Wow, heat can affect speed a LOT!

[MRD]

I took a look at my father's PC (Duron 700) the other day and noticed that on boot it shows that the CPU fan is at 0 RPM and that the temperature is 92C (in the BIOS, w/o load!). He said it has been that way for months. He has also been complaining that his computer is really slow. Anyways, I decided to try to fix that. I took off the case and the fan was so clogged with dust it couldn't move. I took off the heat sink, which took a lot of force, because the green phase change material was burned and cracked. The rubber pads on the CPU had melted right off the chip and were permanently stuck to the heat sink.

I took an old athlon stock heat sink that I had and put it on with some silver thermal goo and the temperature dropped to 38C.

I find three things about this astounding.

1) The CPU could actually continue to work for that long at that temperature with no apparent damage.

2) There was no system instability, no locking up, not freezing.

3) Here's the one that really surprised me, as I've never heard it before. The speed of the CPU went up in benchmarks by about a factor of 3. Some tests (like 2D graphics) went up by a factor of 200.

I was unaware that heat could slow down a PC, I always thought it just caused instability.

MRD

 

[Mr.Radar]

Does that mobo have thermal throttling? If so, that would explain everything (CPU surviving, lack of instability, speed).

 

[SewerBeing]

yeah I am thinking thermal throttling

 

[ElectroX]

I've never heard of thermal throttling on AMD chips, I thought that was just an Intel thing

Don't AMDs self destruct at like 75c?

What mobo was it on?

 

[Sentential]

AMD chips do not have thermalthrotteling. They are just really hardy chips. It is very difficult to kill an AMD chip.

 

[Mr.Radar]

Wierd, theres a thermal throttling option in my NF7-S's BIOS.

 

[cack01]

Another amazing thing, is that the reading of 92C was not on die. The reading was most likely from a thermistor on the mobo.

 

[MRD]

The motherboard is a Biostar M7VKB. I can't find any mention of any kind of thermal throttling in the manual, searched it online for the words "heat", "thermal", and "cool" and got nothing.

Also interesting is that the slowdown isn't what you'd expect just from reducing cpu clock rate. For example, the 2D graphics rectangles test was slowed by about a factor of 200. That's not just strict clock speed reduction.

MRD

 

[boffin2163]

i think that AMD's just dont have thermal throttling built in like Pentiums its done by the M/B

 

[IZON]

What's thermal throttling?

 

[Mr.Radar]

Where the mobo will scale back the CPU's speed/voltage to help reduce heat (usually only happens at insane temps).

 

[infinity9]

AMD chips usually have a max core temp around 90C. I really doubt your thermistor was reading correctly at that temp. It was probably around 80C which is why its still alive.

If the cpu receives a failed code it will try the instruction again I believe. This was cause it to try multiple attempts to run any instruction slowing it down. It's been awhile since I've read up on CPU design but I believe thats how it works.

 

[fac_29]

Heat will definitely affect speed. I noticed just the other day that making just a few changes in my case so my Vid Card got more ventilation gave me a significant boost in 3D Mark. Makes sense when you think about it. Electrical current flows better and more efficiently at cooler temps.

 

[MRD]

I think Inf has the right idea. I believe that instructions were failing and restarting.

I think the temp was correct. All the phase change material was brown/black, dried, and cracked. The label on the fan had burned off, you could just see a few brown/black edges left on it. The rubber feet had burned off the processor and turned into puddles on the heat sink. The dust clogging up the heat sink was brown and burned as well.

I've never seen a chip function at those temps before, no wonder it was slow as all hell.

MRD