Bitcoin Forum
April 20, 2024, 12:08:19 AM *
News: Latest Bitcoin Core release: 26.0 [Torrent]
 
   Home   Help Search Login Register More  
Pages: [1]
  Print  
Author Topic: Mind testing your cards folks?  (Read 1438 times)
Remember remember the 5th of November (OP)
Legendary
*
Offline Offline

Activity: 1862
Merit: 1011

Reverse engineer from time to time


View Profile
December 01, 2011, 07:50:12 PM
 #1

Basically, running a GPU at 99% for an extended period of time surely means one is getting closer to electromigration.

I've tested my 5850 with OCCT which ran at 950mhz for 2-3 weeks, and then back at stock and after 40 seconds into the test at max shader,fullscreen the program started detecting artifacts...a lot of them.
I got over 50 and counting till i stopped the test.

Should i assume it was caused by electromigration?

BTC:1AiCRMxgf1ptVQwx6hDuKMu4f7F27QmJC2
Advertised sites are not endorsed by the Bitcoin Forum. They may be unsafe, untrustworthy, or illegal in your jurisdiction.
cablepair
Hero Member
*****
Offline Offline

Activity: 896
Merit: 1000


Buy this account on March-2019. New Owner here!!


View Profile WWW
December 01, 2011, 07:52:22 PM
 #2

I could totally google it, and then pretend I am super smart but I would rather you explain it to me...

WHAT THE HELL IS ELECTROMIGRATION?

are my GPU's flying south for the winter? ;P
DeathAndTaxes
Donator
Legendary
*
Offline Offline

Activity: 1218
Merit: 1079


Gerald Davis


View Profile
December 01, 2011, 08:11:25 PM
 #3

Electromigration is possible but at stock voltage and 2-3 weeks any issues you are having aren't from EM.  Completely wrong timeframe.

I mean a gamer playing an average of 2 hours per day would rack up 2000 or so hours over 3 year lifespan.  Thats more like 13 weeks as full load.  Some gamers play a lot longer than 2 hours a day and many overvolt the cards.

If you overvolt 20% you may see electromigration after a year of sustained load.

cablepair
Hero Member
*****
Offline Offline

Activity: 896
Merit: 1000


Buy this account on March-2019. New Owner here!!


View Profile WWW
December 01, 2011, 08:19:01 PM
 #4

oh right, ELECTRO migration, my grandma died of that Sad
P4man
Hero Member
*****
Offline Offline

Activity: 518
Merit: 500



View Profile
December 01, 2011, 08:19:28 PM
 #5

Electromigration gets worse with high temperatures though, and depending how his rig was cooled, those 3 weeks might have been a lot worse than 3 years of gaming. Besides, not all cards survive 3 years of gaming. Its always a bit of a lottery.

That said, are you by any chance running GPU-z and/or everest as well as occt? If so; DONT.

cablepair
Hero Member
*****
Offline Offline

Activity: 896
Merit: 1000


Buy this account on March-2019. New Owner here!!


View Profile WWW
December 01, 2011, 08:28:13 PM
 #6

look how you guys are ignoring me, no wonder my self esteem is so low
DeathAndTaxes
Donator
Legendary
*
Offline Offline

Activity: 1218
Merit: 1079


Gerald Davis


View Profile
December 01, 2011, 08:35:48 PM
 #7

look how you guys are ignoring me, no wonder my self esteem is so low

Electromigration is an effect in semiconductors (well in any circuit but we really only care about it in semiconductors).  The high flow of electrons though the gates causes metal ions to be moved.  Eventually this will degrade the performance of the gate and eventually destroy it.

As the gates performance degrades its ability to accurately switch open and close at ultra high speed becomes impared.  Symptoms tend to include errors, crashes, hard locks, etc.  Lowering the clock speed can restore stability because it is now under the cards newly reduced peak performance.  Over time clocks speeds will need to be continually lowered and lowered and lowered to keep the card stable.

Higher current and higher temps increase the rate of electromigration but all circuits are continually being degraded by EM.  Every circuit every built will eventually be rendered inoperable by electromigration.  To degrade the wiring in your house however might take a couple thousand years (the wires are many magnitudes larger than the traces inside a chip).
Remember remember the 5th of November (OP)
Legendary
*
Offline Offline

Activity: 1862
Merit: 1011

Reverse engineer from time to time


View Profile
December 01, 2011, 08:39:30 PM
 #8

Yes, my 5850 did see an increase in voltage. The max i went(but not stay constantly at that voltage) was 1.2mV. Max temps at the time were 85C and 100% fan.

BTC:1AiCRMxgf1ptVQwx6hDuKMu4f7F27QmJC2
DeathAndTaxes
Donator
Legendary
*
Offline Offline

Activity: 1218
Merit: 1079


Gerald Davis


View Profile
December 01, 2011, 08:44:34 PM
Last edit: December 02, 2011, 02:52:49 AM by DeathAndTaxes
 #9

Yes, my 5850 did see an increase in voltage. The max i went(but not stay constantly at that voltage) was 1.2mV. Max temps at the time were 85C and 100% fan.

85C is pretty hot.  High temp + high voltage = reduced lifespan.  The effects of electromigration are non-linear and sadly irreversible.

Don't overvolt (especially w/ todays bitcoin prices & electrical costs it makes no sense anyways) and keep the GPU temp in the 60s.  As a side note 100% 24/7 fan usage is going to kill the fan prematurely too.

It isn't an issue of load.  It is more an issue of the overclock, overvolt, and temp.  Just because you can get a card to xxx clock doesn't mean it is sustainable. 
vapourminer
Legendary
*
Offline Offline

Activity: 4312
Merit: 3506


what is this "brake pedal" you speak of?


View Profile
December 02, 2011, 12:57:48 AM
 #10

as a long time folding @ home addict, I can say that running overclocked at stock volts and decent temps has never caused an issue with several video cards. or CPUs either, and those I do overvolt as well as OC.

most recently I have 2 GTX 285s that have folded more or less non stop 24/7 at 100% load, factory overvolted and mild OCs, for 2 years now (they have aftermarket artic cooling HSFs with 3 x 92 mm fans). no problems with games or folding to this day (its my gaming rig). I have several intel CPUs with mild overvolts and fairly heavy overclocks that have run flat out for several years doing SMP or single core work units, no issues.

keep the temps reasonable and as close to stock volts as possible.
tritium
Member
**
Offline Offline

Activity: 81
Merit: 10


View Profile
December 02, 2011, 03:37:31 AM
 #11

I have noticed a reduction in the clockspeed i can have some of my cards. i have 3 5830s which were completly stable in july at 1010MHz, that has dropped roughly 10MHz every 2 months, the cards run at about 75C for the top 2 and 65 for the bottom one, i havent seen a difference between them at those temps and have all degraded about the same rate

1FCzN34C1xCLsDaLxfY7yB5CQKN74ruGHV
deslok
Sr. Member
****
Offline Offline

Activity: 462
Merit: 250


It's all about the game, and how you play it


View Profile
December 02, 2011, 05:45:33 AM
 #12

nothing strange on a mixture of 5770-5870 cards here, however artifacting can be caused by any number of things and may have been something your card was predisposed to already

"If we don't hang together, by Heavens we shall hang separately." - Benjamin Franklin

If you found that funny or something i said useful i always appreciate spare change
1PczDQHfEj3dJgp6wN3CXPft1bGB23TzTM
check_status
Full Member
***
Offline Offline

Activity: 196
Merit: 100


Web Dev, Db Admin, Computer Technician


View Profile
February 11, 2012, 03:33:34 AM
 #13

Can the motherboard bus and southbridge be causing or contributing to stability issues like in the OP?

What testing should you be doing after purchasing a used card through the internet middler?
If the visual inspection of it appears good and clean, what next?

For Bitcoin to be a true global currency the value of BTC needs always to rise.
If BTC became the global currency & money supply = 100 Trillion then ⊅1.00 BTC = $4,761,904.76.
P2Pool Server List | How To's and Guides Mega List |  1EndfedSryGUZK9sPrdvxHntYzv2EBexGA
Pages: [1]
  Print  
 
Jump to:  

Powered by MySQL Powered by PHP Powered by SMF 1.1.19 | SMF © 2006-2009, Simple Machines Valid XHTML 1.0! Valid CSS!