Can you help my temps?

[kohd]

i7 920 @ 3.42 (Did not increase CPU voltage. Just bus speed from 133 to 163 if this matters at all.)
Coolermaster HAF 932
Thermalright Ultra-120 with a Scythe S-Flex fan
Room temp ~75f

Asus P6T Deluxe V2 Bios. Speedstep enabled. Turbo fans enabled for both CPU & ChassisX3. (supposed to boost fan speeds when temps get higher)

Everest IDLE results:
Core1 = 42c
Core2 = 42c
Core3 = 38c
Core4 = 38c
CPU = 31c

Everest LOAD results: (10m of Prime95 blending test)
Core1 = 71c
Core2 = 71c
Core3 = 70c
Core4 = 70c
CPU = 61c

These temps seem awfully high. As I said, I have the turbo fan speed enabled for the CPU and the 3 chassis fans bios listed. However, during my testing the temps either didn't go high enough to enable the fan changes or something with the bios turbo controller is in-op because the RPMs stayed the same. It seems I have no ability to change the fan speed manually in bios either.

Fan speeds: (constant)
CPU = ~1560RPM
Ch1 = ~740RPM
Ch2 = ~775RPM
Ch3 = ~1110RPM

Any suggestions? I've been reading posts about these issues ever since I built this PC a week ago. It's really hard to use another person's issues though because temps seem to change sporadically depending on chipset and system build (some people report low 20s CPU temp but 40-50s core temps at idle) One post is in agreement that 70c CPU temp is ok, another says anything over ~65c is bad. Some refuse to distinguish between core and CPU temps >< All that considered, I chose to make my own special post.

Do I need to re-seat my HS and do a better job with the paste?

 

[Shadow703793]

Do I need to re-seat my HS and do a better job with the paste?

Yes. Re seat and put in new TIM.

Also, could you give temps in RealTemp? EVEREST is known to be inaccurate sometimes. Also, which S-Flex fan did you get? Remember, the S-Flex varay from ~33.5-75CFM depending on which S Flex you got.

70C is with in Intel's specs, but most agree that anything above 65C (Core) is too much.

 

[kohd]

Do I need to re-seat my HS and do a better job with the paste?

Yes. Re seat and put in new TIM.

Also, could you give temps in RealTemp? EVEREST is known to be inaccurate sometimes. Also, which S-Flex fan did you get? Remember, the S-Flex varay from ~33.5-75CFM depending on which S Flex you got.

70C is with in Intel's specs, but most agree that anything above 65C (Core) is too much.

RealTemp IDLE: (new to this program...can't find the CPU temp ;P)
Core1 = 41c
Core2 = 38c
Core3 = 41c
Core4 = 38c

Realtemp LOAD (10 minutes of Prime95)
Core1 = 71c
Core2 = 69c
Core3 = 68c
Core4 = 68c

I am using the S-Flex F model (63.7CFM) Apparently 1560RPM is max and should produce ~60-64CFM which should be perfectly fine I'd think on a good HS.

Oh one last thing...if I do re-seat the HSF with a better layer of Arctic 5, is it very important to let the computer rest like I've seen mentioned? (i.e. shut down each night for a week or so as opposed to leaving it on) I ask because it's kind of hard for me to be able to do that without another computer to use for work (gets used by 2 shifts of people ;P). Gave my old machine to the parents ><

 

[Shadow703793]

^Some thing's not right with your mounting. Re seat and apply a very thin layer of TIM with your finger covered in cling wrap. Or just put a pea sized drop on the center of CPU.

As far as shutting down goes, Down it for about 5-10 minutes at night should be fine. No need to down it for an entire night.

 

[Conumdrum]

A small pea, bigger than a grain of rice.

 

[kohd]

Just got finished redoing the HSF. Not sure I could do it any better than I did this time tbh. Got rid of the old paste with alcohol and a glasses cleaning cloth...1 small drop dead center on the CPU etc. Temps are close to what they were before.

IDLE with Everest
CPU = 30c
Core1 = 41
Core2 = 41
Core3 = 38
Core4 = 38

IDLE in Real Temp
41c
38c
40c
38c

LOAD with Everest Prime95 10m
CPU = 59c
Core1 = 70c
Core2 = 70c
Core3 = 68c
Core4 = 68c

LOAD with Real Temp Prime95 10m
Core1 = 70c
Core2 = 68c
Core3 = 68c
Core4 = 68c

I hear temps should drop between 2-5c over the course of the next week following a newly fixed HSF but...even if I got that isn't this still too high?

 

[Conumdrum]

Very good temps. Your fine. It's an i7, they run hot.

 

[TexasRattleSnake]

did you increase your uncore/QPI voltage a bit? what is your memory running at?
if you haven't changed the uncore or QPI voltages then these temperatures are a bit high (off by 4-5C), otherwise they are just fine.

 

[geofelt]

Go to the arctic silver web site, they have a pictorial tutorial on their recommendations on applying as-5 to a i7 cpu.
I don't think it makes much difference. The common mistake is to apply too much material so that it acts like an insulator.

I rarely used many of the possible 8 threads, so I turned off hyperthreading. The result was that I could OC higher and run cooler.
If your situation is the same, you might try it.

The true ultra 120 is somewhat noted for having an uneven base. You might want to consider lapping it. Take the cooler off and clean off the residual material. Put a very small drop of water on the cooler and press a piece of glass to it. Look at how the water spreads out. If it spreads out evenly, and the suction alone almost lets you suspend the cooler, then it is flat and true. If not, consider lapping it to flatness.

 

[orangegator]

Very good temps. Your fine. It's an i7, they run hot.

+1

Your hsf mounting is fine. If you want to lower your temps, you will need to manually tweak and lower your voltages to find the minimum needed for stability. Also, many people run the i7 at 80C+ load and have no problems.

 

[kohd]

I have lapped the HS, although maybe I didn't do a good enough job? My base sure didn't look all mirrored and pretty like the pictures I saw ;P (it did reflect well and it did create a good suction with my sheet of glass+water...but part of it was a copper color and part of it was silver) I can say that I spent a good 2 hours with that derned thing and I followed the instructions I found here on Tom's as close as I could.

Anyway, I really appreciate the replies. I don't intend on OCing any further for the moment. So as I stand now it seems that the temps I have are ok. If I plan on upping the voltage manually I will definitely look into lapping the HS again.

And to answer the question whether I increased the voltage or not, I did not manually change anything except the base clock to get 3.42. However, my bios likes to do things on it's own with it's auto settings which seem to make things fluctuate some under load vs idle. CPU-Z shows my core voltage @ 1.24 when idle and 1.28 under load. This is most likely due to the turbo setting in bios (ups the CPU an extra 200ish MHz under load...probably set to increase the voltage 1 click?)

 

[geofelt]

I have lapped the HS, although maybe I didn't do a good enough job? My base sure didn't look all mirrored and pretty like the pictures I saw ;P (it did reflect well and it did create a good suction with my sheet of glass+water...but part of it was a copper color and part of it was silver) I can say that I spent a good 2 hours with that derned thing and I followed the instructions I found here on Tom's as close as I could.

Anyway, I really appreciate the replies. I don't intend on OCing any further for the moment. So as I stand now it seems that the temps I have are ok. If I plan on upping the voltage manually I will definitely look into lapping the HS again.

And to answer the question whether I increased the voltage or not, I did not manually change anything except the base clock to get 3.42. However, my bios likes to do things on it's own with it's auto settings which seem to make things fluctuate some under load vs idle. CPU-Z shows my core voltage @ 1.24 when idle and 1.28 under load. This is most likely due to the turbo setting in bios (ups the CPU an extra 200ish MHz under load...probably set to increase the voltage 1 click?)

Part copper and part silver is a good indication that the sample was not flat to begin with. You most certainly helped by flattening things out. A mirror shine is not important, the flatness is. Good job.