SB Overclocking results?

Also, I've noticed that if I push the CPU too far and lock up and/or get a memory dump, the system is not stable for several minutes at a reduced clock speed. So I have to be very careful about testing my pull-backs. For example, I tried to hit 4.8 again last night after 4.7 proved to be stable. It failed and then I went back to 4.7. Instant failure for several minutes after. Now it's fine. I've only seen that behavior in GPUs.

Ive had identical behavior on more than one 2500k ive tested on a UD4.
 
5ghz!!!!!

It's been folding for an hour now. I had to raise the vcore to 1.36v in the bios (1.416v in cpu-z). So still within the 'safe' limits. There is one thing however, I plugged in a multimeter and the actual voltage is 1.444v -(that's .084 more than specified in the bios) - so go careful when pusing these to the limits.

The temps are in the 70's so I am happy :D
 
Anyway, I think I have 4.7 GHz nailed down, but I can't be sure yet. My system is weird. If I go too far (like 4.8), I can actually pass all stress tests for hours, but immediately after shutting down a stress test, a game, or shutting down Windows, I get a memory dump.

This has happened to me as well. I ran small FFTs for 30min, figured I was good, loaded up WoW and got memory dump 3 min into the game.
 
Same. At 4.8Ghz, it passed all stress tests (Prime95, IBT, Cinebench, 3D Mark etc), but BSODed after about 8hrs of folding. At 4.7, its stable and been folding for about 60 hours straight now.
 
If Prime95 crashes, you're not stable. You won't notice a 100 MHz drop but you will notice if your system bombs during something important.


Maybe, but at some point it is stable enough. I don't run Prime95 as an application, and I don't fold, so if it can reliably run IBT/LinX, and it doesn't crash in games, then maybe it is good enough for day to day use.
 
With subzero cooling....

screenshot013y.jpg
 
^^^Holy chit! What a loser . . . 99.7MHz bus speed . . . can't even handle a full 100MHz. Weak!

;)
 
Maybe, but at some point it is stable enough. I don't run Prime95 as an application, and I don't fold, so if it can reliably run IBT/LinX, and it doesn't crash in games, then maybe it is good enough for day to day use.

You don't run IBT/Linx as an application either but you're arbitrarily satisfied with those results. This is why I've taken OCs around here with a grain of salt for 10 years or so.
 
You don't run IBT/Linx as an application either but you're arbitrarily satisfied with those results. This is why I've taken OCs around here with a grain of salt for 10 years or so.

This isn't a competitive sport (at least not on this forum) - I'm not looking for sponsorships, and neither is anyone else. So if an overclock is stable enough for an individual's use in the apps they use/care about, then it's good enough. There isn't a "standard" of stable, which is why this seems to come up every time a new chip is released.
 
Man... you ain't got a clue. :p
What clue Manley?

Just try to set 45X multi in bios and everything else in stock:) See if you can boot it out

My 2600K wont go over 53X Multi on Biostar mobo even with -30C on cpu cooling. Guess I will get another cpu to test:)
 
Last edited:
wtf Big Typhoon? Mine is in the garage waiting for the copper scrapper to come get it LOL.

How could you even fit it on this socket?
 
It's not strange at all. According to folks at XS, Prime95 stresses different parts of the CPU so everyone's understanding of stability testing has to change for SB.

Don't know about you guys, but one of the best stability tests I've used is Ripbot264 (x264). Doing a 2-pass bluray 1080p encode will peg all eight threads on the 2nd pass for the duration of the entire second half of the encode. If my CPU's clock survives a 2nd pass encode it'll survive pretty much any real world situation.
 
Linx and IBT are not good enough for testing stability. My 2500K can both pass at 4.6Ghz, with V1.32, but instantly fails the the prime95. After passing linx and IBT left the PC idling overnight and found it crashed in the morning...
 
This isn't a competitive sport (at least not on this forum) - I'm not looking for sponsorships, and neither is anyone else. So if an overclock is stable enough for an individual's use in the apps they use/care about, then it's good enough. There isn't a "standard" of stable, which is why this seems to come up every time a new chip is released.

It's not a competitive sport and yet people seem to have no problem "reporting" unstable overclocks in these kinds of threads instead of keeping it quietly to themselves. And, yes, like you said this happens with every CPU launch. It creates an air of BS and an environment of unrealistic expectations. One thing basically everyone agrees with is if you can't even pass the most popular stability tests, something is amiss. Do you honestly disagree with that? It's not like Prime95 is an exotic, archaic stability test that only the lunatic fringe run.
 
It's not a competitive sport and yet people seem to have no problem "reporting" unstable overclocks in these kinds of threads instead of keeping it quietly to themselves. And, yes, like you said this happens with every CPU launch. It creates an air of BS and an environment of unrealistic expectations. One thing basically everyone agrees with is if you can't even pass the most popular stability tests, something is amiss. Do you honestly disagree with that? It's not like Prime95 is an exotic, archaic stability test that only the lunatic fringe run.

I'm not arguing that it shouldn't be able to pass some basic tests, but the craziness starts when people get hung up on what constitutes stable. It used to be Prime95, then it was IBT/LinX, and now with SB it seems to be swinging back to Prime95. Then the discussion goes to the length - is 20 passes of LinX good enough? 8 hours of Prime95? 24 hours of Prime95 while running Furmark? It just gets ridiculous. Let's agree that none of those tests are anything like realistic loads, so the fact that the CPU overheats and crashes after 4 hours of Prime95 is not an earthshaking failure as far as an overclock goes.

Maybe F@H should be the stability test of choice, since accuracy matters there, and it is a real world app that a lot of people use. Until some kind of testing group comes out with a "standard of stability" there is always going to be uncertainty about whether the overclock someone is reporting is "stable" or not.
 
Last edited:
going for 5ghz this weekend. and notice core temp is out of wack compared to cpu id. will have to try real temp.

5372769725_546515361d_b.jpg
 
Last edited:
I'm not arguing that it shouldn't be able to pass some basic tests, but the craziness starts when people get hung up on what constitutes stable. It used to be Prime95, then it was IBT/LinX, and now with SB it seems to be swinging back to Prime95. Then the discussion goes to the length - is 20 passes of LinX good enough? 8 hours of Prime95? 24 hours of Prime95 while running Furmark? It just gets ridiculous. Let's agree that none of those tests are anything like realistic loads, so the fact that the CPU overheats and crashes after 4 hours of Prime95 is not an earthshaking failure as far as an overclock goes.

Maybe F@H should be the stability test of choice, since accuracy matters there, and it is a real world app that a lot of people use. Until some kind of testing group comes out with a "standard of stability" there is always going to be uncertainty about whether the overclock someone is reporting is "stable" or not.

You're right when you say there isn't a "standard" for stability. That's exactly why these discussions take place.

A lot of people have been saying BFBC2 is harder on these cpu's than some torture tests. BFBC2 runs fine for me at 4.7GHz. I ran prime for 11 hours with no errors. Yet last night WoW crashed. So, in my case, running prime for 11 hours wasn't enough. People are just trying to get their machines as stable as possible so they aren't crashing in the middle of a raid or encoding a video.
 
These SB CPU's are totally overhyped and totally not worth the price of admission, right?

:p:p:p

Then the discussion goes to the length - is 20 passes of LinX good enough? 8 hours of Prime95? 24 hours of Prime95 while running Furmark? It just gets ridiculous. Let's agree that none of those tests are anything like realistic loads, so the fact that the CPU overheats and crashes after 4 hours of Prime95 is not an earthshaking failure as far as an overclock goes.

The post I originally responded to said Prime crashed after 30 minutes. Can we at least agree it's a tad ridiculous to call that stable and leave it there?
 
These SB CPU's are totally overhyped and totally not worth the price of admission, right?

so far as there o/c ability and how easy it is for them to hit 4.9ghz on air. and how they will dominate everything yes they were.
 
so far as there o/c ability and how easy it is for them to hit 4.9ghz on air. and how they will dominate everything yes they were.

Just return that overhyped piece of crap then. Having the ability to OC from 3.4Ghz to 4.7GHz with three BIOS settings changes and the ability to hit 4.8-5.0GHz with a little more work isn't all it's cracked up to be. You should've waited for Bulldozer.
 
why cause i called BS on how everyone is going to get at least 4.9ghz on air out of them?

No, because it would appear that you've been around long enough to take the yearly pre-launch "news" (hype) with a grain of salt. Anyone who's been in computing for any amount of time knows that with the release of every next gen chip or GPU, the hype is that it'll GUARANTEE you xx% increase in performance or it'll be TWICE as fast as anything else ever made.

If you bought into that hype, you have no one to blame but yourself. If I believed everything I read on the interwebs, my life would be phucked in so many ways. Take any pre-launch previews/hype you read with a large grain of salt then wait for realworld user results then make your call. Apparently you still found the results good enough to buy a 2600k so just enjoy the damned chip and what it can do for its price point instead of hanging onto that chip on your shoulder that you were the guy that called BS on SB (but owns one anyways).
 
Are you folks stable when stressing a single core? Stressing a single core with EIST and C1E enabled doesn't get the same voltage response as stressing all four cores. For example, my system is completely stable when stressing all four cores (full voltage response) but crashes when stressing a single core (partial voltage response). I don't know how to get around this without disabling EIST and C1E or bumping up the voltage excessively to compensate.
 
Fine here - interestingly the fans spin right down when 4 cores or less are used. I would say turn the voltage up a touch (but not too high!).
 
What clue Manley?

Just try to set 45X multi in bios and everything else in stock:) See if you can boot it out

My 2600K wont go over 53X Multi on Biostar mobo even with -30C on cpu cooling. Guess I will get another cpu to test:)


LOL ....Just razzing you man. No temps monitoring values on both your post though?? :D
 
Are you folks stable when stressing a single core? Stressing a single core with EIST and C1E enabled doesn't get the same voltage response as stressing all four cores. For example, my system is completely stable when stressing all four cores (full voltage response) but crashes when stressing a single core (partial voltage response). I don't know how to get around this without disabling EIST and C1E or bumping up the voltage excessively to compensate.



These are 4 and 8 thread chips. The 2500K is 4 thread and the 2600K is 8 thread. You should be manually selecting that on Intel Burn Test v2.3.

Then stand back with a fire extinguisher!!!!
 
Last edited:
Threads is how the chip is tested by programs. I'm not sure if Prime is completely compatible with these new Sandy Bridge chips. All I've seen published as updated is IBT.

OH...Damn I see now Sorry Proter. 2500K is 4 thread. Fixed! Thanks. ;)
 
Here's a test for you to run: Open Prime95. Set CPU affinity to a single core (CPU 0 w/o HT, CPU 0 & CPU 1 w/ HT). Run Prime95 Small FFTs with enough threads to get 100% load on the core. Now, unless you're manually setting the voltage, you won't get the same voltage response that you would get when stressing all the cores which may result in instability. I admit that such a scenario is highly unlikely to occur with normal usage, but is that reason enough to disregard it?
 
Here's a test for you to run: Open Prime95. Set CPU affinity to a single core (CPU 0 w/o HT, CPU 0 & CPU 1 w/ HT). Run Prime95 Small FFTs with enough threads to get 100% load on the core. Now, unless you're manually setting the voltage, you won't get the same voltage response that you would get when stressing all the cores which may result in instability. I admit that such a scenario is highly unlikely to occur with normal usage, but is that reason enough to disregard it?

Actually, I think I saw that mentioned in a couple of reviews (maybe even [H]'s?). They were Prime stable stressing 4 cores at a given overclock, but running a single threaded app would cause a crash at that same overclock.
 
Back
Top