Best stress testing program for i7 system

bryanW1995

Lifer
May 22, 2007
11,144
32
91
Ok, I finally broke down and bought an i7 system. I've used prime 95 and occt in the past for stress testing the cpu. Are those still good or do you guys recommend something else?
 

Ben90

Platinum Member
Jun 14, 2009
2,866
3
0
I don't think there is anything that puts more pressure on the majority of a processor than Linx/IBT/Linpack.

Ive been prime stable for hours only to crash IBT within 5 seconds, and not just a rounding error, a full blown lock up.

That being said, IBT/Linx/Linpack are hands down the best for checking stability quickly, however the best solution for long term settings is to use a variety of tools. Each tool stresses the system in a different way.
 

ehume

Golden Member
Nov 6, 2009
1,511
73
91
OCCT using Linpack gives you nice graphs at the end of the run. Also, you can customize which temps and voltages you want to track.
 

DrMrLordX

Lifer
Apr 27, 2000
23,240
13,327
136
OCCT's implementation of Linpack uses old libraries. It doesn't stress the CPU as hard as LinX or IBT, or even Prime95 Blend.

I still like Prime95, especially Prime95 Blend. For some reason, Blend really kicks my K10.5 chips in the ass compared to small or large FFTs (not sure why), and it seems to be a good way to pick up memory and/or NB stability problems. LinX is quite a bit more hardcore though.

Also, if you're serious about uncore overclocking and/or RAM overclocking, MemTest HCI is a good one to use. I've been using it lately in conjunction with LinX to diagnose possible NB stability problems. LinX was halting at about 2 hours 30 minutes without giving me a BSOD, so I ran MemTest HCI and, sure enough, I had memory stability problems. The culprit? Me trying to run my NB at 3 ghz again.
 

bryanW1995

Lifer
May 22, 2007
11,144
32
91
I've heard of linpack but not linx. I used occt in the past but it looks like it hasn't been updated in 15 mos or so. I'm using memtest86 to test my memory.

I started looking around b/c I couldn't remember how to get to 100% cpu usage on prime 95. I know that I eventually went to occt on my x3350, but I distincly remember using prime on all 4 cores at once. I'm stuck at 13% right now, so only hitting one core.
 

DrMrLordX

Lifer
Apr 27, 2000
23,240
13,327
136
The latest version of Prime95 should be able to spawn one thread per core. HT might interfere with proper testing of the CPU, sadly. Or, at least, I think that's the case with LinX/IBT. You should get higher gflops with HT disabled, which implies that it works the CPU harder when it isn't spawning logical cores.
 

DrMrLordX

Lifer
Apr 27, 2000
23,240
13,327
136
Normally you can't, but the thing with HT is that the logical cores it spawns utilize resources from physical cores that are being under-utilized. Programs using the Linpack libraries (LinX, IBT), among others, are capable of stressing the cores of an i7 so severely that there are no resources to spare, so using HT while running LinX/IBT actually reduces the performance of an HT-capable i7 (and, hence, reduces the heat output and the stress on the CPU).

I don't have any personal experience stress-testing HT-capable i7s, but logic would dictate that if you can get such a chip stable in LinX or IBT, it should run HT just fine at those speeds. Should. Something like LinX would probably turn up errors/crash the system/etc. at a clockspeed where the chip would be relatively stable in normal apps without HT but unstable with HT.
 

ShawnD1

Lifer
May 24, 2003
15,987
2
81
Prime is my favorite for quick testing things, but OCCT is arguably the most useful. OCCT shows graphs, it shows temperatures, it shows voltages, and it has the PSU test.

OCCT's PSU test is useful for heating rooms in your house. I don't trust a $20 space heat made in China that has no protection against catching fire, but I trust my old computer to not start a fire :)
 
Jul 4, 2010
44
0
0
i would use linx and p95. you can be stable after 20 passes of linx, but still error out in p95, and vice versa. but really, the best stablility test is daily usage. even if you're stable with both of these, your computer can still crash. there's no stress test that will test every aspect of the CPU, although it can be very close. and also note that if you're on the edge of stability, a chance in ambient temps can cause your rig to crash, like the huge rise of ambient in the summer vs winter.

on my old OC'ed desktop, i was stable with linx and p95, but would sometimes randomly crash when streaming HD content. a bump in the vcore fixed the problem though.
 

Elganja

Platinum Member
May 21, 2007
2,143
24
81
the key for linx is to run it with max memory... yeah to takes a lot longer but it's a much better test
 

taltamir

Lifer
Mar 21, 2004
13,576
6
76
i would use linx and p95. you can be stable after 20 passes of linx, but still error out in p95, and vice versa. but really, the best stablility test is daily usage. even if you're stable with both of these, your computer can still crash. there's no stress test that will test every aspect of the CPU, although it can be very close. and also note that if you're on the edge of stability, a chance in ambient temps can cause your rig to crash, like the huge rise of ambient in the summer vs winter.

my approach was to find the highest OC that is stable in all testing programs... then lower the OC somewhat and used the lowered value rather then the highest tested value. That way if I run across a program which is more stressful, I have some extra headroom.
 

bryanW1995

Lifer
May 22, 2007
11,144
32
91
yeah, I've always tried to get my desired 24/7 stable OC (based upon how hard it is to get to 3.6, 4.2, or whatever), fine tune voltages down as low as possible, then drop it back 100mhz or so for 24/7 usage just to be safe.