Page 1 of 1

Strangely High CPU Temps

Posted: Tue Jan 16, 2007 9:52 am
by Ravenlark
So I just put together a system I thought would be pretty quiet and cool despite the Pentium D 820 (I built on a budget). The specs are in my signature.

It's not.

The idle temps are 55c, the temp under 50% load goes to 62-63c. If I set the exhaust fan to low instead of medium it foes up another few degrees.

This seems way higher then other peoples temperatures. I'm getting the readings from Speedfan BTW.

The only thing I can think of that might be effecting the cooling power of the Ninja is that the MCP fan in blowing through the MCP and into the base of the Ninja, but with a Push / Pull fan setup, and the PSU fan sucking in air right next to the Ninja as well....I can't see this raising my temps by 10-15c.

Unfortunately I'm at work, so I can't post pictures and give you a better idea what I am talking about it terms of setup, but any preliminary thoughts? My assumption is that with a fan cooled Ninja and a decent case I should be seeing idle CPU temps of 35c or so and 100% load in the low 50c range. Is this accurate?

*EDIT* Speedfan reports the case temp as 55-57c pretty consistently regardless of CPU load. Side panel on or off did not effect the reading.

*EDIT2* Just remoted into the system now; apparently the case temp goes down to about 47c after a while at no CPU load.

Posted: Tue Jan 16, 2007 1:09 pm
by Plissken
First I would confirm the temperatures with some method other than Speedfan, to make sure you have a problem.

Yes, you should be in the 30s idle. You are evacuating plenty of air, since with the side panel removed the temps are the same. I suspect bad contact between heatsink and CPU. After that, it's the MCP fan I guess... don't know.

Posted: Tue Jan 16, 2007 5:20 pm
by psiu
What kind of temps does the BIOS report? And in your edits, you say the **case** temp is 55-57 and 47C? Do you mean CPU temp?

Because if your case temp is that high :shock:

Replies

Posted: Wed Jan 17, 2007 8:02 am
by Ravenlark
You know, I haven't verified the numbers in the BIOS. I'll check when I get back home.

By case temp I mean the System temp reported by Speedfan. I'm not sure if thats supposed to indicate the air temperature in the case (which was my assumption) or something else.

Whatever its reading must be close to the CPU or MCP though, because if I stick my hand in the case near the mobo, it's only slightly warmer than room temp, except right around the video card, MCP and base of the Ninja.

I turned off the MCP fan last night, and it did drop the temps a little, but I am still getting CPU temps of 40c idle and 50-55c under less than 50% load.

I'm loath to take the whole assembly apart to check the heatsink contact (the ninja mounting on the 775 slot is a pain and requires me to remove the whole mobo). That being said, if the BIOS backs up the temperature readings and no one else has any suggestions, better to hope that poor contact is the answer rather than manufacturing defects in my Ninja or CPU.

Further Tests

Posted: Mon Jan 22, 2007 11:00 am
by Ravenlark
So I reseated my Ninja with a fresh coat of Arctic Silver...no change.

So I decided to move on. I am putting together another system with exactly the same parts for my fiance...this one though is idle at 34c and only 51c at 50% load.

I went back to the first system and verified the CPU temps with a few utilities.

This means (to me anyway) that either one of two things is happening:

1. I have a defective CPU that is generating way too much heat

2. The cpu temp sensor on the one board is defective and giving numbers 15c higher than they are.

Is there any way to find out if a CPU temperature sensor is defective?

Posted: Mon Jan 22, 2007 12:11 pm
by Plissken
1. I have a defective CPU that is generating way too much heat

2. The cpu temp sensor on the one board is defective and giving numbers 15c higher than they are.
3. One CPU surface is MUCH less flat than the other one, but 15C seems like too much difference for this excuse.

4. One Ninja is defective. You never know.

One thing to try, to help narrow it down: since you have (2) exact same systems, touch the heatpipes at the base of the Ninja. You should be able to tell the difference if one is kicking out 15C more. Other things to try are to swap Ninjas to the other CPU, to rule out a defective Ninja, or to swap CPUs to the other mobo. With identical systems, eventually you'll know the answer, by process of elimination.

AFAIK, the only way to see if you have a defective sensor is directly measure with a thermocouple... or rule everything else out.

thermocouple

Posted: Mon Jan 22, 2007 12:51 pm
by Ravenlark
I don't have a thermocouple, so I guess process of elimination it is.

I didn't think of the Ninja; I just blithely assumed it wouldn't work at all if it was defective. Makes sense in hindsight.

Might not get a chance to dissemble everything this week (having to take the whole mobo out to get at the Ninja is a pain), but I'll try and post my results when I get them.

Thanks for the input.

BIOS

Posted: Wed Jan 24, 2007 9:41 am
by Ravenlark
I found the issue.

The latest BIOS for the EVGA 680i (p24) has an issue that causes the temps to report about 10c higher to programs. If you go into the BIOS and monitor the temps there its fine.

I hadn't updated the BIOS on the second box, which is why there was a discrepancy.

Check the EVGA forums for details.