What happens when you take the heatsink off a running CPU?

March 18, 2003 1:18 AM

What happens when you take the heatsink off a running CPU? The people at tomshardware.com find out by trying to fry four different processors. They also produced this excellent little video of them pulling heatsinks off running computers. You don't have to be a geek to appreciate burning silicon. You'll need the divx codec to view the video
posted by skallas (21 comments total)

Did you notice that the article on THG is from September of 2001?
posted by DoomGerbil at 1:42 AM on March 18, 2003

I wonder if it happened before or after 9-11 LOL!
posted by tljenson at 1:58 AM on March 18, 2003

This "simulation" has been widely disparaged by home-build enthusiasts as a pretty bogus attempt by tomshardware.com to slam AMD, allegedly either because AMD wouldn't provide free test hardware to Tom and/or because Intel handed over the freebies like there was no tomorrow.

Though I'm sure it could happen, catastrophic fan failure with no warning is so rare that I've never heard of it happening. If a heatsink fan is going to fail, it'll usually give you some warning, either by vibrating or slowing down, resulting in unusual noise or heat.

It's interesting to note that the version of this "test" that appears on tomshardware.com now has been edited from the original, where claims about heatsink assemblies "falling off" the motherboard played a much more prominent role. That was also considered pretty funny by legions of home builders on hardware forums all over the net, none of whom could remember ever having seen it happen.
posted by KiloHeavy at 2:28 AM on March 18, 2003

If I remember correctly, the conclusion was that Tom's Hardware Guide are Intel biased, hence the existence of this video. They don't show you the prices of the combustible AMD chips as compared to Intel, you may notice. Also, the speed-stepping (or whatever it's called) was the selling point for this generation of chips, given the aforementioned pricing discrepancy.

'You don't need to be a geek', but it helps. I have tried showing this video to non-geeks, and have met with little interest.

On preview. What KiloHeavy said.
posted by asok at 2:30 AM on March 18, 2003

I've had 3 AMD processors (2 x Thunberbird 1 GHz, 1 Duron 900 MHz) break because of thermal problems (heatsink wasn't properly secured). The processor breaks in the 3-4 seconds you need to notice something's wrong. I also have a Socket A motherboard that can't support heatsinks anymore because the plastic holders have been worn almost completely off (not really AMD's fault).
posted by rosmo at 2:35 AM on March 18, 2003

AMD really doesn't play up fact that you get a free space-heater for every processor purchase. It's really quite a good deal.
posted by Space Coyote at 2:40 AM on March 18, 2003

Fairly entertaining video but I don't know why it contained some kind of bizzare "euphoric elevator rave" in the sound channel. Very weird and adds to the file size.
posted by ed\26h at 3:16 AM on March 18, 2003

skallas - 'the framerates drop with one of the pentiums and he puts the heatsink back on and off it goes again.'

That is the speed-stepping (as you probably know). The processsor clock slows down to avoid over-heating, it is the same technology they use in the mobile chips to cut down on power consumption.
My friends with AMD chips have external chip temp. led displays, and/or motherboards with temp. cut-out to avoid over-heating.
posted by asok at 3:26 AM on March 18, 2003

I stopped at the point that I realised that the ignoramus who wrote the article doesn't know the difference between state of the art and run of the mill. Such misuse of simple terminology suggests that the author doesn't know what he's talking about.
posted by daveg at 4:34 AM on March 18, 2003

Yes. What an ignoramus! He is so damn ignorant he does not even say run-of-the-mill. Such misuse! such pomposity! how patronising! Shut up.
posted by ed\26h at 6:06 AM on March 18, 2003

I tried an analog of this test -with an aircooled VW bus- once.
posted by troutfishing at 6:33 AM on March 18, 2003

uh, i guess some people are STILL missing the point of this old (but important) video.

HEAT KILLS PROCESSORS. heat is bad. the hotter the CPU gets, the slower it computes.

SO, if heat is an important factor in stability and life of the CPU, wouldn't it make sense that it had a built-in thermal protection scheme? (yes!)

a better (and much more meaningful) test would have been to just disconnect the heatsink fans -- some newbies forget to plug them in, and some fans are just built poorly and eventually fail on their own. the CPU would still get dangerously hot -- i dunno about the smoke, but the message would be the same.

fwiw, i'm using an athlon xp 1800 right now. and he REFUSES the AMD-supplied heatsink/fan that he came with. won't boot 3 out of 4 times... if/when windows eventually loads, continuous errors. runs like a charm with a gigantic (and extremely noisy) thermaltake volcano heatsink/80mm fan...
posted by bhayes82 at 6:35 AM on March 18, 2003

IIRC, Athlon began shipping CPUs with heat protection because of this. In other words: You wouldn't be able to reproduce this experiment using present-day chips.
posted by cx at 7:05 AM on March 18, 2003

I once had a fan cut out on an Athlon board I was running - the machine was being worked on, and a wire got stuck on the fan. I was installing the OS, didn't know about it, walk downstairs. About five minutes later, I come back to a blue screen of death and a COMPLETELY unresponsive machine. Not even the reset button works. So I cut the power, and think it's just one of those usual windows things. I stick my head in the machine, getting ready to switch out a hard drive, and think "Huh, it's really hot in here." Fifteen minutes later, when the machine cooled off enough to start again, the processor temperature was 214 Degrees Ferenheight. But the processor still worked fine after this.
posted by kaibutsu at 9:00 AM on March 18, 2003

I haven't had a problem with any of my Athlon XP processors, but I'm going to a water cooled jammie next.
posted by adampsyche at 9:12 AM on March 18, 2003

Though I'm sure it could happen, catastrophic fan failure with no warning is so rare that I've never heard of it happening.

I had it happen twice in about 6 months back in the pentium days. Back then most systems wouldn't even create some type of warning beep if things were getting out of hand (something that seems common now). At least one of those 2 cases it also managed to wipe out the motherboard when the cpu went.

In general I'm not too impressed with the fans used on computers, lots of them seem to start sounding sick after a few months. The fans on my laptop have really started sounding bad lately, and given then heat that thing creates with the fans working (think man fries sensitive parts type heat), if those fans go I'm sure it'll toast something.
posted by piper28 at 10:13 AM on March 18, 2003

Holy crap.
I remember this video. I loved it, more for the quality that Divx was able to prove it could deliver (in a relatively small file size) than for the information it was relaying.

That said, this was done with technology that wasn't cutting edge at the time of posting, so its relevance may have worn a little, but the message was still clear: AMD needs to develop a better way to manage the heat their CPUs produce. Pretty simple.

As for claiming Tomshardware was bashing or setting AMD up with this experiment, Toms has proven themselves pretty much in AMDs camp, lauding the fact that a company with a small fraction of the funds that Intel has can still compete (and even outperform) the Microsoft of the processor world.

Just think of this video as an illustration of why you need to be careful placing the heatsink on the processor.

AMDs still produce a lot of heat, but cases are keeping up with keeping them cool enough to make this a non-decisive factor for building your own pc.

Not that many people bother to anymore, since it now costs you more to build your own than just wait for some Dell business computer deal, where you can get a stripped down system that costs far less than the internal components would if you tried to buy them yourself.
posted by Busithoth at 10:25 AM on March 18, 2003

Not that many people bother to anymore, since it now costs you more to build your own than just wait for some Dell business computer deal, where you can get a stripped down system that costs far less than the internal components would if you tried to buy them yourself.

Is this really the case? I've been looking to get a new kinda-low-end machine, and it seems like I could build it for about $300 cheaper than I could get it from Dell ($600 vs. $900). Should I wait for a sale from Dell? Am I missing something?

As a matter of principle, I'd rather build my own machine, but money's tight, and the conventional wisdom is that a prefab system is cheaper. I was surprised that my recent shopping didn't back up the conventional wisdom.
posted by mr_roboto at 11:10 AM on March 18, 2003

The only reason I'm calling double post is because it's the only FPP I have ever done. ;)

posted by linux at 11:12 AM on March 18, 2003

looking at the new dell site it seems they offer less customization than a serious user would really want, and the prices are nothing special. i could definitely build a computer for less money than they are asking. do keep in mind the software you will need ie windows, office, norton that comes with a prepackaged system, if you do not already own or have acces to this stuff it can get very expensive. i would highly reccomend a more independant site such as this one . you could easily find a well built, premade computer for under $600, completely customizable with nifty cases to boot.
posted by sophist at 3:50 PM on March 18, 2003

Bringing this back up to date: Athlon XP CPUs have built-in thermal diodes which measure the temperature of the chip. Modern (KT333/KT400/nForce2) motherboards will shut down if this temperature exceeds specifications.

So, uh, yeah.

Also, the heat output of top-end Athlons with Pentium 4s are pretty equal.

meh.
posted by Coda at 12:57 PM on March 19, 2003