Intel Burn Test completely unreliable

K

kazaakas

Limp Gawd
Joined
Feb 8, 2012
Messages
144
Hello guys, I recently came across IBT/Linpack after years of using P95, and at least for me it seems IBT does a really bad job at checking if the system is stable, even though its commonly believed IBT is the heaviest stress testing program since it makes the system reach higher temperatures than any others.

It seemed just too good to be true, checking out the stability of an overclock in less than an hour...

Though, with IBT, the lowest Vcore I can run the test for about an hour with (didn't try longer) is 1.285V on 4.8Ghz.

Though, a few apps were randomly crashing afterwards and I didn't trust the stability of the system at all, so I ram prime95 and my system just powered off after like 15 minutes (didn't get an error though and my power supply is wayyyyy sufficient, I haven't even received my gpu yet..)

Now even with a whopping 1.305V prime95 still crashed after 6 hours!

So I'd advise anyone to run more than one type of stresstest and would like to hear other peoples experiences with prime95 over IBT. Did prime95 run stable for half a day on the minimal voltage you needed for IBT or did you need more?
 
OCCT linpack test is what I use. Is fairly accurate to a fully loaded system.
 
I don't think anyone recommends testing your system stability with just IBT. IBT is really only good for testing your temps under a heavy (unrealistic) load with regards to overclocking. It does some stability testing, but Prime95 is more or less designed to do long term stability testing. IBT will just tell you a lot faster whether or not your current OC settings will work or not without having to do the 12-24 hour Prime95 test each time you make an adjustment, gives you a kind of starting point.
 
I've had situations where it was stable for hours on P95, and unstable within 30 minutes in LinX, and vice versa. Depends on which part of the overclock was unstable, etc.
 
Yes, this is exactly what I was thinking, however, if I'd only knew to tweak which setting...

I don't think my ram is holding me back in terms of stability btw, its corsair ram at its rated speed..
 
It could be your QPI voltage settings.

I also know that people have said that when you tweak for the lowest voltage in LinX, increasing the voltage can increase the GFLOPS, and thus performance.
 
Wouldn't call it completely unreliable. I have used it and it worked fine for me.

I have heard that LinX is better and I now use LinX with the proper updated code.
 
IBT runs the CPU hotter than anything else. But even Prime95 is more of a load on the processor than any current game.

I've had situations where a CPU isn't stable in IBT at certain settings, but stable in everything else, then goes for a year+ without a single BSOD during actual usage.

I use IBT more as a thermal test of the heatsink than a stability test at this point.
 
More heat means more voltage needed, thats a fact.
So I suppose those who fail IBT but pass Prime95 are just on the verge of "entering the steep part of the heat/voltage graph" with Prime95 and get into that with IBT if I'm making any sense to you..
 
Tsumi said:
I've had situations where it was stable for hours on P95, and unstable within 30 minutes in LinX, and vice versa. Depends on which part of the overclock was unstable, etc.
Click to expand...

Agreed.

I have run into both also, which is why I run many programs, IBT for 30-40 passes and Prime for 24 hours, I have had it pass one but fail the other in both cases. I then do renders, video encoding, gaming and the like, being the most stressful program does not always find something, sometimes the way a program interacts with the system and which parts it uses at once will show a problem that something else will not.
 
I had never used IBT and always relied on P95. My 4 ghz clock was stable in P95 for 10+ hours. I couldn't make it passed 3 IBT runs. Finally had to give up because my temps were way passed max safe (64C yikes!). So I bumped it down to 3.9 and dropped the voltage and just now made it thru 10 runs on maximum but temps are still 59-60.

About once a week, my rig would freeze and the screen would go gray. I was afraid it was an unstable OC which is why I tried IBT. P95 wasn't showing it so IBT worked better in my case. Its gonna drive me crazy not having the coveted 4.0 in my sig so I'm definitely gonna have to look into some better cooling. Guess its time to get under water. But yeah, love me some Intel Burn Test.
 
I used both for my 3770K. LinX using latest libraries (AVX included) for 15 runs (2+ hours) with max memory (14.5GB allocated) plus Prime95 small FFTs for 12 hours. Temps were actually pretty close between the two.

I agree that using LinX/IBT to check for quick stability/impact on temperatures is a good tool. It gets the CPU cores up to max temps almost immediately where I find Prime95 might take a while to reach that level, even on small FFT.
 
No single test is going to prove 100% stability which is basically impossible. One time the best oc test for one of my rigs turned out to be Media Player visual effects! Depends on where the critical path is.
 
I crash in Prime95 after like two hours, only ever tried IBT for about an hour. However for every day use and even up to 10 hours of gaming (longest I've gamed in a single period) it's rock stable and never crashes. I know a lot of people wouldn't consider it a stable overclock but it works for me.
 
fullvietFX said:
I crash in Prime95 after like two hours, only ever tried IBT for about an hour. However for every day use and even up to 10 hours of gaming (longest I've gamed in a single period) it's rock stable and never crashes. I know a lot of people wouldn't consider it a stable overclock but it works for me.
Click to expand...

This.

I don't really waste time on stability testers anymore, I maybe run Prime95 for an hour, and IBT for an hour and then stop. Doing it for hours on end is so pointless, because I've had a 2.5 hour stable IBT still crash on desktop randomly, same with Prime95.
 
You need the latest linpack binaries and you need to test with as much memory as possible. The only way to really achieve that is to boot into diagnostic mode and run linx there, when a minimal number of services and all your startup programs etc are disabled.

Start -> Run -> msconfig -> Click Diagnostic startup on the General tab.
Click OK and Restart.

You can get LinX with the 10.3.9.015 Linpack binaries from this thread here:
http://www.xtremesystems.org/forums/showthread.php?239175-Intel-4.5GHz-amp-5GHz-LinX-Stable-Club
It's the only place that keeps reasonably up to date LinX.

You can get the most updated binaries here(10.3.10.017 as of writing):
http://software.intel.com/en-us/articles/intel-math-kernel-library-linpack-download/
Download the Windows .zip, then extract linpack_xeon32.exe and linpack_xeon64.exe to your LinX folder. Overwrite when prompted, and now your LinX is up to date.

Mind you, in Intel's own latest propaganda on retailedge... They are suggesting LinX with the latest libraries to stress test CPU's when overclocking.
 
xIronCrossx said:
This.

I don't really waste time on stability testers anymore, I maybe run Prime95 for an hour, and IBT for an hour and then stop. Doing it for hours on end is so pointless, because I've had a 2.5 hour stable IBT still crash on desktop randomly, same with Prime95.
Click to expand...

Well I think instability can "build up" so to speak. My 4 gig overclock was perfectly fine running Prime95 for hours on end but every 7-10 days or so, Id get a crash out of nowhere even when I was just browsing a website. IBT showed me it wasnt stable after all. It also showed me my dual fan Hyper wasnt as good as I thought either. Temps were just fine under P95 but way too high in IBT.

Still not sure that was the actual fix as I only did that yesterday but Im pretty sure it is.

I think the best way and the way Ill be doing it from now on is 10 runs of IBT at max or the highest my RAM will allow and if it makes it thru that, then an overnight session of Prime95 Blend test to be sure.

Zoson said:
Mind you, in Intel's own latest propaganda on retailedge... They are suggesting LinX with the latest libraries to stress test CPU's when overclocking.
Click to expand...

I thought IBT was linpack. Whats the difference between IBT and LinX?
 
Last edited:
MacLeod said:
Well I think instability can "build up" so to speak. My 4 gig overclock was perfectly fine running Prime95 for hours on end but every 7-10 days or so, Id get a crash out of nowhere even when I was just browsing a website. IBT showed me it wasnt stable after all. It also showed me my dual fan Hyper wasnt as good as I thought either. Temps were just fine under P95 but way too high in IBT.

Still not sure that was the actual fix as I only did that yesterday but Im pretty sure it is.

I think the best way and the way Ill be doing it from now on is 10 runs of IBT at max or the highest my RAM will allow and if it makes it thru that, then an overnight session of Prime95 Blend test to be sure.
Click to expand...
Honestly, I JUST had a similar issue as you with the stable on any test and then crash after about a week. Eventually one of my sticks of ram died. Took about an hour to figure out which of the six, but when I figured out which one and swapped a different stick in there... No more crashing.

Bad part is the replacement stick was DOA. =\ Still waiting on the replacement.

MacLeod said:
I thought IBT was linpack. Whats the difference between IBT and LinX?
Click to expand...
They both use linpack... Just Linx is a more polished interface(IMO). I've never used IBT myself.

It's also really easy to update LinX. I can't speak for the ease of updating IBT.
 
MacLeod said:
It also showed me my dual fan Hyper wasnt as good as I thought either. Temps were just fine under P95 but way too high in IBT.
Click to expand...

I wouldn't bother worrying about your cooler not being good enough just because you have high temps in IBT or P95 - both of them are going to stress the cooler much, much more than any real world use will. I don't know that I've ever seen actual in-use temps within 20C of my IBT temps.
 
Zoson said:
They both use linpack... Just Linx is a more polished interface(IMO). I've never used IBT myself.

It's also really easy to update LinX. I can't speak for the ease of updating IBT.
Click to expand...

Wut. You claim one is better than the other and then admit to never using it? Wow, just wow.

They are the same program. Only difference I can think of is that IBT has more preset options and a blinking flame.
 
thesecond said:
Wut. You claim one is better than the other and then admit to never using it? Wow, just wow.

They are the same program. Only difference I can think of is that IBT has more preset options and a blinking flame.
Click to expand...

Seen screenshots... And no, it does not have more preset options.
 
Tsumi said:
I've had situations where it was stable for hours on P95, and unstable within 30 minutes in LinX, and vice versa. Depends on which part of the overclock was unstable, etc.
Click to expand...

Exactly, it just depends which is why it's good to use a variety of tests.
 
Agreed. I think its best to run at least 10 passes of IBT and if you make it thru that then youre ready for a 12+ hour session of P95. Pass all that and youre probably good to go.
 
That's essentially what I do. 10 passes of ibt set to "high" or higher stress and if that passes I run prime over night, check on it in the morning, if it's still going I let it run until I get back home from work.
 
I use LinX to test my spike voltage stability. Set the ram usage to like 1-2gb and it will be constantly going from high load to idle which I have found to show voltage spike instabilities.

Dunno a better word to call it but if your motherboard has sluggish power phases you know what I am talking about.
 
LinX is good for testing a quick OC settings, however, it's not really reliable, neither a prime is. The P95 takes forever to find out an error, and even if you pass the test for 24h, you might crash the next run within an hour. I usually run for a few hours and then stop it. I have my rig running for 24/7 and doing all kinds of shit from playing games to processing raw files. I've recently encountered a BSOD 124 while playing the D3. Raised the vcore, and seems good now, but might still crash in the following weeks.
 
GotNoRice said:
......

I use IBT more as a thermal test of the heatsink than a stability test at this point.
Click to expand...

Same here. IBT to see the max temp my cores will hit. Then I use prime95 for 12 hours and if its stable with that I never have any problems with the overclock.
 
You must log in or register to reply here.
Back
Top