Real Temp - New temp program for Intel Core processors

[CompuTronix]

95c is the correct Tjunction Max value for the Q6600 G0. Prior to Real Temp, the G0 stepping processors were wrongly assumed to be Tjunction Max 100c, and the 45 nanometer processors were wrongly assumed to be 105c.

Thanks to unclwebb's research, testing and analysis, we now know that the 45 nanometer processors are 95c. The only exception to this, which has yet to be confirmed, is the 7000 series, which may be Tjunction Max 90c.

Comp

[WoZZeR999]

Look I'm not doubting anything, Pure logic sense now : I just would find it weird that my Q6600@3.6 at 1.42 volts would idle at 23-25°C then under watercooling if it's based on 95 and not on 105°C... Sadly I cannot prove anything as I sold that Q6600 ( think Vid was 1.2125 ) but I know what I've seen during 2 watercooling setups, idle temps in coretemp (based on 105) were 32-35°C idle which sound more beleiveable than 22-25°C unless I live in some cold country and my ambient is sub 20's...

Think my logic makes sense not ?

While true, at the same time when running the calibration on the Q6600, I have a hard time believing that an underclock of .9v and 1.6GHz would run the processor 5 or 10c above the water temp.

On a side note, CoreTemp is using TJmax of 100c(default value) on my Q6600 now, so now the temps are only 5c apart.

[unclewebb]

Leeghoofd: The difference between CoreTemp and RealTemp for the Q6600 GO is only 5C. CoreTemp assumes TjMax=100C and RealTemp assumes TjMax=95C which is based on the testing of my Q6600 G0 with an IR thermometer.

Reported Temp = TjMax - Digital Thermal Sensor Reading

I think everyone assumed TjMax=100C because that gave believable readings at idle. Before RealTemp came along, every program blindly assumed that the temperature curve from these sensors was totally linear but none of them are. Since the first 65nm processor left the factory, the sensors have not been 100% accurate from idle to TjMax. I've seen only one mention of this by Intel in an interview at Anandtech which surfaced soon after RealTemp was released.

Intel designs and calibrates these sensors to accurately trigger thermal throttling and thermal shut down but they were never meant to be used to accurately report idle temperatures.



What I discovered was that if a user took 5 minutes to calibrate his CPU then these sensors could also be used to provide fairly accurate idle temperatures. Using the correct TjMax results in extremely accurate load temperatures. Using the wrong TjMax will help cover up the problems these sensors have at idle but by 60C, your load temperatures will be off by 5C in the case of the Q6600 G0.

On most Q6600 G0 processors, doing things the correct way or the CoreTemp way doesn't make a huge amount of difference. It seems that most of the 65nm chips read about 5C too low at idle so boosting TjMax by 5C made everything look great. On 45nm chips like my E8400 which read too high at idle, using the wrong TjMax combined with ignoring the sensor issues at idle results in reported core temperatures that are a long way off from reality.

Edit: I believe that thermal throttling starts to happen just before TjMax is reached. I've tested a few different processors and throttling has started each time at DTS=2 or 3 as reported by PROCHOT# being set in the processor. It seems to be designed by Intel to kick in at this point to help prevent the CPU from ever reaching TjMax.

[k4vz0024]

I have strange problem when Rel Temp 2.69.1 running under Vista Ultimate pro 32 and 64. When I program started first time and I set up Start Minimized its actual for a while. But now every time when Real temp started under Startup when System boot up I see a Real Temp window, like as Start Minimized off. What’s the wrong with my system?

Tx

[unclewebb]

k4vz0024: This is a problem for some Vista users and I have no idea what is causing it. I will be looking into this further in the near future. The Start Minimized feature has always worked for me in XP. I might have to create a Registry entry to get this to work properly in Vista which so far I've been avoiding since I hate programs that add junk to the registry.

[WoZZeR999]

Have you tried running the program in Admin mode? On ultix64, start min. works correctly every time.

[k4vz0024]

Yes. I´m Administrator.

[sdsdv10]

Uncle did testing on an actual Q6600, I believe that most other programs used Intel docs to find the Tjmax.

This is a common misconception. There is no Intel published literature on the Tjmax for desktop CPU's. If I understand things correctly, Intel has only released this information for their mobile/laptop CPU's. For the desktop conterparts the other temperature measuring software developers are simply assuming a value, either based on a number for a previous generation CPU or for a related mobile CPU (as noted above). Unclewebb is the only person I found on any forum who is developing his software in conjuction with experimental evidence (across all types of CPU's, 65nm and 45nm as well as dual core vs quad core). A big hand to you unclewebb

[unclewebb]

Thanks sdsdv10. You did a great job of explaining the problem. I've always thought that the reason Intel has never published TjMax for the desktop processors is because it would expose the inaccuracies in these sensors at idle. Most people running a mobile chip in a laptop never notice this problem because they're not using water or a massive air cooler.

My first E6400 also had a saggy temperature curve at idle and can report below ambient temperatures even with the correct TjMax=85C. There is no single TjMax value for a processor that is accurate at both idle and at TjMax. That's why RealTemp lets you choose calibration factors to help out down low so you don't have to use the wrong TjMax which will screw up your load temps.

I think Intel TAT influenced what other software uses for TjMax. People forget that TAT is a laptop testing tool and may have no relevance to the desktop processors even though it has that nice Intel logo on it. I decided to start with a clean sheet of paper and so far that's worked out very well.

k4vz0024: Can you try dragging the RealTemp folder to your Desktop and see if Start Minimized works there. You don't have to re-boot. Just try starting the program and it should minimize to the System Tray area. I've been trying to find / understand what's causing this bug but it seems to only be effecting a few users. I can't figure out why it works for some Vista users but not others.



After it starts up if it is still not working does it show Start Minimized as checked and is there at least one core selected so it has a System Tray icon to minimize to?

[unclewebb]

mcoffey: Did you try the calibration procedure as outlined in the RealTemp docs? As the graph a few posts ago shows, none of these sensors are accurate at reporting idle temperatures. Not the 45nm sensors or the 65nm sensors. Intel did not design or calibrate their DTS sensors for that purpose.

You can choose to use TjMax=105C and that will make your idle temps look more believable but I guarantee you that your load temps will not be accurate. Changing TjMax does not address where the problem with these sensors is. My E8400 reads about 6C to 8C too high at idle depending on the room temperature. Your 45nm QX9650 reads too low by a very similar amount. By using the calibration features in RealTemp you can do a pretty good job of correcting for the errors in the DTS. It will never be 100% correct at idle but from idle to TjMax it will be far more accurate than any other software can report.

[WoodButcher]

@ mcoffy, I run a Q6600 in my new box, the loops and blocks are IMO mid -high end and I also watch the temps. Using the D-tek v1, comparing ambient air and water, I've adjusted the idle to 3-4c over water. This will change as you add vcore. My water temps are taken before the rad, for a look at the box here is the thread on Anand.

[i43]

@ unclewebb

here my HEAT-UP run
system :
QX9650 / Maximus Extreme (bios v1007) / True120 *! FanLESS !*/
for this run everithing was set to [AUTO]/ Factory Default / in bios
this means C1E ; EIST ; CoreTM function - [ ENABLED ]

Ambient measured 23~24C*

i HEAT-UP (forced heat-up!) on stock speeds, but fanless and run small FFT PRIME - on all 4 cores / ( no affinity or such things touched )

Throttling triggers somewhere @ 93~94 degrees C* - read by Realtemp

and only CORE#0 was triggered ..aka '' acted as switch '' ..
i let it run for few more minutes .. but it push back multi and reduce load, what in turn, -- lowered temperature of the whole CPU unit,
so it remained Prime Stable even under throttling ... any way
i did not see any core dropping in prime, no Crash or Screen hangup until this stress
-----------------------
i think i get it correct .. quad consist of 2 dual's so
core#0 & core#1 = [ Pair No#1]
core#2 & core#3 = [ Pair No#2]

dont know is this difference something to worried about ?
the [ Pair No#2] stay @ 88~89 C* // 6-7C* To TjMax remaing
-----------------------

few rows from RT Log
~ ....
13:56:50 93 93 89 87
13:56:51 93 93 89 87
13:56:52 93 93 89 87
13:56:53 93 93 89 88
13:56:54 93 93 88 87
13:56:55 93 93 89 87
13:56:56 94 93 89 88
13:56:57 93 93 89 88
13:56:58 93 93 88 88
13:56:59 93 93 88 87
13:57:00 93 93 88 88
13:57:01 93 93 89 88
13:57:02 93 93 89 88
13:57:03 93 93 88 88
~~~ .......
(check the time stamp )

here a shoot

------------

CY

[unclewebb]

mcoffey: My experience with calibrating with water is zilch but I think WoodButcher has some good advice to shoot for 3C to 4C over your water temperature. With the numbers you posted, 3 of your 4 cores would have no problem being adjusted to that level and I think even your last core will be really close to that range after adjustment. We're not looking for perfection here. The sensors aren't perfect so your idle temps won't be perfect even after a calibration. They should be very close though. Many users with 45nm Quads have had good success using RealTemp with TjMax=95C. Maybe someone running water has a good air cooler in their closet and they could run both set ups at low MHz and low core volts on the same CPU to compare idle temps between the two.

i43: Your results are very similar to what I had when testing my Q6600 on the last page. My Quad peaked at 93C on core0 and core1 and then core2 and core3 only got up to 88C as thermal throttling held the processor at this temp as Prime continued to run without any problems or crashes. I might give it some more core voltage to create some more heat next time to see what happens.

The difference between the two sets of cores is normal. I tried this test at idle with no heatsink attached and the temps across all 4 cores were within a degree or two which showed that this isn't a sensor issue. With a heatsink mounted and the CPU fan on high, the difference between cores was about 3C instead of 5C. Maybe a smart person who understands heat pipes, heatsinks and thermodynamics could explain this difference in the two sets of cores within a Quad.

[TAOTAO161]

hello all,
to who i need to believe ?
the rig is in the signature.

on PRIME95 TEST

REALTEMP show core1+2 - 63C core3+4 - 60
everest show core1+2 - 68 core3+4 - 62
TAT show cpu0 - 63 cpu1 - 65

pic

[WoodButcher]

Maybe someone running water has a good air cooler in their closet and they could run both set ups at low MHz and low core volts on the same CPU to compare idle temps between the two.

I knew those shoes didn't belong to my wife!
I laugh,,,, take my wife,,,, Please!

Anywho, I do have a U120, I think it was the 2nd revision before the extreme. I need to do a little tubing work so I'll see if it will fit w/o a complete rebuild, if I can manage I'll do it, no promises though.

I'd still be guessing though unless someone else has the same cooler mounted on a Q6600.

[unclewebb]

At low MHz and low voltage at idle, there isn't much difference between a good cooler, a great cooler or even water. Heat transfer is never perfect so core temps at idle should always be a few degrees above either your air or water temp. An Ultra 120 would be perfect to do some testing with. The wife is on holidays so it's tough to sneak too much new hardware into the house without looking suspicious! What's that???

TAOTAO161: Read the RealTemp docs and see what makes sense. You should be able to do some testing and answer your own question about what software is telling you the truth.

[WoodButcher]

Haha, ok, like I say if I can squeeze her in I'll do it. Kinda tight but I want to make a few changes to the fans and tubing so it may work.

[unclewebb]

mcoffey: Thanks. I really appreciate that you took the time, did your homework, did some testing and were able to come to your own conclusions. Being skeptical is a good thing. Too many users for far too long have put their trust in programs that have no real world testing to back up the temp numbers they put out.

It's very difficult to argue with the results from the test I did on my 45nm E8400. TjMax=105C is simply impossible.
http://www.xtremesystems.org/forums/...&postcount=573

Quad core processors are simply two Dual Cores strapped together so I found it tough to believe that TjMax=105C for them either. Any user with a Quad that does some honest testing usually comes to the same conclusion.

Spread the word. There are still a lot of users hanging on to the old way of doing things.

WoodButcher: Maybe you could rename your computer, "Welcome to the jungle!"

[kpo6969]

Is 2.69.2 out somewhere? I saw someone using it on another site. Thanks

[unclewebb]

2.69.2 is available from the beta section.

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

I didn't announce it because there were no real changes. The only change was a custom profile for one user to get his 45nm Quad looking more believable. He provided me with some hopeless looking data and I created a custom profile for him. He's happy now!

If you are using 2.69.1 then there is no need to download 2.69.2

I might have time later today to add a couple of features. It's too nice a day out at the moment to be inside programming.

[kpo6969]

Thanks

[TAOTAO161]

their is a specific calibration for Q6600 ? my vid is 1.3000V

[unclewebb]

their is a specific calibration for Q6600 ? my vid is 1.3000V

Read the docs and follow the procedure as outlined there.

I have found that with an open case when you bring a Q6600 down to 1600 MHz and about 1.10 volts you should see idle temps about 5C or 6C above your air temp with a good air cooler. If you have a good water cooling system then try to get RealTemp to report idle temp on all 4 cores at about 4C above your water temp. It's not an exact science but a guide line which will get you pretty close to the real temperature.

Individually adjust your Idle Calibration factors to get all 4 cores equalized at Idle. This calibration will improve the accuracy of your reported temperatures from idle to TjMax.

Edit: I did a quick calibration of my Q6600 about 30 pages ago. I got lucky and found it on my second try!
http://www.xtremesystems.org/forums/...postcount=1006

[unclewebb]

Finally had a chance to add in an alarm feature to keep CompuTronix happy.



It's only a visual alarm at the moment where you will see the application and the RealTemp system tray icons flash. I will be adding an alarm that you can also hear as soon as I find an appropriate .wav file.

I've found that Quad core processors at full load will usually have the first two cores reporting higher temps than core2/core3 so I've created the ability to set two different alarm points if you need to compensate for that. Single and Dual core processors will only have one alarm temperature to choose. The temperature range is from 0C to 125C and if you don't want an alarm then just leave it unchecked. You won't likely ever hear an alarm if you set it to 125C since that is believed to be the thermal shut down temperature of the desktop processors.

It's available for download in the beta section:
http://www.fileden.com/files/2008/3/...alTempBeta.zip

Download, unzip and copy the new RealTemp.exe into your RealTemp directory.

To make room for this option in the Settings window I moved the Core Order to the main screen. It has been renamed APIC ID which is the technical term for it. If you have a Dual core you won't see this information since your APIC ID will always be 01. This information is only used for the grouping of the cores of a Quad core processor.

The other small feature added was a user selectable anchor position.

By default, if you do a double right mouse click on the RealTemp GUI, it will jump to the top left corner of the screen. If you would like it to jump to a different location then position RealTemp where you want it, hold down the Shift key and do a double right mouse click while holding it down. This will set a new anchor position.

Now when you double right click on RealTemp it will jump to this new custom anchor position that you have chosen. I realize that setting a new anchor position is a little cryptic but I'll try to document it early on in the New Features section. Other than that I guess it will be a special feature for the XS guys that RTFM!

If anything isn't working quite right with these new features then let me know. If you have any sounds that you would like to hear for an alarm then e-mail them to me. I'll try to code the final version so users can custom choose their own .wav file to play for an alarm.

[WoodButcher]

Finally had a chance to add in an alarm feature to keep CompuTronix happy.

It's only a visual alarm at the moment where you will see the application and the RealTemp system tray icons flash. I will be adding an alarm that you can also hear as soon as I find an appropriate .wav file.

A toilet flush would be appropriate!

I've ordered new fans so when I get them I'll see about doing a little air / water comparisons. I'll record ambient air, water, idle and load temps uncalibrated, cpuz, screenshots when possible, is there anything else that would help?