Ivy Bridge 3570K Testing, Opinions, Results, New Bios, 4.5Ghz At 1.236v

[BallaTheFeared]

LinX is worthless since it only stresses FP, IMC, and RAM. I guess if you were using programs that ran pure AVX and/or FP code then it would be moderately ok. But you're still not hitting cache, which on SB is overclocked directly with core speed

 

[Don Karnage]

Guys want any benchmarks ran? I'm off today so i actually have time to post things. I'll run anything that can easily be downloaded

 

[JAG87]

Guys want any benchmarks ran? I'm off today so i actually have time to post things. I'll run anything that can easily be downloaded

Can you find the highest clock at stock voltage (no offset in BIOS) that is prime/linx stable? Even just 5 minutes stable, just to get a rough idea.

Thanks

 

[Grooveriding]

LinX is worthless since it only stresses FP, IMC, and RAM. I guess if you were using programs that ran pure AVX and/or FP code then it would be moderately ok. But you're still not hitting cache, which on SB is overclocked directly with core speed

Sure. You should run some prime and some linx to test stability.

Trying to dismiss linx as worthless because I labeled cinebench worthless is pretty silly and being argumentative for the sake of being so. Linx is one of the tests you should be running to test stability on your overclock and cinebench is not.

If I can get to my desktop without a BSOD, I can generally run a pass of cinebench... Good luck making it through linx under those same conditions.

 

[Edrick]

LinX is worthless since it only stresses FP, IMC, and RAM. I guess if you were using programs that ran pure AVX and/or FP code then it would be moderately ok. But you're still not hitting cache, which on SB is overclocked directly with core speed

LinX does run AVX code. Why else do you think SB saw jumps into the 100's (Gflops)? I feel LinX is a must to test for system stability.

 

[LagunaX]

Hey Don see if you can do a Prime95 just to see what your temps are.
Try to see if you can lower your vcore closer to 1.30 also.

 

[BallaTheFeared]

Sure. You should run some prime and some linx to test stability.

Trying to dismiss linx as worthless because I labeled cinebench worthless is pretty silly and being argumentative for the sake of being so. Linx is one of the tests you should be running to test stability on your overclock and cinebench is not.

If I can get to my desktop without a BSOD, I can generally run a pass of cinebench... Good luck making it through linx under those same conditions.

I should run whatever it is I do with my cpu to test my own stability needs. Prime95 and LinX aren't helping me find gaming stability, all they're doing is locating prime numbers and crunching floating point, not something I need to max out my cpu on. Not to mention when using offset the voltage levels will be different during gaming than during a high stress program such as Prime95 or LinX.

It's worthless unless the programs you expect to use follow the same type of stress pattern, do they? Great use it, if not who cares, unless you're trying to use a generic form of stability for website validation, I guess, or max temps that you'll never see in any other application :whiste:

Cinebench isn't a stress test it's a benchmark application for rendering, which oddly enough doesn't stress your cpu like Prime95 or LinX so being stable in either is meaningless if you're using your cpu to render with Cinema 4D. If your cpu can render 24/7/365 does it matter if it's stable in Prime95 or LinX? No, odviously which is the point I was trying to make before you took it personally and rolled your eyes.

Great I'll remember that the next time I crunch AVX all day, note to self, LinX is a good stress test if you're doing pure floating point workloads with memory usage exceeding cache thresholds.

You do know SB cpuNB (cache, IMC) overclocks itself with core overclock right? And you do know that cache is purposefully bypassed in LinX and Prime95, right? You do understand that a cpu works different depending on the applications needs, right? Generic tests like Prime95 and LinX are only good for testing stability in applications like them, of those I can name off the top of my head there are none like either.

So what are you really "stable" for?

 

[Grooveriding]

snip

So you don't care if your system is not 100% stable so long as you can run Cinebench and play a game.

Fair enough, to each their own. I prefer to run a system that will not BSOD under any circumstances and is stable no matter the circumstances. You prefer one that can only perform allowed functions on your list of must-haves, and not others that will cause it to crash...

Random BSODs ftw.

Thank you for enlightening the community on the worthless nature of Linx as a stability test. Going forward we will all do Cinebench runs and call it good.

 

[Edrick]

I should run whatever it is I do with my cpu to test my own stability needs. Prime95 and LinX aren't helping me find gaming stability, all they're doing is locating prime numbers and crunching floating point, not something I need to max out my cpu on. Not to mention when using offset the voltage levels will be different during gaming than during a high stress program such as Prime95 or LinX.

That really does not makes sence. Unless you do not care if your system is unstable for everything other than games. And if that is the case, then all the power to you. But considering you are one of the very few here to think that way, why bother trying to argue your point?

 

[BallaTheFeared]

People passed Prime95 and LinX all day with SB, then crashed at idle... So much for assured stability from those programs.

Nice little passive aggressive comment at the end there, all it did was show how everything I said went right over your head though

 

[pelov]

So you don't care if your system is not 100% stable so long as you can run Cinebench and play a game.

Fair enough, to each their own. I prefer to run a system that will not BSOD under any circumstances and is stable no matter the circumstances. You prefer one that can only perform allowed functions on your list of must-haves, and not others that will cause it to crash...

Random BSODs ftw.

Thank you for enlightening the community on the worthless nature of Linx as a stability test. Going forward we will all do Cinebench runs and call it good.

There's no need to straw man. He's absolutely right. For someone who won't stress their PC to such levels the value of LinX or Prime95 provide little value. I've had cases where I've ran prime95 for 3+hours only to start a game and crash suddenly. Such programs serve a particular purpose but they do so by addressing the extremes to which most users will never get to, thus aren't necessarily an accurate representation of a "stress test." The stress I put on my PC is different then the one you put on yours.

This isn't to say that I don't use them. I do. If I want to find the load temps, power consumption figures. For stability I run LinX/Prime95 and then do a hours-long gaming session. If it survives them all then I know it's fine at its current settings. But you know what? I also know that I can safely push it up a couple more notches in vcore and bump the multiplier up and still be as safe despite not being able to pass Prime95 or LinX

 

[Edrick]

People passed Prime95 and LinX all day with SB, then crashed at idle... So much for assured stability from those programs.

...and Cinebench would have solved that no problem LOL.

 

[BallaTheFeared]

That really does not makes sence. Unless you do not care if your system is unstable for everything other than games. And if that is the case, then all the power to you. But considering you are one of the very few here to think that way, why bother trying to argue your point?

Do you have statical data to back up your claim? Otherwise I'm not sure why you're trying to argue with my perspective using "the masses" without any hard empirical evidence to back up the claim.

My point was stability only matters is so far as what you're actually using your computer for. Neither Prime95 nor LinX provide anything close to a real life workload, and therefore stability in either is of no consequence.

Perhaps you missed it, but when sandy bridge first came out LinX was easy to pass (still is if you can manage the heat), games would crash, and prime95 was only moderately decent if you used blend, anything else was a waste of time. However neither Prime95 nor LinX were able to stress the cpu enough to rout out the idle bsod errors, nor were they capable of exposing issues within the cache which was quickly exposed by games.

Stability is a matter of what you use your computer for, that is all I'm saying. If you run linpacks all day, or find prime numbers in your off time, by all means, use those programs. If you watch videos, render, encode, transcode, game, or whatever else it may be you do, use those programs for stability testing.

I have never once stated Cinebench was a stability test, not once. Unless you're using your processor to render only, then it would be enough to try longer duration tasks, wouldn't it?

 

[Edrick]

Do you have statical data to back up your claim? Otherwise I'm not sure why you're trying to argue with my perspective using "the masses" without any hard empirical evidence to back up the claim.

You are not backing up your claim with stats either. All you said was many people passed Prime95 and LinX all day (24 hours?) and yet crashed idle. I never heard of that, so please present cases. (excluding cases where people did something stupid in their BIOS that was quickly resolved.)

My point was stability only matters is so far as what you're actually using your computer for. Neither Prime95 nor LinX provide anything close to a real life workload, and therefore stability in either is of no consequence.

Again, if you only care about your computer being stable for games, then all the power to you. Thats your decision and yours alone. My point is most people care if their system is stable for everything, and in that case, LinX and Prime95 are very important.

I am not argueing that you should change your method. But you are coming across as if your method is the only correct method. And that is not the case.

 

[Grooveriding]

There's no need to straw man. He's absolutely right. For someone who won't stress their PC to such levels the value of LinX or Prime95 provide little value. I've had cases where I've ran prime95 for 3+hours only to start a game and crash suddenly. Such programs serve a particular purpose but they do so by addressing the extremes to which most users will never get to, thus aren't necessarily an accurate representation of a "stress test." The stress I put on my PC is different then the one you put on yours.

This isn't to say that I don't use them. I do. If I want to find the load temps, power consumption figures. For stability I run LinX/Prime95 and then do a hours-long gaming session. If it survives them all then I know it's fine at its current settings. But you know what? I also know that I can safely push it up a couple more notches in vcore and bump the multiplier up and still be as safe despite not being able to pass Prime95 or LinX

It's really not a straw man though. I'm just indulging and pointing out the irony of trying to call linx a worthless benchmark for stability/temps - which as you yourself have said it isn't - because of my initial branding of cinebench as such, which definitely is worthless on both counts.

Question, you take your new shiny CPU from Intel home and toss it in your mobo. The expectation is that you can run any and every properly coded program on that CPU without a BSOD as a result of the CPU's stock performance, all other components in your system being stable ? Yes/No ?

Intel is not selling us CPUs that are only stable so long as we don't subject them to certain programs. Many users expect the same out of their overclocked systems as well and ensure stability. I'd prefer not to get hit by random BSODs because I'm trying to see 100mhz higher numbers in CPUz just for the sake of it and making stability compromises for such.

Regardless, it's wasted forum space. I pointed out accurately that for us to make judgments on the 3570k posted, that cinebench is a worthless indicator of system stability and load temperatures whereas linx is a much better indication of such. Then the thread gets spiraled off into wasted posts explaining how linx doesn't cover all variables. Greeeeeat, the thing is linx does cover plenty of bases in stability and all bases for temps. Hence why it's used to test stability and temperatures whereas cinebench is worthless for both. As I originally said.

Now we have a tangent just for the sake of being argumentative and I'm not going to contribute to petty nitpicks that pollute the forum.

 

[BallaTheFeared]

You are not backing up your claim with stats either. All you said was many people passed Prime95 and LinX all day (24 hours?) and yet crashed idle. I never heard of that, so please present cases. (excluding cases where people did something stupid in their BIOS that was quickly resolved.)




Again, if you only care about your computer being stable for games, then all the power to you. Thats your decision and yours alone. My point is most people care if their system is stable for everything, and in that case, LinX and Prime95 are very important.

I am not argueing that you should change your method. But you are coming across as if your method is the only correct method. And that is not the case.

No you're right, I'm just using my personal experience of running a SB chip from stock to 5.3GHz. However I'm not attacking your perspective, am I, I'm merely stating mine.

I've only owned my chip and have been overclocking it for 16 months now, what is your experience and qualifications?

Never heard of the idle bug?

http://www.overclock.net/t/1053162/fixing-the-sandy-bridge-idle-bug/0_100

http://forums.guru3d.com/showthread.php?t=349315

http://www.xtremesystems.org/forums...he-fix-for-Sandy-Bridge-idle-light-load-BSODS

The CPU is prime95 / Linux stable for hours but will BSOD with either 101 or 124 usually watching video or surfing (which seems to be the most common I've found)

Anyways you're completely missing my point, perhaps you should go re-read what I have said as it seems to have fallen on deaf ears.


We used to have a passing joke for sandy, needs more vcore… Perhaps you've heard it?

 

[pelov]

You're pretty much right on all accounts. Cinebench is worthless as a stress test and people don't want random BSODs so these programs, LinX, memtest, Prime95, etc, clearly provide a purpose.

Still, though, the term stability is itself a very relative term. Where someone runs their PC as a bitcoin or F@H machine 24/7 or use it as a workstation I would only run a few games as the toughest task. LinX and Prime95, though admittedly I do use them, serve little purpose to me with regards to system stability. Not useless, as we both agree, but I wouldn't exactly say they represent actual system stability per my usage =P

 

[Edrick]

I've only owned my chip and have been overclocking it for 16 months now, what is your experience and qualifications?

You got me there. I have not owned the same SB chip for 16 months. I bought me a 2600 the day they were released (Jan 9th, 2011). But I only held that for 4 months and decided I wanted to play with the 2500K. I had that until the day the 3820 was released and I moved on to my current setup.

No you're right, I'm just using my personal experience of running a SB chip from stock to 5.3GHz. However I'm not attacking your perspective, am I, I'm merely stating mine.

I do not run my cpus that fast for the simple reason I want to be 100% stable at everything. Who knows what I might throw at my system from day to day, and I dont want to ever worry about it crashing. But again, thats just personal opinion.

I respect other people's opinions. I just do not take kindly to people putting down my opinions.

 

[n0x1ous]

Question, you take your new shiny CPU from Intel home and toss it in your mobo. The expectation is that you can run any and every properly coded program on that CPU without a BSOD as a result of the CPU's stock performance, all other components in your system being stable ? Yes/No ? Intel is not selling us CPUs that are only stable so long as we don't subject them to certain programs. Many users expect the same out of their overclocked systems as well and ensure stability. I'd prefer not to get hit by random BSODs because I'm trying to see 100mhz higher numbers in CPUz just for the sake of it and making stability compromises for such.

Well said, good analogy +1

 

[LagunaX]

:|
Ok guys back on track.

The Cinebench runs are useful just to see the capability of the chip but by no means for stability settings for 24/7 crunching or other hard core utilization - this we all know already.

From what most infer from "stable" or "stability" usually is associated with passing a stress test of some sort.

Otherwise ppl usually disclaim,"gaming stable only," or something to that effect, as ppl run WPrime and other such "benchmarks" that do not say much for the stability of a system.

Aftermarket "custom boutique" PC Gaming builders that sell overclocked systems definitely linX/IBT/or Prime95 stress test their systems before selling them for sure - otherwise the returns or complaints would be enormous!

In my Sandy Bridge experience, I used to run linX or IBT with SP1 and AVX as that was the test everyone else was running - however I would still get BSOD's every once in a while.

When I switched back to Prime95 custom blend utilizing 75% of your ram (say 6 out of 8gb, leaving the OS 25% for background functions/programs) I rarely ever saw a BSOD again.

Just my 2 cents :whiste:

 

[Tempered81]

Thank you for enlightening the community on the worthless nature of Linx as a stability test. Going forward we will all do Cinebench runs and call it good.

hahahah.

Guys want any benchmarks ran? I'm off today so i actually have time to post things. I'll run anything that can easily be downloaded

Can you find the highest clock at stock voltage (no offset in BIOS) that is prime/linx stable? Even just 5 minutes stable, just to get a rough idea.

Thanks

Any progress on this Don Karnage? I'm ready to see some of you LinX Temperature testing too

 

[Don Karnage]

Whats a good program to do cache latency besides sandra?

 

[Don Karnage]

 

[IntelUser2000]

Whats a good program to do cache latency besides sandra?

Everest, but it only shows single threaded results. Sandra shows full multi-threaded one.

 

[frozencopper]

With 2X2gb BBSE ES CPU EO stepping 8