Printable View
Wasnt trying to insult. I spent a lot of time running blend, and linx, and no not 5 passes. The idea of not running the system at a given OC because one or both of these 2 programs say it is not stable no longer makes any sense to me. I have passed them both and had issues with certain games. Its been nice not feeling I have to pass these test to enjoy my systems at whatever speed I am running, so long as it runs the things I use it for without issue.
You just need to look at stability tests from a different angle: instead of using them as proof the system is stable, take failure as a way to detect instability as fast as possible.
Stoned,Quote:
Originally Posted by st0ned
I stopped stress testing at 5 ghz.
Why?
Because I don't want to blow up my CPU at 95C loads in Prime, when the absolute highest temp it ever gets in a game is 75C. Even 8 core Deep Rybka 4 topped out at 78C.
and 2) stress testing at 5 ghz caused my good 2600k to degrade over all clock ranges slightly. 4.5 ghz was stable at 1.236v load for two months. Then, after several hours of prime abuse on repeated occasions at 5 ghz (I'm guessing maybe 12 hours total, and many, many hours of gaming sessions), it now needs 1.260v at 4.5 ghz to pass 12 hours blend. That's not something I'm happy to see, when it's my money being spent and not someone else's. Now if someone else paid for the CPU, then sure...I'll run it at 90C for 24 hours if requested. It was the priming that caused it to degrade rather fast (compared to the gaming).
4.5 ghz priming at low vcore didn't degrade the chip. 5 ghz at high vcore did (quickly) and some gaming (much slower, since the temps weren't as high). Even though it wasn't a lot, it was still some, and I don't want to see it keep happening, since as I said, the chip was stable at 4.5 ghz 1.236v for 2 months.
BC2 at 5 ghz made it degrade a bit more (so that 1.248v at 4.5 ghz became 1.260v; this took much longer to happen) but so far hasn't gotten worse. And I'd like to leave it like that.
I hate bsod's like you do, but what's the point when all you're going to do is degrade your chip trying to see if it won't crash at 95C (my second, worse 2600k gets to this if I try to prime on it at 5 ghz and 1.512v, and it will NEVER see those types of temps in 24/7 usage depending on what you use it for? (YMMV). And I had it for several days at that and haven't had a single BSOD (except once when I had offsets+c3/c6 enabled and had a BSOD in a game lobby at very light load...disabling C3/C6 fixed that instantly).
I'm done with stress testing at high vcore. Unless someone is willing to donate a chip to me, I'll only do it at low vcore.
This is my main focus :yepp:Quote:
Originally Posted by danielkza
1.260 is like my VID for 3.4Ghz... Im envious of that chip even if it degraded.
Weve mostly always agreed on everything , but ill completely disagree with your opinion and state my reasonQuote:
Originally Posted by Leeghoofd
first reason : only people who brag about sandy bridge high clocks are the people who cant prime or do a suitable stability test and like to show high overclocks 4.8 and 5.2 and say" i test my system gaming scenarios "
second reason : if a computer cant pass a torture test but can play a game what's the use of being at high clocks say 4.5 pass prime vs 5.2 prime crash ,
and the in gaming you get just 3 fps second more ? if your a bencher then i understand , but my Photoshop and decoding wont like gaming scenarios
and oh ya you can believe at the end of the day if i build a $3.600 system it better be able to do a simple task like passing prime . . .or my apple friend well say you see that why we choose apple ack ,
just my opinion but a pc should be as stable as it is fast
Stability is only upto a point where you test it, if you can prime95 for 3 days then its only stable for 3 days.. if you can do 2 full weeks then its only upto 2 weeks, to each to his own. I BSOD after 16hrs of 4.5Ghz with all voltages in auto but it never fails me with my encoding and gaming tasks, so I just went with 1.360v and got done with it. finding the lowest voltage or making it 1 month prime stable wont get me better encoding times or better FPS.
Plus you WILL see some pretty serious degradation if you try priming for 1 week at 5 ghz 1.5v, that would be nowhere near as bad as 1 week of playing Bad company 2. Why? Simple logic--the temps.
If you're playing BC2 and the temps are maxing at 75C tops, and you don't crash, A 8 thread chess engine like Deep Rybka 4 tops at 80C and doesn't crash, but doing a prime blend stability test brings it to 95C, what do you think is going to happen degradation-wise?
Furthermore, what if you managed to get the chip cool enough to pass 24 hours of prime at 5 ghz 1.5v load, and then you started crashing in games, and then due to excess stress and degradation, your 24 hour prime stable system is bsoding in 45 minutes, when you start panickng and decide to run Prime again? And unfortunately, nothing besides raising the vcore even more will help...(you tested VTT, PLL, etc...)
Guys, its perfectly OK to disagree, but its *NOT* Ok to tell someone else what to do with their own money! Please do you want to do with your money and use your cpu and stress or game or fold on it as you see fit. But don't try to say that someone else shouldn't do what they're doing just because you're doing something different, while you can give recommendations on what works for you.
Remember also that VID is not the same thing as vcore. 1.52v VID does NOT mean that the CPU is running in its electrical limits at 1.52v LOAD voltage. If you read Intel's documents carefully, you will see that they NOT account for loadline calibration, and they assume you are running the CPU according to full VRD 12 specifications, which CALLS for vdroop by design. This is right in the documentation and proven by charts. And 1.52v VID (set in BIOS) is actually about ~1.39-1.4v of full load voltage. Please consider that, before you go crazy with stress testing at high vcore. In fact on ALL of the intel CPU's released in the last several years, the max VID has NEVER, ever been the max vcore. This is why the core i7 had a max VID of 1.55v--if you add vdroop to that, this turns out to be about a 1.42v vcore voltage (most boards will droop over 0.1v at these voltages and clock speeds). Also notice that there is a loadline slope, showing how vdroop is supposed to work.
LLC on motherboards, while catering to us overclockers, go beyond Intel's specifications.
Something to think about.
I totally agree with you Falkentyne and I run all my systems according to Intels specs with wdrop.
Here's an example.
Vcore is set in bios to 1,43 but drops to 1,344 under load.
http://img571.imageshack.us/img571/6900/70435633.jpg
Been running at this speed for four months running F@H and WCG and the chip still does 5,65GHz while I'm benching it.
well i think on most part , that there should be a better program to torture test than prime or linx , something made for sandy bridge or even better a sensor in the bios that say exactly what fails , but for now there's not and prime is it
I agree that everyone should test there system according to there choice of stability , but ill disagree a proper test of a system is a game test :up: overnight with sandy bridge everyone became a leet overclocked and have no proof there system is stable , but say hey i can run 54 all day and night :D
without a standard test who is really reaching what ? :confused:
Droop is necessary but i hate it on my ud7 with llc 1 ,grrrr i have to run a higher idle to compensate
Seeing as the name of this Forum is XtremeSystems Forums and not Xtreme24/7stableSystems Forums, I think the debate on proper testing methods for 24/7 stability should be left at the door. Just my opinion.
If someone wants to run blend all day, have at it. If someone runs 32m and says "good enough for me" then good for you, enjoy. Who really cares if someone elses system meets their own personal standards for stability, I dont. My current 24/7 settings are stock, I have a couple good chips now, a 56x and a 57x that makes it to windows load screen at 58x on water, it just needs cold I think/hope. I wont be stress testing either of these chips with blend or linx any time soon.
good wording on that one OC NUB lol . . .Quote:
Originally Posted by OC Nub
I am in love with my new 2600K L046B028. 52x pll overvolt disabled, 57x pll overvolt enabled. 5ghz at a hair above 1.4V, I am finally done binning I think.
L046B028 seems to be a pretty strong batch, my last one did 56x max multi. Got em from the local Microcenter.
I totally agree on this, I mean why push the hardware to the point of breaking it and to just prove how large the e-peen is.Quote:
Originally Posted by OC Nub
Try using offsets and C1E, but turn off C3/C6 since it doesn't even do anything except park cores at idle. I did this and now my idle vcore is 1.12v, my light load vcore is 1.52v, medium load is 1.512v and full load is 1.5v. All of this with LLC1. Sort of like LLC2 in reverse, except you don't degrade your CPU by leaving it at idle and something like 1.58v.Quote:
Originally Posted by the_real_7
After seeing just how well offsets work, I'm hooked thanks to having the chip at 1.12v at idle instead of 1.58v. C3/C6 cause problems with light load...had a BSOD just in a gaming lobby from it (insufficient vcore at light load/one thread). Disabled C3/C6 and it never happened again.
Falkentyne - nice find there :up:
Demo and Nub - each to his own BUT if you use your cpu for gaming as well then its nice to have a 24/7 clock where its nicely overclocked but not at high voltage, so personally I will prime a lower voltage clock but also test in game, as demo said often a pc will crash in gaming use where it wont in stress testing, but I still stress test first. I think linX is more stressful than prime though right?
and testing 24/7 clock doesnt usually prove any e-peen points
:up:
Im really not sure which program is best for testing SB. I have spent very little time stress testing this platform. I would always do long blend runs in the past and if it didnt pass, even if it failed at 12-14hrs I wouldnt call it stable and would retest. I ran LinX mainly to see how many chips I could get in the 4.5GHZ LinX thread, but other than that I didnt really care if it passed or not.Quote:
Originally Posted by aussie-revhead
This is the first time (with SB) everyone has had the chance to taste 5ghz+ without any type of extreme cooling, including myself. Im just enjoying it while it lasts because it wont be forever. I would hate to not play at say 5.6ghz because a program says Im not stable. I just like seeing the high scores at the end of the test. I use this rig for fun so it doesnt have to be 100% stable.
Some cant afford to not be stable and I understand needing an oc that you dont have to worry about crashing or corrupting data. But for the rest of us who dont really care, this is a lot of fun.
The difference in opinion on 24/7 stability is this:
People that lobby for prime and LinX stress testing understand that 5 minutes or even 5 hrs is not sufficient to fully detect instability.
People that lobby against prime and LinX systems say "it was stable in prime and LinX but not in games" but they never say how long they ran prime for. They probably ran it for 1 hr and called it quits because they don't have the patience to wait 8 hrs +
I stress all of my systems to minimum 8hrs of prime. I am also a firm believer in LinX with AVX. My system that was 7.5 hrs prime stable and standard LinX stable (20 runs+) failed LinX w/ AVX after 3 passes.
If you have a system that faults under any circumstances, it is not stable! The best way to test this is by performing long duration, heavy load testing.
I understand that this is just my opinion - but so many people fail to state how long they ran prime or LinX and whether or not they used AVX LinX...
And that, too, is part of the problem. Anybody can boot a 2500K or 2600K into windows at 5GHz with enough voltage - and everybody does. Then they fail prime and LinX because temperatures are 80C+...Quote:
Originally Posted by OC Nub
Then, in a desparate attempt to show everyone that they have a good chip, they claim that prime and LinX are not good stress testers because they are stiill able to play games with their chip... :shrug:
L051B468 *A1014 56X and probly higher with more cooling...This one is hot, took my old SS to positive cpu core temp in vantage:eek:
Nice, chip looks good, 8 threads runing no problem.Quote:
Originally Posted by Dumo
Quote:
Originally Posted by dr_sharp
The problem seems to be when person A takes the time to stress test using linx and blend to assure his clocks are stable. Then person B comes along that didnt go through all the trouble, or didnt pass but still runs his rig at high clocks because he can game or do whatever without issue, so he says "stable". Person A seems to take offense to this but, it isnt a contest. Hats off to the guy who passsed all tests with flying colors, this Bud's for you :up:, but no need to frown upon person B because he didnt, or chose not to go through the headache, or has a chip that wont pass all the "required" tests but is still able to enjoy his high clocks. Its not person A's responsibility to project his standards onto person B, cant we all just get along lol.
:ROTF:Quote:
Originally Posted by OC Nub
Yep, I think you are right, basically a disagreement in terminology. :)
Quick note with c3/c6 disabled, my idle temps are about 3c higher.
Id like to add that a lot of people using manual voltages and dont really understand the low/medium load crashing since they only have like 2 voltages.. Idle and loaded.. nothing in between.