Ridiculous core i7 temps?!?

[bmbvenom]

I have a major problem.
My 920 is (Dummy) OC to 3.33 (160x21), HT on, Turbo on.
I have a Noctua NH-U12P SE1366, using the Noctua thermal paste.
While running p95 (8 instances) my core temps in RealTemp are super high (94,99,85,100), but the display on my EVGA x58 mobo says 83.
Then when I'm at idle RealTemp reports 46,48,46,47 and my mobo says 33.
I know if what RealTemp says is correct I'm in big trouble, but that seems hard to believe with the cooler I have.

What am I missing??

 

[Ancient_1]

I would guess you have a bad mount of some sort causing those temps. With that cooler and that OC you should have much better temps.

The differences in temp readings you are getting is that your evga display is reading tcase and realtemp is reading you core temps.

I just re-did my TRUE and got a good 8°+ drop in temps.

Old

New

I just did a sensor test with realtemp and here are my results, I am showing two shots since I keep all power saving features enabled and need to load the cpu to show the correct speed. With my settings I idle at 1.136v and load between 1.36 and 1.375.

 

[CompuTronix]

I have a Noctua NH-U12P SE1366, using the Noctua thermal paste. I know if what RealTemp says is correct I'm in big trouble, but that seems hard to believe with the cooler I have.

What am I missing??

Your Noctua NH-U12P is perhaps somewhat less robust than you might think. It currently ranks #43 on Frosty Tech's cooler tables - http://www.frostytech.com/articleview.cfm?articleid=2383&page=5

 

[Ancient_1]

Your Noctua NH-U12P is perhaps somewhat less robust than you might think. It currently ranks #43 on Frosty Tech's cooler tables - http://www.frostytech.com/articleview.cfm?articleid=2383&page=5

From everything I have seen that Noctua is a very good HS and about as good as it gets with quiet fans (and it comes with 2 well matched fans with that model).

The problem I have with frosty is that they use a die sim and not a real cpu and will have a much more even temp across the contact area than a real cpu which will have hot and cool spots. Also they don't test with a common fan and the stock fan which would give a nice apples to apples comparison when possible to see which HS is best rather than the combination of HS and fan.

Here is a test of a lot of the top i7 coolers that I do like. It shows performance with the stock fans at different settings and with a common fan for the top models. It also shows the max OC achieved with each (too many only go a moderate OC) with different fans and speeds equal on each.

There are a couple other reviews that aren't bad from Mad Shrimps and bit-tech that show similar results, there are many more that all show the Noctua in much better light than Frosty.

 

[bmbvenom]

OK so I re-seated my HSF, found out that the thermal paste hadn't spread itself out very well (hence the two hotter cores) and things cooled down a little bit. However, it still seems warm compared to what others are reporting.
Any other ideas?

 

[one-shot]

From everything I have seen that Noctua is a very good HS and about as good as it gets with quiet fans (and it comes with 2 well matched fans with that model).

The problem I have with frosty is that they use a die sim and not a real cpu and will have a much more even temp across the contact area than a real cpu which will have hot and cool spots. Also they don't test with a common fan and the stock fan which would give a nice apples to apples comparison when possible to see which HS is best rather than the combination of HS and fan.

Here is a test of a lot of the top i7 coolers that I do like. It shows performance with the stock fans at different settings and with a common fan for the top models. It also shows the max OC achieved with each (too many only go a moderate OC) with different fans and speeds equal on each.

There are a couple other reviews that aren't bad from Mad Shrimps and bit-tech that show similar results, there are many more that all show the Noctua in much better light than Frosty.

I wonder why they didn't test the Xigmatek HDT-S1283. According to Frosty Tech, the Xiggy beat the TRUE in a number of tests. I guess I'll put the Xiggy on the i7 920 when I get it later this week.

 

[Ancient_1]

They did (the dark knight version) here but it was more recent and not included in that group.

 

[one-shot]

They did (the dark knight version) here but it was more recent and not included in that group.

The Xigmatek HDT-S1283 does perform a little better than the Dark Knight. The Dark Knight does include the retention bracket, but the bracket only costs a few dollars. I want to see one of the original S1283. According to the review you listed, it is stated that the Nickel layer on the Dark Knight is about 4 degrees more inefficient than coolers without the coating.

 

[CompuTronix]

That is my understanding as well. The more recent coolers with direct touch heat pipes and coolers which feature push-pull fans have extended the perform envelope once again. It wasn't long ago when the selectiion of high-end air coolers was much more limited. The technology has cetainly come a long way.

 

[micky_lund]

http://www.legitreviews.com/article/880/12/
noctua has great temps here... dunno bout ur situation...

 

[bmbvenom]

I feel like such a n00b....
I looked up a diagram of the core i7 and found this on pcstats:

So I realized that putting a 'pea-sized' dot thermal paste in the middle wasn't going to be enough.
I took my computer apart (yet again) and reapplied the thermal paste in a line that covered all four of the cores and my temps went down by over 25 degrees Celcius!

Here's the proof

 

[one-shot]

Good job and nice temps!

 

[micky_lund]

so what happened? u used more thermal paste, and it worked?

 

[bmbvenom]

Try reading my post again, I explained exactly what I did: Not more thermal paste, just in the right spot.

 

[micky_lund]

how much of the bottle did u use all up? in all 3 re-applications?

 

[bmbvenom]

Not too much, probably about the size of half a pencil eraser per application; so maybe around a third of the tube? As I stated before, it's not about putting on gobs of thermal paste, it's about having it in the right area so that it conducts the heat more efficiently.

 

[micky_lund]

yeah... i know (now, thanks to u ) so u just put it more to the 'top' side of the chip, where the cores are situated, and it made that much difference?? thats sweet

 

[stan116]

What are your under load temps ? I own a hardware review site, and we reviewed this cooler, on a core i7 920 at 4Ghz.

Temps were at at 75c under load with the Noctua, 73c with the Dark Knight, 72c with the Thors Hammer, and 74c with the Akasa Nero.

The Xigmatek Coolers , and Akasa Nero at idle were at 40c and the Noxtua at 43c. What was nice about the Noctua, is that even at high speed you could just about hear the fans.

 

[bmbvenom]

The highest any of the cores got under an hour of blend torture testing, courtesy of prime 95, was 68c, but they stayed pretty consistently at 65c or under.
I hope this helps you out somehow.

 

[-713king-]

why dont you show us how thinly the line should be because i want to reduce my temps as well..

 

[bmbvenom]

Sure thing, I'll even make you an animated picture

I looked and found the diagram of the Core i7 chip (as shown in one of my previous posts), then I found a real picture of a Core i7 chip and just made it so the yellow triangle matched. The picture also shows about how much thermal paste I put on (maybe a little more, but not too much). Just click on the link and download the swf file. Sorry but I can't figure out how to directly embed from google docs.

https://docs.google.com/a/biola.edu/leaf?id=0BwCfXPhzhJVqNTNlZDE4N2YtYWM4Yy00MDM4LWFjNWMtMjE2YmUwNjQ2ZmNk&hl=en

 

[-713king-]

ahh nice i will do that to my CPU then.. when removing paste can i use alcohol 91 percent??? do you think i should switch out my Dark Knight fan to a Noctua fan??? to have much cooler and better performance or just leave it how it is..

 

[-713king-]

this works wonders!! i was running at 45c and up now im running at 35c with HT on and once it settles i think it should lower down the temps?? if not 35c is fine! lol thnx bmbvenom

 

[cboath]

What happened to the SWF file? It's just text right now and it's only been 1.5 weeks...I'd really like to see it because I think I have the exact same problem.

 

[DragonQ]

I used the same method (line across the cores) yesterday and it didn't improve my temps at all (I used the blob method at first).