did anyone come up with a reliable stability test yet?

rodrigox

rodrigox

rodrigox

Associate
Joined
25 Sep 2009
Posts
114
Yeah tittle says all...I've been using a i5 750 @ 3,6ghz and I can tell you its rock solid because its been running for 3 years without bsods or anything... I was running it with 1,21v with 180x20. Rest voltages default, now i wanted to mess a bit and put 4ghz but cant seem to get it right every 5 or 6 hours i have a bsod or a crash....bumped vcore to 1.4v and now seems to be stable but for real. Do you guys have any reliable program that shows if the cpu is stable?

Its pretty retarded at this point that there are no tools that can test a cpu 100%.
 
Depends what you mainly use it for. If its a gaming rig then use 3dmark, bit of everything then try pc mark or stress test with intel burn test.
 
As it's an old gen CPU you can still test using P95 or Intel Burn Test but I do like Real Bench myself. :)
 
I've always found OCCT in AVX linpack mode to be the most torturous of stress tests. Can past 100 runs of IBT and 12 hours of Prime 95 but an instability will be very quick to show it's head with OCCT.
 
Disco_P said:
I've always found OCCT in AVX linpack mode to be the most torturous of stress tests. Can past 100 runs of IBT and 12 hours of Prime 95 but an instability will be very quick to show it's head with OCCT.
Click to expand...

yup I run that for an hour with 64bits and avx compatibility and it is much better and punishing poor overclocks compared to intel burn test and realbench.

I had strange hitching problems in games and ran memtest64, GPU tests, SSD tests and everything and the problem was down to the CPU even though it passed Realbench.
 
You have to run IBT with under task manager performance Free amount of memory not "available ram" many people bung it on low med high very high etc I think is correct to enter custom amount and your free ram amount.

These days I like Asus real bench and IBT and the games I play.
 
RealBench, Intel XTU and AIDA64, pass all those you're pretty good.

Don't bother with p95 and OCCT, they have been shown to draw crazy current on newer CPU's.
 
harry5522 said:
And can cause damage to newer CPUs p95 is not recommended anymore at all.
Click to expand...

That's more down to poor hardware cooling/design and aggressive clocks/voltages at stock, it doesn't change the fact that Prime95 is a superb stability test. If you have hardware that can't cope with Prime95 stability test it just means that you'll have to settle for a lesser tester.

The whole concept behind stability/stress tests is that they use the processor as efficiently as possible, in other words try to use every bit of it concurrently. If the hardware overheats or throttles then that's not a failure of the software.

You're right to warn people that it may damage their procesor but it's not that the software is inherently bad or flawed, just that it's much better at the job than most other stability tests and some hardware these days is not physically designed to be used to its full potential, that might change again when we start seeing die shrinks.
 
3D mark, intel burn/LinX, then just use the dam thing.
The best test is running the system and using it if it works for your needs then its stable.

Dont use prime its poo, i had a 2500k(4.4Ghz) 3hr stable with prime, when i opened firefox it blue screened.... was not stable.
 
For lynnfield here's my settings the higher clocks like higher vtt especially the gigabyte p55 boards

205 x20 = 4.1ghz

vcore = 1.41 via adaptive voltage
qpi/vtt=1.41
pch core= 1.120
cpu pll =1.920
dram 1.66

heaven / valley benchmark
asus real bench
games you actually play
IBT (set it to 4 cores and custom amount of ram and always your free memory amount) 50 passes = rock solid imo

certain elements of prime and ibt can cause haswell cpus onwwards to overvolt on some boards i think but can be solved by a setting in bios i think cant remember what though
 
Last edited:
I did a little reading about prime95 yesterday. Apparently it's the volt regulators in haswell chip that are not up to the job and they let more bolts through when stressed with prime95.

So yes prime95 can damage your CPU but it's not software fault.
 
The best stability test is to just use the machine. If it doesn't crash, then it's stable.
 
koooowweeee said:
I did a little reading about prime95 yesterday. Apparently it's the volt regulators in haswell chip that are not up to the job and they let more bolts through when stressed with prime95.

So yes prime95 can damage your CPU but it's not software fault.
Click to expand...
It's been an issue since ivybridge, the non soldered ihs assembly means that mainstream chips run hotter than previous generations which were soldered. The problem got worse on haswell due to the addition of the vrms on the chips. AVX versions of prime 95 caused extremely high load temps on haswell at stock, same with the refresh devils canyon that was supposed to feature a new improved thermal paste. If anything DC actually ran hotter than the first generation haswell cpu's.
 
You must log in or register to reply here.
Back
Top Bottom