CPU Cooling - Does temperature impact performance?

[nicklmg]

NCIX: http://bit.ly/1EmUx6S

Amazon: http://geni.us/3zaZ

 

Does a colder CPU get more performance with the same clock speed? It turns out that testing this may be more difficult than it first appears...

 

[MeltingPoint]

Looks like I'm the first viewer 

[Dabombinable]

Lol, I actually just finished a similar test with my Xeon.

[wolfgang]

Dont do this.

Dont do this.

Dont do this.

Dont do this.

Very informative video. What was the video originally about?

[Sam Z Man]

I love the ghettoness of the new videos, it makes it fun, really looking forward to see the dual 16 core server build based on what everyone said about it.

[beebskadoo]

Linus WhatNotToDo tips

[19_blackie_73]

Linus WhatNotToDo tips

exactly:-)

[Guest]

This is more comedy than tech. Like one of those Top Gear challenges, but for computers.

'We always say temperature doesn't affect performance, but one of the producers said we should actually test the claim and we agreed. I mean, how hard could it be?'

[beebskadoo]

This is more comedy than tech. Like one of those Top Gear challenges, but for computers.

'We always say temperature doesn't affect performance, but one of the producers said we should actually test the claim and we agreed. I mean, how hard could it be?'

"Top Gear!!! Ambitious but Rubbish!!!"

 

R.I.P

[yourfavoritenumber]

Cooling the CPU on the same clock speed actually slows it down, since the oscillator is oscillating a little bit slower, which results a slower clockspeed, though it wouldn't be more than a few milliseconds a day, and you would notice only if you compared the time to an external clock.

 

Today's processors however have some techniques to compensate the varying temperature effects: http://www.freepatentsonline.com/7541878.html

 

Does temperature impact performance?

Yes it does, but you will never notice it.

 

Did Linus use the right method to measure it?

No he didn't. This measurement requires an external clock.

 

Is it possible for Linus to measure the difference?

No it's not. At least not in his current office.

[The Knome]

As I remember it, 4000mhz and up is, for CPU's, where it gets difficult to push the speed any higher as a lot of the energy is lost in heat (and heat creates inefficiency). If the CPU is designed to stay efficient at 5000mhz then super cooling it wont make a difference on that hypothesis.

If you are looking to push it to the point of destruction on that top speed then the afore mentioned issue comes into play (its actually physics, the copper particles as an example will move more blocking electrical movement, like a car trying to cross a busy junction).

There is a reason at these overclocking comps they use liquid nitrogen, because the afore mentioned physics really kicks in like crazy and its the only way to keep it stable (as well as stop it melting or setting itself on fire).

Its also the reason for multi-core and multi-threading coming into play years ago. 

My really dumbed down version but its all verifiable factoids. 

 

[MyNameIsTaken]

We had this happen to a bunch of servers in a production high performance compute farm.

The management software screwed up and stopped monitoring the temps and ran the fans at minimum. (for months, possibly years, no one noticed)

I think these were supposed to run at 3.3Ghz with all core turbo when properly cooled vs 1.2Ghz @ 95C.

 

so yeah... dumber things have happened in real life.

[Psykomantis00]

Does anyone have a tldr? 

[Jumper118]

that was cringe worthy. i could barely watch. 

[.spider.]

And now do it again and measure the power usage.

[Jumper118]

And now do it again and measure the power usage.

the phase will use around 1-1.5KW

[mrthuvi]

We had this happen to a bunch of servers in a production high performance compute farm.

The management software screwed up and stopped monitoring the temps and ran the fans at minimum. (for months, possibly years, no one noticed)

I think these were supposed to run at 3.3Ghz with all core turbo when properly cooled vs 1.2Ghz @ 95C.

 

so yeah... dumber things have happened in real life.

This is hilarious. :lol:  What were everyone's reactions at your workplace?

[.spider.]

the phase will use around 1-1.5KW

What do you mean by phase?

[Fishsticks]

We need to implement a cringe counter whenever he does this kind of things

[peanutsrevenge]

From my GPU mining experience, lower temps make a significant different to efficiency, for compute performance .. it's within margin of error.

 

Wouldn't save enough power to offset the running cost of sub-ambient though

[aselwyn1]

why does all the cheero stuff have to not have shipping to canada

[euzen01]

Interesting

[Jumper118]

What do you mean by phase?

a phase change cooler

[Lays]

that was cringe worthy. i could barely watch. 

 

 

IKR

 

I think he should do another one showing that you can get more frequency out of the item @ same voltage, with colder temps.

 

IE: 1.3 volts on phase and 1.3v on high end AIO.

[Shadow-R]

Terve!

I remember a quote from a fast as possible: "Oh Linus, this is why we love you."

Matches this perfect. I'm alway in for more "how not to do" things!

You should also do a training for "how to mount your CPU vertically, if the IHS does not stick".

That's what I had to deal with in the past...

Sayonara!