Well, you could be "under the down under"Quote:
Originally Posted by dinos22
Out here they find them buried outside Las Vegas but damn, you guys got the whole of the outback to "lose" someone in!:rofl:
Printable View
Well, you could be "under the down under"Quote:
Originally Posted by dinos22
Out here they find them buried outside Las Vegas but damn, you guys got the whole of the outback to "lose" someone in!:rofl:
IBT tests like prime 95 without the IMC stress. It's why so many ram guys love to use it for reviews and call it stable......
If it's not blend 3 hours plus it aint stable.......IMC wise ;)
Of course running any of this crap is harmful to your system especially when running the edge of safe voltages.
Also I agree that crunching can pick up :banana::banana::banana::banana: that not even stress tests can, I ran a boinc system way back in the day for dave on phase 24/7. It required more fiddling around to keep it running and error free 24/7 balls to the walls than any normal rig I have run daily.....
Side note, I know how you love car comparisons.
Buddy rips his cylinder head off, oilpan and connecting rods, and removes pistons. Hones out cylinders in car, installs new parts and cleans and reassembles it the best he can in a car.....10K later he tosses a main bearing.
Moral of story. There is no true substitute for time.
Had he built his engine properly like mine he could rev his 2.0 single overhead hampster to 9500 rpms like me and still be driving and dragging it 20K miles later.
Back to hibernation for me, see ya in a few years or so.
:rolleyes: Your're killing me MM. This was my go to app for stability testing. :( I've never gone past 20 passes. Temps always monitored and less than 75C 1366 and 1155. 10-15 passes usually prove WCG stable. Will that really kill a chip???? :confused:Quote:
Originally Posted by Movieman
My belief based on what I know is that IBT slowly degrades the chips and as such is a mistake to use it.Quote:
Originally Posted by WFO
You mention WCG. When I set a machine to run WCG I look for that sweet spot between high voltage and high overclock.
When I set up the SR2 last year with the X5680's I could run 24/7 100% load at 1.35V with temps in the low 60's.
Now I could also run that system with 1.4V at 4400 and still within what I consider safe temp margins( 65C max)
but the increased elec draw for work done wasn't worth it.
For testing I did push that machine to 4600 with 1.4v using WCG and it was stable with temps right at 65C
This was with TRUE's and 113cfm Delta's
Bottom line is test with the apps your going to run every day as each has different factors.
Noted. Thanks for the quick reply! :up:Quote:
Originally Posted by Movieman
You do not threaten my account TROLL (I did state this thread only)!Quote:
Originally Posted by Movieman
Advise you shut your mouth!
My advise to AG/OP, report him for derailing your thread which is about your kind FREE work!
Not really, he was not called any names when he still came in spouting more BS and crying saying he had been called names, so his amount of posts derailing this thread is TROLLING!Quote:
Originally Posted by dinos22
Post count or any donations given in past do not change that fact and also I knew his previous post was not his last for the holidays as he just cannot help himself!
And running your CPU at stock also slowly degrades it, running overclocked also degrades it a big faster but still still last you more years then you will want to use it by times its old tech unless your stupid and put very high VCore/PLL through it added to not cooling it well!Quote:
Originally Posted by WFO
Oh-oh, I know the difference between mud and crappola. Jeesh, at least you can't say you've not been warned.Quote:
You do not threaten my account TROLL (I did state this thread only)!
Advise you shut your mouth!
My advise to AG/OP, report him for derailing your thread which is about your kind FREE work!
Anyway, back on-topic: As someone here has mentioned, under normal operation, you have slow degradation of your CPU, as it is subject to electron-migration. If you start increasing the Vcore, you accelerate this condition in a CPU. Now, with pretty high Vcores, as well as subjecting the CPU to stability-testing, by using IBT, as an example, you may accelerate electron-migration to such a point that instead of having a CPU last 20 years+ under normal conditions, you can shorten this to mere months.
We have seen enough of the 'sudden-death' cases to realise this is a real concern.
So, humeyboy, show a bit of respect to guys like Dave (Movieman), who has been around the block more times than quite many of us here, and heed his advice, as it's good advice. Behaving like a spoilt brat is not conducive to good relations, and you can pretty quickly find yourself out in the cold.
Let's keep it civil.
ps: Dave, you can sent Dinos to Antarctica - I think the Aussies are laying claim to a 3rd territory Down Way South :rofl:
You were warned but I sort of knew you'd be back and use the word so realistically you've banned yourself.Quote:
Originally Posted by humeyboy
I've seen this before. It's what I call the marytr syndrone.
When warned the person feels compelled to push the matter even though they've been told they will be banned if they do so.
Done
My apologies to the rest here.
I repeat, my only purpose in posting in this thread was concern for the members investment in their hardware and my comments were based on good info from
two different engineers at Intel.
I'm grateful for that, movieman.
I still remember when Linpack binaries first appeared, and I think they appeared during the beginning core 2 days, and there was a note saying that these binaries were used for *internal* CPU testing. But someone or something requested that they release them, and then came the "new stress tool released: now find if your cpu is truly stable! more accurate than prime" etc...
While I do understand Agentgod's point that this was used to be a time saving mechanism, I also fully understand the other person's post in that he said "there is no substitute for time."
I guess a question I can ask Movieman AND AgentGod is:
What is more harmful to a 1.45v 5 ghz CPU?
5 passes of Linpack/LinX/IBT, or 8 hours of prime 95?
Which will cause more degradation?
The problem of course is, even prime 95 can degrade cpu's if they are running highly out of specification. I've already degraded two 2600k's with prime, and neither one of them even saw a linpack run, except a quick 5 loop test at 4 ghz stock voltages, which was safe, and that was a gflop test. I'm lucky that I can still do 5 ghz at 1.45v. And since I'm a gamer, after all, I think some good old Black Ops and Battlefield 3 works better for me than burning my CPU up with hours of prime.
Of course, for folders, who crunch data that needs to be accurate, prime can be a good test, but I STILL remember a post on the [H], where a person who folded 24/7 at 5 ghz 1.5v, had to reduce his overclock to 4.6 ghz after 3 months, due to degradation.
Degradation has happened to enough people now, so that no one questions whether it's "real" or not (unlike that famous GNDS thread, where many of us were called incompetent overclockers, until the problem became so widespread, no one could deny it anymore).
I guess this is all old hat now. The BIG question is how Ivy Bridge will overclock, and how well the 22nm process tolerates voltages over long term. And of course, no one will know without guinea pigs to test their processors....
That would IMO depend on the amount of current the application draws and how much the degradation is accelerated not only with overcurrent but time. Should be interesting to see how prime95 performs when fully optimized for AVX. BTW AFAIK Linpack is just a benchmark to measure throughput of floating point computation whilst giving an indication of instability and not as a bining tool.Quote:
Originally Posted by Falkentyne
While increased voltage will produce more current not many people seem to be interested in how much that current is but just interested in vcore. I would think overvoltage (a voltage that causes a junction to breakdown) will kill a cpu instantly while overcurrent would be more subtle by degrading and once degradation starts will only get worse with continually excessive load before finally failing.
So maybe the real question is how much current is too much, how far past maximum can you push and still have a decent life expectancy. Temperatures are likely to affect results too.
Movieman is beyond hilarious.... so im not even going to bother with my explanation, some peoples' "beliefs" are just.... ...
True and Valid?
I fully agree with everything Movieman says. If you don't agree with him, then go blow up some chips on your own with your 8 hour linpacks. We don't need to hear about it. Seriously.
You know, insulting a mod of a forum is not the best way to keep membership here....
I think that guy didint realize who Movieman was and that's why he went over his head... :)
Anyways, recently i started using this program, regardless of the dangers and i have also upgraded my RAM to 8gb from 4gb. Just by using more RAM i have noticed that my stable overclock from before wasnt stable when i used 6.5Gb ram with IBT. Now i am down by about 30mhz in order to stabilize 7 runs at 6.5GB ram (110Gflops/run). Also i found that if i stabalize my CPU with Prime 95, IBT would instantly blue screen with 101 error. IMO 5 runs per session of IBT in the course of 3-4 days(5 runs per day) is a better testing strategy for stability. This way you dont put a massive stress on your CPU at once, but rather spread it in a few days. Furthermore, the hardware would not "cook/degrage" as much and it would give you some idea of your stability. Same thing with Prime 95, it is better off to test for 1 week (30 minutes a day) rather than a 24 hour run that would damage your expensive CPU. If there is some sort of instability of the CPU it might not show up in the first day but it would eventually show up later on. ;)
my 2cents
Dave you sorta promised the community to put Dino down ! Just do it !Quote:
Originally Posted by Movieman
Maybe AG can make a normal version with loops selection to eg max 20 and an Extreme one clearly stating the risks of frying CPU, motherboards plus an inbuild-no-nagging tool on forums...
Merry X-mass and a happy New year.... darn them Aussies are gonna beat us belgians again by partying a day earlier... Life is so cruel..
I run nothing less than 2000-3000 iterations depending on memory size just for the hell of it and then drop 100Mhz. And above 1.200V may degrade the chip doing so, but <1.200 should be fine.
Using IBT 2.53 to stress test my i7-2600K. It's stable for 10 runs on Standard and is Prime 95 stable for 9.5 hrs. But if I try IBT at High or Very High stress level, it fails after the 2nd run. Should I be concerned about my system stability?
Yes.
I for one took MM's advice to heart. :yepp: Instead of trying to get 15-20 IBT loops at Max stable, I just go for a 5 loop standard run. If that's stable, I go for a 5 loop test on "High." If that is stable, it's usually good enough to prevent BSOds in WCG or F@H. After that it's just about adding enough more v-core to prevent computation errors on the WUs. :) Quick and dirty crunch/fold worthy OCs in an hour or less. :D BT is valuable and I appreciate AG's work.!!! :up: It used to take more than twice as long with P95. :(
Ummm... unless I'm suffering xtreme memory loss, Fugger works for Intel. How long have MM and Fugger been friends??? How many industry contacts does MM have? :shrug: I wish I knew at least 10 more people like MM!!! I'd probably have to go back 50 years to find them. :rolleyes: People that know right from wrong and tell the truth regardless of whether it is popular or will take a dollar out of their pocket are very rare today. ;)
(Don't let that go to your head MM! You are on the FBI's most wanted list of cell phone battery killers! :stick: :rofl: )
Thank you..Quote:
Originally Posted by WFO
Updated to v2.54!
- Linpack binaries updated (7-20-2012) + patched for AMD CPUs
- Improved effectiveness of Xtreme Stress Mode
:up: This thrad neds to be stickied.
^ Agree.
You probably have a memory issue. Very High tries to use all the memory it can get. Do about 10 runs of memtest86+. I know 10 runs seems like overkill, however i have seen where errors did not show up until the 9th run.Quote:
Originally Posted by wongnog