Coldbug Drift?

[Patch]

Often I'll get a new CPU or mobo and get all excited when it doesn't have much or any coldbug the first time I freeze it, but I can never attain those temps again.

Any ideas as to why?

Most recent example:
980x: on Classy 762 seemed to cold bug in -135 range initially, but with later benches can't get much colder than -120. Same CPU on R3E benched at -150 initially without bugging (didn't try colder at the time), then started to bug at -135, now bugs at -127.

Is this an insulation thing? Mobo? CPU? Or is it simply tightening too tight? - Maybe I get more aggressive as I get more comfortable with a rig....... But loosening never really seems to help.

I've had this problem with 2 980x's, 3 i5 670's and a number of Bloomfields. A couple of the i5's would even go to "full pot" on the first run, then cripple me at the -130ish range thereafter.

Do CPU's just tolerate less cold after a while?

[MaadDaawg]

I seem to have the opposite "problem", at least during one bench session. The longer I bench the colder I can go without bugging out ?? As example, first CB may hit at -128C, subsequent one kicks in at -138C, then it jumps to around -158C

[Hondacity]

1. cpu degradation
2. cap degradation? ive yet to get my capacitance meter adn test this
3. cpu pot mount....

thats my list for now...

[RomDominance]

not goin with degradation as the cause....my theory lies in moisture as the cause.
If I take my eraser off and open the socket to the elements for a period of time, the cb seems to be what it was initially.

[Splave]

mounting pressure?

[ReverendMaynard]

retraction and expansion?

Here's a read about Martian surface temperatures and solder.

http://trs-new.jpl.nasa.gov/dspace/bitstream/2014/38569/3/05-3335FN.pdf

[Patch]

Thanks guys. Very good tips. And you get the prize RM for the coolest article referenced on a forum. :thumbsup:

Gonna de-insulate before the next session and see what happens. With my half-decent fermi and a rockin IMC on the CPU, if I could get it really cold I think I could crush a few benches.

[Dentlord]

I discovered something weird the other day ... when running 2 sticks of ram I have no CB, with 3 installed it bugs at -120??

Which explains why I could run triple channel on my SS just fine, but could never run it on ln2.

[Hondacity]

i sense world records :D

[Patch]

I discovered something weird the other day ... when running 2 sticks of ram I have no CB, with 3 installed it bugs at -120??

Which explains why I could run triple channel on my SS just fine, but could never run it on ln2.


Shammy posted that info on a new locked multi GT the other day. He had CB -150 with triple channel and no coldbug with dual channel.

Holy crap - can't wait to try it. I'm off next week. Almost time to play. :cool3:

[FACE]

Does this mean when quad channel drops it will cb at +10c? :laughing:

[Patch]

Eliminated coldbug on the R3E by moving the QPI LLC jumper back to position 1-2.

Coldbug changed from around -130 to no coldbug at all. Performance seemed to suffer, so it's a trade off.

Couldn't reach much higher absolute clocks with last night's tweaking. But 3D CPU tests completed with a higher OC than they did with higher temps.

Time to freeze the fermi for some 06/vantage action tonight.

[Chuchnit]

Could you be getting a little cold slow action? Maybe there is a sweet spot that's lower than absolute max ln2 temps?

[Patch]

Hmm...don't know.

With jumper in 1-2 position it got rid of coldbug, but I couldn't clock the CPU very much higher - and high uncore/ram were tougher to achieve. I'm hoping I can find the magic settings to have the best of both worlds.

But, going colder is a definite advantage in 06 and Vantage CPU tests. Will help only a tiny bit in Vantage because GPU was already the bottleneck, but I have hopes for actual gains in 06.

[Splave]

also lets you focus on the gpu pouring :)