Server CPU Temp at 87c!

[CHR1S]

Tell me the HP monitoring software is pants and its not really that high!

 

System Overheating (Zone 2, Location CPU, Temperature 87C)

 

Current temp with a portable aircon unit sitting by its side

 

CPU 66 C

 

Hmmmmm

[danrhodes]

Take a more accurate reading using Core Temp but I think you will find the HP one to be quite accurate.

 

D

[CHR1S]

Oh, ******!!

 

Just what I need!

[danrhodes]

Oh, ******!!

 

Just what I need!

 

I guess it was accurate then, that's pretty warm, what processor is it, whats the max it can support?

[CHR1S]

Gives the "This Intel processor is not supported" its a Xeon.... hmmm

[danrhodes]

Xeon what, there are hundered in the Xeon range!

 

Try http://www.techpowerup.com/realtemp/ its also very accurate.

 

D

[Hightower]

Oh, ******!!

 

Just what I need!

 

Thing is though, you really can't trust modern software these days to do a manual job. Most accurate way is to power down the server (very important), take off the heatsink and fan, and lick the CPU. If you feel a burning sensation on your tongue is probably is too hot!

[danrhodes]

Thing is though, you really can't trust modern software these days to do a manual job. Most accurate way is to power down the server (very important), take off the heatsink and fan, and lick the CPU. If you feel a burning sensation on your tongue is probably is too hot!

 

Is this a BECTA approved method for checking core temps? LOL

[Hightower]

Is this a BECTA approved method for checking core temps? LOL

 

What's BECTA? (Seriously, what is BECTA now the government plan to scrap it hehe)

[tmcd35]

Thing is though, you really can't trust modern software these days to do a manual job. Most accurate way is to power down the server (very important), take off the heatsink and fan, and lick the CPU. If you feel a burning sensation on your tongue is probably is too hot!

 

You'll be hearing from my solicitors once the swelling on my tongue has subsided enough for me to tell them what happend

[CHR1S]

3.4Ghz Irwindale

Minimum/Maximum operating temperature (°C) ? 5 - 73 5 - 79

 

Oh dear! Weird thing is tho, its a 3.4Ghz CPU but its running at 2.8

[danrhodes]

Oh dear! Weird thing is tho, its a 3.4Ghz CPU but its running at 2.8

 

That's more than likely Enhanced Intel® Speedstep Technology in play there, it will lower the FSB and voltage of the chip to reduce power usage.

 

64-bit Intel® Xeon® Processor 3.40E GHz, 2M Cache, 800 MHz FSB with SPEC Code(s) SL7ZD, SL7ZK, SL8P4

 

D

[CHR1S]

Ahh, cool, well I wish it was. What you think then, dodgy sensor or something iffy?

[Hightower]

You'll be hearing from my solicitors once the swelling on my tongue has subsided enough for me to tell them what happend

 

Please ensure you take photo's of your injuries. I doubt they'll help much in your legal case, but if you post them up here we'll be sure to get a laugh out of them =P

[danrhodes]

Ahh, cool, well I wish it was. What you think then, dodgy sensor or something iffy?

 

What is real temp reporting? If your core temps are still high then you may want to look into a better cooling solution for your server room.

 

D

[CHR1S]

It wont run for that Xeon chip

[srochford]

Does sound a bit high :-)

 

We have a threshold temp of 35C; above that alarms go off and people fix things ASAP (I know that there was a problem last night, for example)

 

You need to get some of this: TROX | Data Centres

 

The liquid CO2 cooling basically puts an extra "door" behind your server racks. This contains the liquid CO2 coolant and so air drawn through it is cooled. It's much more efficient than blowing air randomly into a server room and hoping some of it gets to the servers!

[SYNACK]

Assuming the sensors are correct this is probably the break down of the thermal compound between the CPU and its heatsink. Over time and with moisture or contaminants this compound does break down and no longer conducts heat properly to the heatsink.

 

First remove the old compound (Arctic Silver do a great cleaning product ArctiClean by Arctic Silver) and then apply some new high performance thermal compound to the CPU. This should better thermally link the CPU and heatsink allowing the fans and cold air to cool it properly.

[CHR1S]

Does sound a bit high :-)

 

We have a threshold temp of 35C; above that alarms go off and people fix things ASAP (I know that there was a problem last night, for example)

 

You need to get some of this: TROX | Data Centres

 

The liquid CO2 cooling basically puts an extra "door" behind your server racks. This contains the liquid CO2 coolant and so air drawn through it is cooled. It's much more efficient than blowing air randomly into a server room and hoping some of it gets to the servers!

 

Server Racks? The servers sit in a store room, no aircon .....

 

 

Assuming the sensors are correct this is probably the break down of the thermal compound between the CPU and its heatsink. Over time and with moisture or contaminants this compound does break down and no longer conducts heat properly to the heatsink.

 

First remove the old compound (Arctic Silver do a great cleaning product ArctiClean by Arctic Silver) and then apply some new high performance thermal compound to the CPU. This should better thermally link the CPU and heatsink allowing the fans and cold air to cool it properly.

 

The heatsink is red hot to the touch, I just think the ambient temperature for the room is just too much for the type of server, with it being a rackmount (minus the rack and proper cooling)

[Butuz]

What is the ambient temperatuture in your room? If its 30 or more then yes you just need to get air con!!

 

If ambient is low (under 25) it could indicate a problem with the cpu fan / heatsink itself.

 

Butuz

[CHR1S]

Ambient temperature is 20c, server temps are as follows -

 

1 IO Board 34 C

2 CPU 63 C

4 Power Supply 25

5 System 24 C

[Mako]

Echoing SYNACK's suggestion here. Try applying some new thermal paste. You need to remove possible problems before isolating the fault. You'd be suprised at the heat build up caused by crumbly old thermal solution, especially the stock stuff. Clean it off and throw some Arctic Silver on it, or equivalent, and see if it makes a difference. Your other temperatures seem to be within reasonable operating ranges, so I would look at the heatsink/cpu contact first. If there's more than one server in that room, and no others are throwing up overheating errors, then it points to the individual machine's cooling efficiency. Ambient temperature is fine.

[CHR1S]

Ok, will give the thermal compound a go and report back.

 

Thanks

[tmcd35]

Please ensure you take photo's of your injuries. I doubt they'll help much in your legal case, but if you post them up here we'll be sure to get a laugh out of them =P

 

Oh, I dunno - I'm sure pictures of the 'Intel Inside' tongue tatoo will make for some pretty compelling evidence

[CHR1S]

Well, I stripped and cleaned the CPU and HS then applied the Arctic Silver 5 and set it all back up again. It seems to idle at 65c and with less than 1 minute of prime95 it went mental hit 87c and was still climbing and shut down

 

Going to let it settle over night then check the compound tomorrow and re-apply etc. as required.