Only one CPU has died (out of dozens used), and it may have been because the mobo got confused about how much juice to send through it.

All CPUs are from ebay, and some were sent to me in visibly poor condition. Also, others look in OK condition but may have been overclocked/volted. (I'm not entirely to blame for all problems encountered.)

Some stability problems appear to have been BIOS settings/driver versions/software related, because they've been fixed.

Most experiments are just frivolous 30 minute tests, and not part of any "real working system". (My primary goal is to work out which CPUs work with which mobos.) Then I'll get (hopefully better quality) replacements from ebay, and look after them properly from the outset. Naturally I'll use paste when I make a "real working system". I use good paste in my main machine. It's got an E8400 in it and idles at 25C, with core temps of 35C and 40C, and its fan is just 800 RPM!

All tests are done with a heavy heatsink sitting vertically on top of each CPU, not with the mobo (and heatsink) positioned at a 90 degree angle inside a case.

The maximum working temperature of these CPUs is around 100C, which is extremely high.

I'll probably incur the Wrath of old Thrashbarg (which sounds like a cheesy Dungeons & Dragons adventure), but here are my test results - which is just me messing about, and is not to be taken seriously.

CPU = Palomino 1900+ (I couldn't use my Palomino 2000+ because I accidentally fried it.)
Only joking.

42.8C (10 minutes of idle desktop usage.)
I then ran coolon.exe and waited for 10 minutes. Temperature = 22.8C. That's an amazing reduction!
I didn't run Prime95, because I didn't want to risk damaging the Palomino CPU as they do tend to get a bit hot.

(I then let the system cool down, and then applied thermal paste to the CPU.)

33.4C (10 minutes of idle desktop usage.) That's an amazing reduction! (From 42.8C without using paste.)
I then ran coolon.exe and waited 10 minutes. Temperature = 21.1C.
I then ran Prime95.exe and waited 10 minutes (with coolon.exe still "active"). Temperature = 39.9C. I guess that's reasonably good, considering I'm using a heavy heatsink, paste and also coolon.exe.

I have been testing about a dozen Athlon CPUs (Durons, TBirds, Palominos, TBreds) today - all of them using thermal paste. I am applying a very thin coating of paste to the CPU die using a cotton bud (or "q tip"). Also, when I remove each CPU in order to try the next CPU, I wipe off all the paste with another cotton bud. Also, I have been able to wipe away old existing dried paste quite effectively with a cotton bud. Even some of the dried stuff which appears to be quite firmly stuck to the CPU die, and also its surrounding area has been wiped off. (A few bits still remain, here and there.) Are all of these actions OK? I mean, there's no danger of me "statically damaging" the CPUs by doing this?

Just make sure you don't get cotton in the paste. That would be very undesirable. 😀

This is probably the only guide I'd take seriously. I don't do the "apply to heatsink" step though.
http://arcticsilver.com/instructions.htm

Thanks a lot. Earlier today, I've been using the end of a cotton bud to smear the thermal grease (very thinly) over the entire CPU die head area. - Quite different to the "line or blob" in the middle of the die head approach, as seen in those photos on that web link you provided. After testing, when I remove the heatsink, I see very little (if any) "grease spillage" over the side of die head. Naturally, there's a "smudge" of grease on the heatsink. The stuff I'm using seems quite "solid" - not "liquidy". It's dark grey, and not white.

Please can you tell me why getting a strand of cotton stuck in the grease would be bad? I think that happened once today, but it's since been all wiped off, as I have been testing and changing quite a few CPUs today.

Incidentally, I've had a good testing day today. At last, I finally managed to get a slightly upgraded QDI Kinetiz board. I tried another version 2.2 board today, and it was even more unstable than my original/initial board. But then I got a 2.3 version today, and everything has worked without any failure. (Every hardware component, software "component", BIOS version and settings identical for both 2.2 and 2.3 tests.) Now, Memtest86 always works. (Previously, it would always fail for every CPU faster than a TBird.) 3DMark always works. CL2 SDRAM works 100% too, in all tests. I'll post a meaningful message tomorrow about it, probably in the "more fun and games with KT133A" thread..

Cotton is bad because it will not become as thin as the paste and will thus prevent the heatsink from resting as close to the core as it otherwise would. Cotton is also a great thermal insulator.

swaaye wrote:

Cotton is bad because it will not become as thin as the paste and will thus prevent the heatsink from resting as close to the core as it otherwise would. Cotton is also a great thermal insulator.

I agree. The paste will also get adsorbed in between the cotton fibres, producing uneven smear.