A question about thermal paste

**Hammerstein** · 10-07-2003, 09:38 AM

What happens if you don't remove and reapply thermal paste to the heat spreader on a P4 after moving a CPU to a different motherboard?

**speedstream5621** · 10-07-2003, 09:51 AM

1. You allow dust, etc. to stick to the thermal paste.
2. You will not have an evenly spread layer when you install the heatsink.

Just clean it and reapply it...

**Hammerstein** · 10-07-2003, 09:59 AM

The question then is - would it be bad enough to cause processor throttling on a P4?

The non-visible type where the thermal control unit only throttles internal parts while maintaining the clock speed.

**mikead_99** · 10-07-2003, 10:44 AM

Possibly, it depends on how bad the compound gets contaminated, how uneven it becomes, how much air gets in the gaps, etc. IMHO it's better to be safe than sorry. I have found that after re-applying the compound my temps improve over previous, probably because pores are now becoming filled in.

**Hammerstein** · 10-07-2003, 01:29 PM

So, then, that could be the source of performance issues, I suppose?

It's a long shot but my system has had tons of trouble and I can't get it up to spec in terms of memory bandwidth and CPU performance. Motherboard is fine (double replacement!) so it's either CPU or RAM. Going to this last shot before CPU replacement...

**speedstream5621** · 10-07-2003, 01:50 PM

It's a long shot but my system has had tons of trouble and I can't get it up to spec in terms of memory bandwidth and CPU performance. Motherboard is fine (double replacement!) so it's either CPU or RAM. Going to this last shot before CPU replacement...

If its a matter of performance, you need to tweak certain settings. Having a poorly applied layer of AS3 isn't going to affect your memory bandwidth, and it certainly doesn't need replacing.

**mikead_99** · 10-07-2003, 01:59 PM

speedstream5621:

If its a matter of performance, you need to tweak certain settings. Having a poorly applied layer of AS3 isn't going to affect your memory bandwidth, and it certainly doesn't need replacing

I'm not so sure about that....after my wonderful Abit nb hsf popped off (partially) I lost about 25% of the memory b/w I should have had. Also the memory b/w is measured by exercising the CPU/mem interface, therefore CPU, mem, or nb could all affect the result.

Another thing to try is to put your RAM at stock timings and bump the VDIMM to 2.7 or 2.8 and test. If you are running the timings in your sig it could be that you are able to run but are wasting b/w with error correction. If that helps you can start moving voltages/settings down 1 at a time to see where your sweet spot is.

**Hammerstein** · 10-07-2003, 03:14 PM

Did up the voltage a while back coupled with reducing timings back to 2.5-3-3-8...no help there, I'm afraid. Bandwidth isn't a matter of just dropping a bit - it's at 1600 MB/s in Sandra memory tests (around the benchmark of the old AMD760 chipset) - it's basicaly terrible and CPU scores also reflect some major issue, performing /far/ below spec.

**mikead_99** · 10-07-2003, 04:30 PM

I hate to ask, I hope I'm not insulting you. Are you sure you have the sticks in the right slots for dual channel?

I think that's probably low for even single channel but I had to ask.

**Hammerstein** · 10-07-2003, 04:32 PM

Not insulted - stranger things have happened

But, since there's only two slots on the motherboard as it is a small form factor machine, no

**mikead_99** · 10-07-2003, 05:13 PM

Hmmm, that might explain it. I'd look at the manufacturer's website to see if the mobo supports dc, or even an 800 MHz fsb. As I mentioned earlier, it still looks low if only in single channel, but if fsb is say 400 MHz then you just have the wrong proc for that board. Also cooling in those small form factor cases can be problematic, maybe you are throttled as well.

Thinking about it though, I bet the mobo is probably made to run with the P4B's. Yours should still work, you just are missing out on a lot of features built into the chip if that's the case.

**Hammerstein** · 10-07-2003, 05:22 PM

It's a i865G motherboard so it supports both the 800 MHz FSB and dual channel DDR. I know that it works since this is my third attempt and the first one I had was running perfectly in dual channel and performing within the normal margin of error on benchmarks, etc.

**mikead_99** · 10-07-2003, 06:31 PM

Hammerstein:

It's a i865G motherboard

Whoops. Yeah, I couldn't see your sig while I was replying.

The only other thing I can think of is to try running your bench again with the case open in a cool room. If there is improvement this will point to a cooling problem. I know it's a pain in the butt to RMA critical components in a PC, if there is improvement maybe you can correct by reseating the CPU an nb heatsinks rather than sending it back in.

You've verified clocks, memory ratios, etc in your BIOS I'm assuming.

**Hammerstein** · 10-07-2003, 07:19 PM

Yeah...everything I could think of in terms settings. Even tried the open case. This is a sort of 'last resort' theory before I RMA processor and then, if that doesn't help, RAM. Recieving a new iDeq tomorrow so I get a fresh mobo to try on. If three motherboards in a row are defective electronically (last one had a defective HSF retention system!) but worked electronically, I guess it is something else...

But, if my stupidity has caused the P4 to just go into higher level throttle of some sort (since I read the white paper on the P4 thermal system which seemed to indicate that the processor could throttle internal components while maintaining core clock speed), then that would be great and I'd be done :/

Thread: A question about thermal paste

Thread Tools

Display

A question about thermal paste

Posting Permissions