HSF mod and CPU temp question?

[KS1]

System:
A7V8X
AMD 1800+ T-bred A + Thermosonic ThermoEngine Alu HSF
Ambient 18C constant
Temp reading from MBM5210

Condition:
1. I lapped CPU die together with a Thermaltake copper shim. (I found that CPU die is not flat at all)
2. I replaced the stock fan with a 80mm Delta AFB0812SH to a heatsink.
3. I also inserted a 6mm copper plate the the base of the heatsink.

Result:
CPU temp is 35C when o/c 1916MHz 11.5x166 @ 1.7V, ambient 18C

I did a new experiment--replacing a 6mm copper plate with a 3mm silver plate.
As silver is better at conducting heat, I expected the temp to be lower.

Result:
CPU temp went all the way up to 48C and CPU would not run at 11.5x166 anymore. I got to go back to 1533MHz.
I took a silver plate out and re-lapped again and again. I made sure that the silver plate made full contact with the CPU Die. But the result didn't change significantly +/-1C
I went back to a copper plate again. CPU temp went down to a normal level 35C and ran at 1916 MHz comfortably.

Do you have any comment/idea/suggestion?

[kiz]

i thought copper conducted better than silver?

[KS1]

I don't think so. I did a research and found that silver is second only to diamond at conducting heat.

[naughty_guy]

yea, i think silver conducts heat better too.

[DRaG0nBLaD3]

...why else wouldn't they make

ARCTIC COPPER

? After all, copper's WAY cheaper than silver.

[felix88]

are you sure there was good contact between the base of the heatsink and the silver plate?

[KS1]

I'm pretty sure that all components made good contact.
I lapped the heatsink first. Then a copper plate. Then a silver plate.
In the testing process, the only change was a copper and a silver plate. I re-lapped the silver plate many times since I also suspected that it did not make good contact. Every time I lapped, I checked and found that the surface was 99.9% flat already.
AS III was applied very thin on all the surface.

[KnightElite]

Originally posted by KS1
I'm pretty sure that all components made good contact.
I lapped the heatsink first. Then a copper plate. Then a silver plate.
In the testing process, the only change was a copper and a silver plate. I re-lapped the silver plate many times since I also suspected that it did not make good contact. Every time I lapped, I checked and found that the surface was 99.9% flat already.
AS III was applied very thin on all the surface.

I suspect that the reason it was warmer with the silver plate is that it is thinner. If you were to try a 6mm silver plate, your results will probably be quite similar to those of the 6mm copper plate. This is probably because the thicker plate allows the heat from the die to "spread out" better and therefore gives more even heat distribution to the heatsink. However, not being a mechanical engineer, I couldn't tell you whether this is even remotely correct or not.

Wha I can tell you is that silver is only about 10% better than copper at transfering heat. Diamond is much better, and graphite is the next best, though only under certain circumstances.

[KS1]

That's possible.
The reason I bought this silver plate is that I read from a Peltier guide of overclockers com that a thin silver plate does as good as a thick copper plate. The thinner the plate, the lighter the weight. Hence, less weight on the CPU core.
But the result is disappointing.
----------
Update:

Side effect:
After the experiment (and let the CPU at 48C for a day or two), my 1800+ does not run 1916MHz at 1.7V any more. I need to up the volt to 1.8V to run 1916 stable.