If your struggling with high temps...

Things change I've changed · 22 Mar 2010 at 17:28

Then a high ambient temperature could be your culprit, I know this sounds obvious but until recently I really didn't realise how much of an impact a small rise in ambient temperature can have to your CPU's temp.

Today I wake up and check my idle temp for my recently over-clocked system (in sig), 46c, on idle

with a room temp of 26.7c, so I whack on the portable air con and only 30mins later with a room temp of 22c my cpu is now @ 35c, that's over 10c lower just for cooling my room down a few degrees.

This thread may seem a bit pointless but I think many people including myself overlook how important and how much of an impact ambient temps have on the temperature of your cpu.

I think where I went wrong was thinking that say 1c rise in room temp will equal only 1c rise on the cpu, but that's certainly doesn't seem to be the case (not sure why or how though).

RJC · 22 Mar 2010 at 17:47

Ideally you need to check the air inlet temperature to the case and cpu cooler.

How close was the a/c unit to the pc case.

Things change I've changed · 22 Mar 2010 at 17:53

RJC said:
Ideally you need to check the air inlet temperature to the case and cpu cooler.

How close was the a/c unit to the pc case.

The other side of the room, its a double room but I do have the vents pushing the air across the room towards the PC area.

JonJ678 · 22 Mar 2010 at 17:58

Heatsinks tend to have a fairly constant C/W value, so for a given heat load should sit a consistent distance above ambient temperature. I'm unsure what additional factors contributed to you seeing a decrease in temperatures of double the decrease in ambient, and they may be hiding well, but they'll be present. Increased air circulation and uneven temperature distribution through the room perhaps?

Howitzer · 22 Mar 2010 at 17:59

It's a very good point!
I remember the IT suite in my old school had permanent cool air-conditioning basting (even in the middle of winter!) for this very reason.
That's not really practical in a home environment but ambient temps (and their variations such as summer/winter) should always be concidered (especially when OCing)

miniyazz · 22 Mar 2010 at 18:11

Two options.
1) the room temperature is being measured in a place not representative of the air intake temperature of your computer.
2) warmer components generate more heat than cooler components via ↑ resistance (true but I'm not sure how significant this is)

Jokester · 22 Mar 2010 at 18:18

miniyazz said:
2) warmer components generate more heat than cooler components via ↑ resistance (true but I'm not sure how significant this is)

Warmer components actually generate less heat due to the increased resistance - though silicon if I remember correctly has lower resistance the hotter it gets, but most of the heat generated in modern CPUs is from the current in the copper interconnect.

JonJ678 · 22 Mar 2010 at 18:32

The hotter doped silicon gets, the higher the fraction of electrons in the mobile energy band and the lower the resistivity. At roughly constant supply voltage this leads to higher currents, which make it hotter, thermal runaway occurs and it all goes wrong very fast. Not applicable to modern processors, but probably contributes to the high fraction of dead 741s in workshop draws the world over.

Interested to hear that it's the current in the copper that leads to most of the heat generated in a processor, didn't know that.

miniyazz · 22 Mar 2010 at 18:39

I appear to have made the mistake of assuming GPUs and CPUs are Ohmic!

Jokester · 22 Mar 2010 at 18:41

miniyazz said:
I appear to have made the mistake of assuming GPUs and CPUs are Ohmic!

V=IR
For a fixed voltage, if you increase resistance, current decreases, leading to:-

P=IV
Same voltage but less current = less power.

mmj_uk · 22 Mar 2010 at 19:03

The best thing I did was go watercooling, not only are temps now easily under control but it doesn't require noisy fans.

JonJ678 · 22 Mar 2010 at 19:03

Thats assuming resistance is independent from current (and indeed from voltage) though, and as current tends to make things hotter and resistance varies with temperature, it's not always applicable. That's before taking into account NP junctions which aren't even remotely ohmic and are probably quite popular in processors.

Jokester · 22 Mar 2010 at 19:18

JonJ678 said:
Thats assuming resistance is independent from current (and indeed from voltage) though, and as current tends to make things hotter and resistance varies with temperature, it's not always applicable. That's before taking into account NP junctions which aren't even remotely ohmic and are probably quite popular in processors.

Indeed they are, but as I said before, the current in the copper interconnect generates about 70% of the total power consumption in a modern CPU

miniyazz · 22 Mar 2010 at 20:01

I'm just going to stop digging now and leave this conversation to those who did A-level physics/electrical engineering