Originally Posted by alancsalt
I remember when OCing the E8500 on an X48-DQ6 and a bunch of guys had a mad thread going (bit like this one) where they were enthusiastically trying to get a stable 533 FSB on the EP45-UD3P. Well one of them got his to pass 36 hours of Prime 95 only to have it crash next day at idle (No Cstates or Speedstepping enabled), and the response was that "idling is the new stress testing!
I used to run Prime for ages too, pass without failing, and then have games like GTA4 crash the OC. Don't have quite as much faith in stress testing as I used to.
LOL! man ... same feeling here ... at certain clocks (mostly higher ones), offsets can be a light load stress tester
hence why usually stabilizing the OC on offset calls for a little more vcore than with a fixed voltage. not a real problem though ... until I tried stabilizing 5.0GHz on my 3770K to be rock solid for daily running on offsets ... impossible so far after nearly a week of tweaking ...
It's actually my first board supporting offsets. In past I used only EVGA boards and those are fixed vcore only ... I had to settle for 4.9GHz on offset and realy all automatic settings and no problem, once I go to 5.0GHz it starts randomly BSOD'ing (124) on light load, anywhere between 5 minutes or 22 hours from boot up LOL!
My latest impression is that actually it's the BIOS programming responsible for this behavior (lack of offset tweaking after certain multiplier is passed), why? because on older BIOS for my board it takes up to 1.52V offset vcore (some number like +0.340) to be mostly stable at 5GHz while on latest BIOS I can reach the same level at only 1.48V actual vcore under load (offset +0.295 or so). ... Mind that with fixed vcore it takes only 1.46V vcore to be stable at 5GHz on either of BIOSes ...
Originally Posted by VonDutch
if offset is done right, there should be no problem with oc's,
i always use offset, never had a problem, maybe needs some more
fine tuning then fixed vcore, but thats about it
with my 4.8ghz oc i used 0.160V offset, i thought that was to much tho,
but anything i use(d) between 0.020 and 0.050V offset ..no problem at all..
yea, best is to do a bit off all, play games, run ibt, prime and cinebench, even idle a while ..lol
maybe a idea for some programer to create a program that can run different load,
and do some stresstesting like prime
but still, 24H prime is a very stable oc,
and upping vcore a 1 notch or 2 after that, if possible, makes it even better
I thought so too, but it seems to be true for my board only up to 49x multi, above it is like forget it
I stress test the same way, CIne first, then IBT quick runs, then some max mode runs, some prime, gaming, daily using, etc. if no WHEA ever then it's all good
However, I am not a fan of running Prime95 for 24hrs ... got better things to do with my system
Originally Posted by kgtuning
i agree, I run IBT like crazy then prime on blend for 24hrs and game then I call it a "good" overclock.
try cinebench, it takes only a minute and sometimes proves more stressfull than IBT and Prime95 both together LOL!
Originally Posted by snelan
What are you're thoughts on just using Cinebench for stability testing? Some 12h stable Prime OC's have failed in seconds in Cinebench for me.
Also, is 1.36VCore for 4.7GHz a little high? I'm thinking I need to adjust something else, considering it only takes 1.315VCore for 4.6GHz.
I have tried:
47Multi @ 1.36VCore
1.2V CPU VTT
1.6V CPU PLL
No other voltage changes, aside from the LLC adjustments/etc.
Cinebench is my favorite stress tester
, usually when I make something Cine stable (with no WHEA in logs) then I have no trouble with any other stress tester or gaming or anything .. unless we are talking about offsets at 5GHz+ and light load BSOD LOL.