digital Sensor Reading

Both programs are reading data from the same digital sensor. It's obvious that if you assume a higher TJMax value that the reported temperature will also increase by the same amount. RealTemp uses TJMax=90C for a Q6600 B3 so HWMonitor will typically report temperatures 10C higher.

Flip a coin and use whatever TJMax value you think is correct. You can even split the difference and use 95C. Personally, I wouldn't do that because I believe that core 0 and core 1 and core 3 are likely very close to 90C and core 2 is very close to 92C.

I don't believe a lot that was said at Intel's Developer Forums last year when it comes to TJMax but their updated list of TJMax values shows a Q6600 B3 has a TJ Target = 90C.
I believe that one.

[JohnZS]

I am wondering if Intel's engineering department ever read the specs for TjMax
Or was there such a large acceptable % of error on this equation?! but providing most CPU's fitted the line/slop then that was the given value.
Anyway Any news on the QX Multipliers and % load engima?
John

[ekerazha]

ekerazha: Recent versions of RealTemp have been switched to using the internal timers to determine the multiplier. This is a more accurate method except when a CPU like your Atom doesn't have any of these timers. Try adding this to your RealTemp.ini file:

MSRMulti=1

Thank you. I've also seen SpeedFan does report about 10°C more than RealTemp, probably it uses a TJmax of 100°C instead of 90°C for the Atom 330.

[unclewebb]

I just had a look at the Intel docs and most Atom CPUs are listed with a thermal specification of 90C. The Atom 230 and 330 however show a thermal specification of 85.2C.

As with most Intel documentation, it's unclear whether this is a TJMax value, a Tcase max value or maybe a TJ target value or maybe just a mistake. Who knows. Actual TJMax might be 85C or 90C or 100C or any number in between. I've never tested an Atom so 90C was my best guess.

Here's the list of Atom CPUs on Intel's website:
http://processorfinder.intel.com/Lis...008&SearchKey=

I don't think that actual TJMax is only 85C for your Atom 330 so I'm going to leave RealTemp as is. You can manually adjust it to 85C if you think the Intel docs are accurate. With a passively cooled CPU or one with a tiny heatsink and small fan, I don't have any method to try and calibrate these.

[TEDY]

Don't know where to post but is OCCT 3.10 the last version? And MEMTEST is grayed :/

[topfuel]

what should be the calibration/tjmax settings for i7?

as there seems to be many variation of screenshots showing i7 in diff temps

[burebista]

For Nehalem/Lynnfield you don't need to calibrate TJMax because is written in MSR.