Intel Burn Test: 20 passes, 2 of them half the gflops

McGraw

McGraw

McGraw

Soldato
Joined
16 Jan 2006
Posts
3,046
Hi.

Last night I had another crack at stress testing my PC to make sure my tame (limited by quiet fan on my H60) overclock was stable.

Currently I'm at 4.5 on my 4790k using 1.212vcore rising to 1.224 in AIDA and 1.236 in ibt.

I had a bsod ages ago when I was rapidly alt tabbing in and out of FSX trying to figure DSR out and I've not had one since but I know 1.2v isn't AIDA stable so I've not got much breathing room.

Last night I did a 8gb encode in handbrake whilst watching sky go, vlc player and a flight in the ngx on fsx. Constant chopping and changing focus didn't even crash fsx.

So after an hour of this I figured I was pretty safe so started to close everything down except handbrake which hadn't finished. I moved by pc and accidentally disconnected my card reader and then I was rapidly clearing min/max in HWMON to see if my temps were dropping due to better airflow.

At this point I got a bsod but I figured it happened when doing something abnormal so wasn't too concerned.

I then did another hour of encoding and 20 passes of IBT with no problems...except two of the gflops results were half the others (60 vs 120).

I don't think it was throttling although my temps did touch 91.

Are some of the passes different in some way or do I have a problem?

I can't go any higher on vcore.

Thanks.
 
Gflops varies depending on CPU load, if you're running another program along side IBT then that's going to affect the amount of CPU resources IBT uses and thus reduces the Gflops rating.

I love IBT personally and always use it on my system.

20 runs at all RAM tested seals the deal for me.
 
OhEsEcks said:
Don't use IBT! See the top of page 65 (and the following discussion) in this thread: http://forums.overclockers.co.uk/showthread.php?p=27360443

You'll probably get higher stable clocks at better temps using something other than IBT or P95. IBT sent my chip to 80c+ and I quickly "noped" that and stopped the test.
Click to expand...

I used P95 and I was stable for 3 hours, go and use IBT for 2 minutes and crash so I don't touch P95 anymore.
 
Hmm I'm pretty sure it wasn't doing anything else

I'm going to consider it ok and knock it down to 4.4 if I get any problems in games or normal use.

Thanks
 
Doogles said:
I used P95 and I was stable for 3 hours, go and use IBT for 2 minutes and crash so I don't touch P95 anymore.
Click to expand...

IBT is the extreme of the extreme. You're unlikely to ever tax your chip the way IBT does in any real world scenario. Even compiling Gnome from source on FreeBSD my chip wouldn't be taxed like IBT does it.

I've just hit 4.8GHz at 1.275v with a max temp of 78c, using Asus Realbench. IBT tortured my chip at stock. If I'd tried clocking it this would likely have been the result.
 
you only need to do 5 runs ibt,it will crap out within that time if its unstable (depends on the stress level,max stress for me but those chips run hotter)

I find prime95 blend test good for finding unstable ram/not enough memory controller voltage
 
The only torture test that I'm happy with (so far) is IBT.

If the system passes IBT then it never, ever, EVER crashes while playing ARMA.

When I overclock and torture test with the others I always end up with a weak 'stable everywhere else but arma' overclock.
 
Threepwood said:
The only torture test that I'm happy with (so far) is IBT.

If the system passes IBT then it never, ever, EVER crashes while playing ARMA.

When I overclock and torture test with the others I always end up with a weak 'stable everywhere else but arma' overclock.
Click to expand...

I play arma but only single player and I doubt I'll do much more than coop missions online.

I know multiplayer is very taxing. Do you test by playing online or use a bench?

Thanks.
 
Id only run IBT/P95 on a haswell chip thats been delidded, on a standard ihs chip you run a good risk of it throttling. Theese cpu's just arent designed/built for such high loads.
 
McGraw said:
I play arma but only single player and I doubt I'll do much more than coop missions online.

I know multiplayer is very taxing. Do you test by playing online or use a bench?

Thanks.
Click to expand...

Testing stability by playing the game takes too much time and it is too unpredictable.


All I know is that if the system passes IBT (yes if it is going to crap out 99% of the time it is within 5 runs, but I've seen it crap out after the 5th run also) then it WILL be arma stable.
 
Last edited:
Ok I've been doing some more testing whilst getting ready for Xmas and I've managed to iron out an nvidia driver crash in asus stress test by disabling my igfx service but I seem to be left with a seemingly repeatable bsod.

If I'm using handbrake either on it's own or alongside other tasks it's fine.

If I stop handbrake, close other stuff down then start handbrake again (without closing it down) it blue screens pretty quickly.

Any idea what can be causing this? Everything else runs great including stress tests and I don't even use handbrake normally.

Thanks.

Edit. Could it be that asus bench uses handbrake so I go from running two instances to one?
 
Last edited:
I love IBT, 100 runs at 7GB ram is my final hurdle for an overclock to be stable. I've had runs crap out at 40+ passes.

@OP what is the error given by the bluescreen? Try going back to stock and seeing if the bluescreen still happens. I had a bsod issue that happened at stock due to a dodgy Intel driver install.
 
Cheers for your reply.

I'm hoping that the igfx thing was also part of the bsod but if it wasn't then I'll probably just up the volts rather than spend hours tracking down the source of the problem.

I'm starting to measure my fun time against the price of a new cpu and if this one dies around broadwell time then I'll see what intel says.

In other words I don't think 1.248 and 80c will be a problem.
 
You must log in or register to reply here.
Back
Top Bottom