Discussion in 'RealTemp' started by unclewebb, Jun 28, 2008.
Nehalem is hot like hell. DTS readings don't lie. At least not in full-load.
But dont you think that the idle temps are too hot? And why do you think the movement is so high?
burebista is right again. The Core i7 does run hot as hell. The sensors on these chips are far better than the previous 45nm Core 2 generation. Here's an example of one that is water cooled.
He also has Sensor Movement in the mid twenties and maxes out in the 70C range. Depending on room temperature, etc., your results don't look a lot different than his.
During the Sensor Movement test, a well overclocked Core i7 increases in temperature probably twice as much as the average, well overclocked, Core 2. That's a pretty good sign of just how much heat these things put out. You could get a lot of performance out of Core 2 with air cooling but if you're overclocking Core i7, water cooling isn't really an option anymore. With these chips, you need all the cooling you can get. [H]ard did a good heat test and video when these things first came out and I think they were using the same cooling that you're using.
That puts my mind at ease thanks, I'll just have to work on improving the cooling to go any hogher than I am at.
I have one question: why does Intel TAT heat up to 82/83C but running Prime95 for hours only leads to 68C temps?
I know that this has nothing to do with the topic I'm in. sorry for that.
Try prime95 Small FTTs.
Anyway, TAT, Linpack, Core Damage have another methodology to heat up CPU. Unfortunately I don't know what it is but frankly in daily use prime 95 Blend Test is very close to temperatures I see in games.
BTW, for all-in-one heat-up package give it a try: OCCT 3.0.0. beta 16.
Lol @ OCCT CPU too hot !
The temp hit 79-80C and OCCT stopped. haha
hehe. you have to change the default of 80c so that it doesn't cancel when the temps hit 80c. I tought that when the Cpu is detected that it automaticly changes the setting depending on the detected CPU.
Slightly off topic, but real temp works fine in Windows 7!
After a couple of months of waiting unclewebb managed to wrote a beta (for now) RealTemp plugin for RivaTuner.
You can get it there.
Works with Nehalem, Quad's, Core2Duo's and it's a big step forward from "old" Core2Duo plugin.
Try it and report if something is wrong so uncle can fix it.
Quick question, i have my Realtemp all calibrated correctly, is the correct tjmax for a Q9450 95 or 100?
TJMax is not a fixed, written in stone number. Actual TJMax will vary slightly from one CPU to another even if they have the exact same model number.
The values that Intel released last year are called TJ Target numbers. Intel says that their manufacturing process and quality of sensors they use are not accurate enough to be able to set TJMax across all processors to one number. Actual TJMax should be equal to or a little higher than Intel's TJ Target number.
Intel's TJ Target for a Q9450 is 100C but I've seen ones where actual TJMax might be closer to 110C on some cores. The best thing to do is a CPU Cool Down Test. That will show you how your 4 cores compare to each other. On a 45nm Quad, it's very rare to have 4 cores that mirror each other. They're usually all over the place.
Post a pic of your cool down results and I'll tell you what I see.
With a little luck maybe we'll see 2 cores grouping toghether and of course with different TJMax for each group.
That's why I keep tell to rely on distance to TJMax, in full-load is the most accurate number for your CPU. Keep that distance >20 and forget about it.
so this IS normal?
Go back a couple of posts and download a newer version of RealTemp. Do a Cpu Cool Down Test and post your results. The sensors on your CPU are probably a bit screwed up which unfortunately is quite normal. Your Q6600 looks a lot like my Q6600 where cores 2 and 3 seem to have a TJMax 5C higher than core 0 and core 1.
Intel says the TJ Target is 90C for a Q6600 - G0 but I think my Q6600 is closer to 100C / 105C.
yeah it does seem like its a bit screwed up huh?
don't know which temp to go by
I don't have time at the moment for a full analysis but just try what I do.
Set core0/core1 to TJMax 100C and set core2/core3 to TJMax=105C. It sounds a little crazy but after you do this you'll be surprised how well your 4 cores track each other from idle to TJMax. Start and stop Prime95 Small FFTs a few times and watch how they track.
There is a slight difference in slope error between the two sets of cores but most of the difference is that TJMax is not the same for all 4 of your cores. Send me an e-mail if you want a more thorough analysis when I get the chance. If it's in my e-mail box then sooner or later I'll get to it:
Your Q6600 is similar to the one I'm presently using.
Edit: You might also want to try 106 for core2/core3. Some of the 45nm Quads have cores that are closer to 110C.
My bet now.
Put 107 for Core2/3 and try Core Damage. Let's see who's close, me or uncle.
results for 105
107 w/ core damage
Does anyone have an Intel Core CPU and like to game and do you use RivaTuner?
I just added a stand alone RealTemp plug-in which you can use in RivaTuner. Download RealTemp here:
Go to the Settings window in RealTemp and click on the RivaTuner button to install the plug-in and then start up RivaTuner. RealTemp does not need to be running after this plug-in has been installed.
RivaTuner will allow you to see some new numbers on your G15 and you can graph data for CPU MHz, CPU Load, CPU core temperature and Distance to TJMax.
Xiphos: I kind of like 106. Try running Prime95 Small FFTs. I find it's more consistent across all 4 cores.
to get it even across the board I had to go 100/100/108/108
high temps is with IBT, core damage is a nifty little program, but IBT is still king of torture tests
106 small FFTs
Man, you're fooling yourself with temps. If you take a look distance to TJMax is the same, even with different TJMax, so you can put TJMax 80 if you like to have 30C in IBT full-load. IMO keep an eye on distance to TJMax and forget about TJMax value.
Quads consist of two separate Dual Core processors so it's possible that at full load they might run a degree or two different but very unlikely that they run significantly different. I've seen so many screen shots of Q6600 CPUs that my only explanation is that TJMax is not always the same for all 4 cores. Your temperatures look a lot more believable across the board when you use slightly different TJMax values.
The other thing I've seen with 65nm is 99% of the time, core 2 and core 3 report lower temperatures. If this was all random chance you would expect to see this only about half the time. I don't see anything in the design to explain this.
burebista: How's the new RivaTuner plug-in working? Was it easy to install? Remember to say Yes to encourage other users to give it a try!
You have a full report on PM on XS.
I took a couple cooldown tests on two other CPUs that i have, a Q9300, and a diffrent Q9450 pics are attached. - Note that i dont really care what temps these two chips are running at this is just for information.
I will be running the cooldown test on my 9450 later this evening and will compare with these results.
This pic is of my Q9450
Separate names with a comma.