Real Temp - New temp program for Intel Core processors

[rambler358]

Unclewebb - any chance you could make RealTemp compatible with the Logitech LCD keyboard displays, like CoreTemp?

Also, in v3.00 of RealTemp, the individual core temp readings in the system tray come up defaulting to be out of order core-wise. Not a biggie, just something I've noticed.

Thank you for all your efforts, and keep up the great work!

[unclewebb]

Also, in v3.00 of RealTemp, the individual core temp readings in the system tray come up defaulting to be out of order core-wise.

We call that a Windows 7 feature. Randomly ordered core temps in the System Tray. Their OS is designed so a program can insert one system tray icon. Any program that needs to insert multiple icons like a monitoring program needs to do, has a difficult time getting them to line up in the correct order. They get inserted randomly, usually to the left or the right of the previous icon. There's no way I've found to predict the order that Windows 7 will give me. The same code works 100% in Vista or XP at keeping the tray icons ordered.

I use the RealTemp / RivaTuner plugin so I can see some data when using my G15. I'm in the process of updating that at the moment so it might be buggy. Hopefully in another day or two it will be 100%.

I might add G15 code directly to RealTemp someday. I'll put it on the things to maybe do list.

Edit: Here's the plugin if you use RivaTuner.
http://www.fileden.com/files/2008/3/...507/RTCore.zip

I just got word that it is working very good. The only bug that still needs fixing is the JohnZS QX load bug.

[rambler358]

Thanks for the info Unclewebb. I don't use RivaTuner, but use EVGA Precision instead (same author though) - a simpler interface with all the features I need.

[JohnZS]

Hi JohnZS:

If the RealTemp Load meter works fine when you are using the default multiplier for your CPU but screws up when you are using a different multiplier then that would be a bad thing. A user with an early QX processor was helping me with this issue a long, long time ago. If you want to be the new helper then I can send you some stuff to test for me. Volunteer testers with unique CPUs are always needed.

As for the endless question, "What is TJMax?", I can honestly say, "I don't know." Intel doesn't seem to know and neither do I. If you still have that wonky Quad that you used to have then I think TJMax varies by as much as 10C from core to core. I don't know if your cores are TJMax 95 to 105 or maybe TJMax 100 to 110 or some combination in between. Intel never released enough information to clarify this. With sticking sensor issues and sensor slope error, it can be very difficult to accurately prove this. A sad situation but I'm only the messenger.

I just had a look at my code and here's a note I wrote a long time ago:

"Intel says QX9650 TJ Target = 95C but it seems more like TJMax = 100C during testing so leave it as is"

The truth is that there probably doesn't exist a QX9650 that has all 4 cores with the exact same TJMax. Once I get the RivaTuner plugin updated, maybe we can look at your numbers one more time to try to come up with a good guess at TJMax.

Hi Uncle
Those are the symptoms my wonky Quad Core is experiencing, (along with a guestimation for the TjMax by Intel, and sensors which might as well be random number generators...once again courtesy of Intel)

You'd have thought with Intel's famed high quality fabrication process they would have put fairly decent sensors into their Quad Core Extreme Range

John

[unclewebb]

The sensors on the QX processors almost seem worse than the rest of the Quad processors. They're probably exactly the same but when you pay a premium price for a CPU, your expectations go up and I know from your previous testing, your sensors definitely don't live up to that. Intel learned their lesson. The Core i7 sensors are world's better than the 45nm Core 2 sensors. Whoever decided to save a few pennies when sourcing the 45nm Core 2 sensors must have been reassigned to a new department.

I sent you a testing program in a PM. Thanks for helping me out John.

[Marvin82]

Is RealTemp 3.20 The last Version
Thanks

[randomizer]

Is RealTemp 3.20 The last Version
Thanks

Whatever you download from the URL to the RealTempBeta that has been posted many times is the latest. unclewebb always overwrites the same uploaded file. Sometimes you have to clear your cache though because it will for some reason download the old version (for me anyway)

[JohnZS]

The sensors on the QX processors almost seem worse than the rest of the Quad processors. They're probably exactly the same but when you pay a premium price for a CPU, your expectations go up and I know from your previous testing, your sensors definitely don't live up to that. Intel learned their lesson. The Core i7 sensors are world's better than the 45nm Core 2 sensors. Whoever decided to save a few pennies when sourcing the 45nm Core 2 sensors must have been reassigned to a new department.

I sent you a testing program in a PM. Thanks for helping me out John.

Thanks Uncle, all received, you have mail
Yes it is a bit disappointing that the premium processors have the dodgy sensors, I guess it is because I purchased this back in December 2007 (just after release), I would imagine the Quads with the newer C1 and E0 steppings have better sensors?

(One would hope)

Thanks for your help
John

[unclewebb]

I would imagine the Quads with the newer C1 and E0 steppings have better sensors?

That doesn't seem to be the reality. All 45nm Core 2 Dual and Quad core CPUs have sensor issues. Some processors are better than average and some are worse but they all have issues. Intel's suggestion at IDF was to buy a Core i7. They have better sensors and you certainly need them considering the heat they can put out.

[JohnZS]

Doh! So all in all the Core2 45nm manufacturer process of the sensors was quite... poor?
Do Core i7's run a lot hotter than Core2 Quads?
I am considering upgrading in 2010 to the Corei7 Extreme Refresh (32nm?) Maybe I will be reunited with dodgy sensors
John

[gymenii]

I picked a good one

[OC Nub]

I have had 5 45nm CPU's. ALL have had least 1 stuck sensor at idle. Have an e8400 CO that has both sensors bad at idle and load, ran it for well over a year at 4.05ghz, never knew what my temps were.

[unclewebb]

Coming from the RealTemp guy this is going to sound a little dumb but temperatures really aren't that important. Cool them as much as you can afford to and they will be able to overclock more. If they are not stable, cool 'em some more or back off your overclock. That's really all there is to it. No fancy temperature monitoring program needed.

gymenii: Your CPU looks like a winner. Two sensors that actually work and both of them on the same CPU.

They must have manufactured that one on a Wednesday. No way it could have been built on a Monday or a Friday.

JohnZS: When you start overclocking Core i7 and feeding them lots of voltage, they start consuming and putting out a lot of heat energy. Far more than Core 2 Quads, especially when you enable hyper threading and run 8 threads of something crazy like LinX.

[-X-hellfire]

Done some more testing as I got my other computer sorted out,
updating a firewall creates a millions of popups that needs to be
redone which takes forever and forever but finally I got some
time to test and calibrate RealTemp


bios

I have previously always had the default Vcore 1.275V at 3.2GHz but
after a I rebooted while installing some software I got a three beeps from the
PC-speaker and It had done a safe reboot and set the cpu to 2.1GHz and 8x266
. . . well the cpu might need some some juice so I upped Vcore to 1.300V and it
has been stable so for through 4 x Prime95 for several tests.

I wondering why RealTemp reports a Core VID . . . MAX 1.2750V when bios (1.) shows

T-Balancer has a digital sensor (cpu-D1) on the bottom of the TRUE
and is the idea is that when it is calibrated it offers enough guidance
to automatically adjust the fans

Cooling

cpu-fan to 100% and case-fan to 100%
. . . don't want to overclock the mem at this point and keeping
it at default 800Mhz

cpu-D1 is 35C
1a
1a-00-idle-0xC1E-3200-8x400-1V30-4x(000)-P95-mem800-0cal





cpu-D1 is 35C
1b
1b-99-load-0xC1E-3200-8x400-1V30-4x(99)-P95-mem800-0cal






I entered what you suggested earlier unclewebb with a
TJ Max of 100C 100C 105C 105C

RealTemp calibration 0.5 0.5 2.2 -0.5 (known as cal1)
cpu-D1 is 33.5C
2a
2a-00-idle-0xC1E-3200-8x400-1V30-4x(000)-P95-mem800-cal1-settings





cpu-D1 is 35.5C

2b
2b-99-load-0xC1E-3200-8x400-1V30-4x(99)-P95-mem800-cal1





cpu-D1 is 34.5C
2c
2c-00-idle-0xC1E-3200-8x400-1V30-4x(000)-P95-mem800-cal1






Pretest

Tried to bring up temperatures to about 65C by lowering the
cpu-fan to 22% and case-fan to 80%
( don't want to overclock the mem at this point and keeping
it at default 800Mhz, otherwise I might had tried to overclock the cpu to 3.6Ghz )
. . . anyway the Q6600 cores reached 63C 63C 65C 64C

cpu-D1 is 40.0C
3a
3a-99-load-0xC1E-3200-8x400-1V30-4x(99)-P95-mem800-fan1-cpu22-fan4-case80-cal1-c65




cpu-D1 is 34.0C
4a
4a-00-idle-0xC1E-3200-8x400-1V30-4x(000)-P95-mem800-fan1-cpu22-fan4-case80-cal1-c65





Heat Up Phase

Tried to bring up temperatures to about 65C by lowering the
cpu-fan to 22% and case-fan to 80% as I mentioned before in the
. . . the pretest (3a) went on for about 10 min but the "Heat Up Phase" was about
half of that time maybe less and Q6600 cores reached 60C 60C 63C 61C

cpu-D1 is 37.0C
4b
4b-99-load-0xC1E-3200-8x400-1V30-4x(99)-P95-mem800--fan1-cpu22-fan4-case80-cal1-c65





When the "Cpu load" was 30% I noticed that the Prime95 dropped
from 100% to 0% in the core graphs and I stopped the 3 x Prime2004
as fast as possible.
cpu-D1 is 37.5C (TRUE-temp is rising even though the cpu is idle)
4c
4c-00-idle30-0xC1E-3200-8x400-1V30-4x(000)-P95-mem800-fan1-cpu22-fan4-case80-cal1-c65






cpu-D1 is 36.5C (TRUE-temp finally going down as the cpu is idle)
4d
4d-00-idle-0xC1E-3200-8x400-1V30-4x(000)-P95-mem800-fan1-cpu22-fan4-case80-cal1-c65









I might try to enter the data logs from RealTemp and T-Balancer into Excel
and make some pretty graphs that might some more datapoints show and how the
TRUE-cpu-D1 digital sensors is trailing some 30 to 60 seconds behind
RealTemp (I placed it where it didn´t ruin the looks of the TRUE too much)
which depending on how you set it up can work reasonable as the TRUE
has some overcapacity it won't hurt the cpu too much.

It does however create some difficulties to master all areas that needs
cooling when overclocking and have the CrazyCool thing enough cooled
at all times even when the fans are fine tuned to run the computer as
silently as possible (its a matter of taste how much heat you think is ok even
if seams as it's stable some stuff might be hurting real bad and create
problems later on)





It looks like the core #2 it the one that needs some more
fine tuning.

What do you think about this first calibration unclewebb?







.

[unclewebb]

When not overclocking and when your voltage is set to AUTO, the motherboard is supposed to use the VID information inside your CPU to set the appropriate core voltage. It uses the high VID information to set the voltage at full load and it uses the low VID information that RealTemp displays to set your idle core voltage when EIST is enabled and your processor's multiplier drops down to 6.0. You typically end up with slightly less voltage than the VID number due to vdroop. Most motherboards ignore the VID information within a CPU as soon as you start overclocking or if you set the voltage manually.

You need to open up the Settings window in RealTemp so I can have a look at what calibration settings you are using.

You'll get more consistent results if you use Prime95 instead of Orthos or SP2004.
http://mersenne.org/freesoft/

That let's RealTemp control the exact shut down time of Prime95 on all 4 cores simultaneously so you don't have to do this manually. Then you can go for a coffee and come back 10 minutes later and see the results. You get more consistent and repeatable results when you do things this way.

At the moment it looks like you have a split TJMax similar to my Q6600. It might be TJMax=100,100,105,105 like I use or maybe core 2 and core 3 are 104.

I wouldn't use any calibration factors on core 0 and core 1 and then I'd use some calibration factors on core 2 and core 3 so all 4 cores report the same temperature when you are idle at your normal overclocked settings. Let's see your calibration settings so I can have a better look.

[OC Nub]

Coming from the RealTemp guy this is going to sound a little dumb but temperatures really aren't that important. Cool them as much as you can afford to and they will be able to overclock more. If they are not stable, cool 'em some more or back off your overclock. That's really all there is to it. No fancy temperature monitoring program needed.

:

My temps on that e8400 CO were read at 20c all the time no matter what the vcore was set to. The temps would go up under full load running Prime 1.256v bios at 4.05ghz, it would show 38 to low 40's c. I guess there is no way to know actual temps? Its a great clocking chip but never pushed it any further due to the sensor issue. Back when the chips came out people were saying to add 15c to what the temp program was reading but I doubt that is accurate. I put the chip back in a few weeks ago in hopes that the newer versions of real temp or core temp would read it correctly but no luck. Guess the sensors are bad.

[-X-hellfire]

@unclewebb, look at the 2a image/post and you will find the calibration settings as well, well it was a long post so it's easy to overlook it ans scroll past it. :innocent:

I have downloaded prime95 and used it in the test above but when it runs it only uses 1 core in the RealTemp "Test Sensor" test. Is it really that much difference between prime95 and prime2004 when testing the "small FFT" ?

I cropped the 1280x960 images to 800x960 as it is easier on the weblayout on forums and not much more was seen except that that it showed 1 x prime95 and 3 x prime 2004 at the clock, so the names of the images are abit misleading as "4x(99)-P95" means 3 x Prime2004 at (99-100%) and 1 x Prime95 at (99-100%) taken a "Print Screen" at the "Heat Up Phase" as close to 100% as possible but it then shows 99% so it's pretty close anyway.

Have only tried the C1E before but it was unstable and crashed games like Crysis and I haven't tried the EIST which was in the bios when they updated my bios from F6 to F7 when then motherboard was on RMA. If EIST works better then I will try it later on.

Well if you look in post/image 2a you'll see that the in RealTemp 33C 33C 33C 33C, and core 0 and core 1 showed to low idle temp at 0.0 calibration ( 32C ) and had to be turned up a notch to show 33C, compared to T-Balancer cpu-D1 digital sensor even that I let it run for a while to see if it stabilized it was the closets to 33.5C as possible. It is of course possible that cpu-D1 gets influenced by other heatsources at idle as the CrazyCool surrounds the bottom of the TRUE and memory blows by it. There are many possibilities for temperature errors when the probe is not drilled into the heatsink etc.

[unclewebb]

Sorry -X-hellfire, too much information. I missed the calibration settings before.

The newer versions of Prime95 let you run as many threads as you like.



It makes life easier when testing. Just set it to 4 and Small FFTs and you're done. You'll have a full load for testing.

I'm using version 25.9 build 4
http://mersenne.org/freesoft/

There are many possibilities for temperature errors when the probe is not drilled into the heatsink etc.

That's always the problem with this. Every digital sensor has a margin of error, even the IR temperature gun I use. Where exactly is your cpu-D1 sensor located? How much is it offset from the center of the CPU and what sort of calibration check if any did you do to it?

The original calibration formula I came up with when I first started working on RealTemp would have worked better for your CPU. Based on Intel's presentation last year at IDF, I changed my calibration formula and simplified it but the end result was not an improvement in accuracy. I assumed the information they were presenting was accurate but I assumed wrong. I was expecting an engineering type presentation based on real data and facts when all it turned out to be was more of a PR presentation to introduce the upcoming Core i7. Unfortunately, a little guy like me can't argue with a big company like Intel. When they said, "This is how it is", I had to nod my head in agreement like a chump. Listening to the guys in the blue coats and adjusting my calibration fomula based on that was a mistake that I regret.

Based on testing and not fluff presentations, I now understand that TJMax is not a fixed value across the 4 cores in a Quad. I'm going to add another variable to my calibration formula so it better represents the testing that rge and I have done and your testing as well. Your Core 2 is out significantly and needs a better formula than what RealTemp is presently using. Stay tuned.

Can you try running a single CPU Cool Down Test using just Prime95 Small FFTs with 4 threads selected? Adjust your MHz and core voltage and fans so it maxes out at around 65C if possible. Turn off Task Manager and all other monitoring programs to reduce any interference to as little as possible. This will provide me with the data I need to adjust my calibration formula to better track your 4 cores.

Edit: Your APIC ID 0132 means that most software is going to reverse the data coming from core 2 and core 3. Compare RealTemp to another monitoring program and you'll see what I mean. No worries, RealTemp gets this right, even if you re-boot and your APIC ID gets shuffled to something new. RealTemp will always correctly report that core 2 is sagging the most.

[-X-hellfire]

Yeah I know unclewebb, too much information often results in tl;dr;

As my methodology of the test was "Print Screen" to capture as much realtime data from various sources, it was a tedious process, and with the prime2004 it probably introduced some other error sources as well

The prime95 I had previously download in the "Sensor test" window in RealTemp was an older prime95 ver 24.4 which didn't have the "Number of threads" field. Now I got the same you got, prime95 ver 25.9 and it worked like a breeze now. Thanks for the link unclewebb.

Will do more testing later on, got to sleep know been awake for 24+ hours, and nothing good comes out while working with computer too long and to tired.

Willl post pics later on with the cpu-D1 location and answer the other questions

Goodnight

[stasio]

Real Temp 3.30 RC6 is available for download

[wutsup]

hey this is my firs tpost on this forum, and i have a questoin. i ahve my q6600 at 3.4ghz (425x8) and when i click on sensor test( i try to make sure my comp is at idle) i get 13,12,11,7 my last core is always 5-6 lower than the first. what does that mean?

[unclewebb]

what does that mean?

I don't know.

That initial Sensor Movement test was more designed for the 45nm chips that had sensors that didn't move at all. A reading of zero is a bad thing. At least all your sensors are moving. When one sensor moves a lot less than the other 3 it's possible that sensor is getting stuck at idle but that doesn't happen very often with the 65nm CPUs like your Q6600.

The best thing to do with your Quad is run a full CPU Cool Down Test and post your results. That gives me more information to go by. When you run this test make sure that Windows has had a chance to settle down first and go have a coffee for 10 minutes while the test if running. No surfing the net while testing is in progress.

[wutsup]

here you go webb

[unclewebb]

wutsup: I've been using a Q6600 - G0 for the last few months and my best guess is that Intel offsets the TJMax for core 2 and core 3. Your CPU seems to show the same thing. We had a discussion here just a little while ago and came to the conclusion that if this is true, Intel likely did this so all 4 cores wouldn't thermal throttle at the same time. If they set core 2 and core 3 slightly higher, core 0 and core 1 would thermal throttle first and if that didn't fix the problem then core 2 and core 3 would thermal throttle about 5C later.

I set TJMax = 100C, 100C, 105C, 105C on my Q6600 and I would suggest you try the same thing. You'll be surprised when you see how much better your 4 cores track each other now from idle to full load. Start and stop a program like Prime95 a handful of times using the Small FFTs option and watch how the temps track each other. I prefer this program when testing because of its ability to put an equal load on each core. LinX creates more heat but is not as consistent.

[wutsup]

so should i just set the tjmax for the last 2 cores to 105 instead of the default of 100 and call it a day?

p.s. my temps are high because its hot in my room (80degrees +)