• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

What temperatures do datacenter CPUs run at?

Yuriman

Diamond Member
Just curious if anyone knows. I know it will vary between server manufacturers and CPUs, but for example how hot are typical Haswell or Broadwell Xeons allowed to run during continuous operation?
 
Not sure on datacenters specifically, but plenty of military server racks/server rooms/whatever you want to call them sit in places where ambient can be north of 50c and the systems are full of sand and dust and shit. They aren't happy and don't last long without regular maint like that but they can run.

I'd presume as long as it's not above 90c for extended durations it's probably fine for the lifecycle of the system.
 
Just curious if anyone knows. I know it will vary between server manufacturers and CPUs, but for example how hot are typical Haswell or Broadwell Xeons allowed to run during continuous operation?

At up to official Intel temperatures, but large datacenter rooms have environmental controls keeping the room ambient temps constant at the selected temps.

In many cases (super computers, Huge Data Centers and Could servers) the annual cost for cooling is very high.

And this is a nice way to use water cooling (by using the sea water temperature differential) and save tons of annual energy cost
and also have the server closer to the market target (lower lag due to shorter distances)

Ocean Submerged Servers
 
Take the Broadwell Xeon E5-2695 v4 as an example.

http://ark.intel.com/products/91316/Intel-Xeon-Processor-E5-2695-v4-45M-Cache-2_10-GHz

It has a Tcase of 84c, and my interpretation of Tcase was that it was the highest temperature the surface of the heatspreader was allowed to run. Given that the heatspreader is soldered onto these CPUs, the gradient between core and IHS surface is probably much smaller than on 1151 chips, but would it be safe to assume that these Xeons run at 90-100c core temp at full load for years?
 
What an incredibly useful link. Apparently Intel rates most of their Xeons to run between 90 and 103c at full load for "long-term reliability".
 
would it be safe to assume that these Xeons run at 90-100c core temp at full load for years?

In a data center in general you aren't managing core temperatures, but server and rack temperatures. High CPU temps would cause an alert, but what that alert point is is up the server vendor. So an operations team doesn't know or care what the CPU temperature is as long as the management software is reporting "green".

For example:
http://www.cisco.com/c/en/us/produc...s-central-software/data_sheet_c78-717193.html

But to answer your question, that's what Intel warrants the Xeon's to do. It's one of many reasons why validation is hard.
 
I'll let you know as I have new dev boxes base don two 2640 v4's that I intend to use supermicro's hyper-speed on the CPU to let me run near max turbo on all 10 cores per cpu (3.5ghz is single core max) .

I'm hoping that I am able to get a reliable 3.0 ghz all core 24x7 and full load temps under 80c
 
In a data centre you care far more about temperature variance and humidity then the absolute temperature. As a result ambient can be almost anywhere, traditional datacentre was 21-23 degrees c.
 
Some of the cloud datacenters use machine learning to determine power draw at different temperatures (with everything held the same, a CPU at 80* has significantly more leakage vs a CPU at 60*) vs. the cost of cooling the air to determine the most efficient temperature to run at.
 
Back
Top