stress tests = useless?

E

erorr404

Gawd
Joined
Jan 30, 2005
Messages
646
i oc'ed my P4's FSB from 133 MHz to 147 MHz.
i ran prime95 blend test for 10 hours and S&M on max load and max time twice, and didnt get a single error or warning from them. i wake up this morning and take a screenshot of them to keep a record of my stable 147 MHz FSB overclock, and when i paste the it into Paint the system freezes. obviously it was very unstable at that setting, but the programs i ran said nothing.
can someone please explain this to me? these programs are supposed to test CPU stability, are they not?
 
We use Prime 95 to stress test, but we never run it and use it alone as a stability tester.
 
Prime doesn't test every aspect of a system. It does fine for the CPU and some aspects of the memory, but it won't, for instance, tell you a lot about the video. I've had systems that will run Prime for 24 hours, but will lock up under 3D Mark.
 
then not enough vcore?

also Prime95 is just one more tool we use to test stability..stress testing isn't just running one program..it's running several for different aspects of the system
 
If you up the FSB without putting in a divider or something, it's going to up the memory too.

Personally, I use a few hours of fragging for stress test...not the most professional way to do it, but it works.

It's also quite possible that it WAS stable, and crashed for unrelated reasons. Possible.
 
gwai lo said:
then not enough vcore?

also Prime95 is just one more tool we use to test stability..stress testing isn't just running one program..it's running several for different aspects of the system
Click to expand...
not enough vcore is probably the cause of the instability, but the problem here is that i dont know when its stable or not, at least not judging from prime95 and s&m. its strange that they ran for hours without any problems but when i paste a screenshot into paint it crashes. next time i will try some more applications like 3mark01, which actually crashes at as low as 139 MHz overclock, while prime and S&M dont crash at even 147 MHz. :rolleyes:


ashmedai said:
If you up the FSB without putting in a divider or something, it's going to up the memory too.

Personally, I use a few hours of fragging for stress test...not the most professional way to do it, but it works.

It's also quite possible that it WAS stable, and crashed for unrelated reasons. Possible.
Click to expand...
i use a 1:1 divider so that my memory is underclocked even when i overclock my FSB. the only possible cause of instability here is the CPU. i highly doubt that it was stable and crashed due to unrelated causes, what are the chances of that?
 
erorr404 said:
i oc'ed my P4's FSB from 133 MHz to 147 MHz.
Click to expand...
erorr404 said:
i wasnt overclocking RAM or video card. the only component being overclocked in my computer was the CPU.
Click to expand...

Somehow I took it from your first post that the CPU wasn't the only thing in the system that was overclocked....
 
CIWS said:
Somehow I took it from your first post that the CPU wasn't the only thing in the system that was overclocked....
Click to expand...
heh, i should have been more clear
 
erorr404 said:
i use a 1:1 divider so that my memory is underclocked even when i overclock my FSB. the only possible cause of instability here is the CPU.
Click to expand...
1:1 divider means that FSB goes up 1, memory clock goes up 1 (2 if you're counting by DDR pseudoclock). So you are in fact overclocking both RAM & CPU. Although if you have PC4400 or something it's not technically overclocking, but it still has to have appropriate voltage and timings.

erorr404 said:
i highly doubt that it was stable and crashed due to unrelated causes, what are the chances of that?
Click to expand...
On a computer running Windows? High enough. ^_^

I'm a suspicious bastard though so I blame it on the OC until proven otherwise. Plus I have this nagging belief in linear cause and effect that they hammered into me at a young age.
 
ashmedai said:
1:1 divider means that FSB goes up 1, memory clock goes up 1 (2 if you're counting by DDR pseudoclock). So you are in fact overclocking both RAM & CPU. Although if you have PC4400 or something it's not technically overclocking, but it still has to have appropriate voltage and timings.
Click to expand...
i realize that RAM speed goes up as FSB speed goes up but, what i meant was that my RAM was underclocked from 166 MHz to my FSB speed which is 133 MHz. as long as i dont push my FSB past 166 MHz (unlikely), my RAM will stay below stock speeds.

ashmedai said:
On a computer running Windows? High enough. ^_^

I'm a suspicious bastard though so I blame it on the OC until proven otherwise. Plus I have this nagging belief in linear cause and effect that they hammered into me at a young age.
Click to expand...
i really think it was because the high FSB made the system unstable, since 3dmark01 crashes for me at any speed over 141 MHz. ill test stability in 3d apps as well from now on, other than just running S&M and prime.
 
OOOOOH I get it now.

How's your cooling? If it ran test for 10 hours and then crashed, and the memory is clocked well below its rating, it's possible your CPU lost stability as it gained heat.

Sorry if me ragging on MS confused you any. ^_^
 
cooling is fine, i forgot what exactly it was when i stopped the test and took the screenshot, but it was in the 40's. i never saw it go over 45 after i cleaned the dust out of the case a few days ago and put in a new 120mm fan. could it be that 3d apps like 3mark01 and games stress the CPU in areas that those programs i ran do not?
 
Nothing like killing a few dust bunnies to drop the temps.

Floating point type tests are god for checking out the CPU, but 3D stuff will load the system as a whole differently (calculating primes is rather unlikely to load the GPU much for example unless your 3D card just sucks that badly). It's a good idea to do both, plus memtest to make sure that's not going to flip out. Just because you're running the memory under rated speed doesn't make it innocent of suspicion, your crash could just as easily be because the memory is defective.
 
i actually ran memtest86 1.55 and S&M 1.5.1 yesterday to see how stable CAS 2 would be. i got errors in both programs so i went back to the stock CAS 2.5 and that was stable. so i think we can eliminate RAM.
the video card has heating problems because the fan recently died, but.. when i ran 3dmark01 with FSB OC'ed to 139 MHz it crashed on the nature test, and when i ran 3dmark01 nine times at stock FSB, it never crashes. so i think video card can be eliminated as well.
 
stress testing isnt useless when you do it right. Run Prim, memtest, all 3dmarks(well i dont think 2000 runs anymore)
Games, Render, whatever. The point of a stress test is to test everything to the limit to see if it brakes.

Run Prime95 and 3dmark05 at the same time, enjoy the scores :)
 
will do, i found that 3mark01 is the only program that shows CPU instability for me, while S&M and memtest86 work at testing memory instability.
 
I would put it the other way.
Because you can run Prime95 at a higher FSB than 3dMark without failing then I would say its part of vidio subsystem that fails and not your CPU.
I would try to find other vidio stress tests and see if thats the case.

Luck.......:D
 
Maybe, but I ran 3mark01 nine times in a row on stock FSB and didnt get a single crash, while I got crashes on the first loop of tests when the FSB was overclocked to 139 MHz and 141 MHz. It could have been a coincidence though because my Ti4200's fan is dead and that might be the reason for the crashes.
BTW, thanks to everyone for taking time to reply here. :)
 
The cooling fan on your 3d card is dead, and a 3d benchmark crashes your computer? Okay.
 
.............
the 3d benchmark crashes when the FSB is overclocked.
the 3d benchmark does not crash when the FSB is not overclocked.

and other than the 3d benchmark crashing, there are more reasons to believe that the FSB speeds tested in Prime95 and S&M are indeed unstable. one time, after a 10 hour prime test was complete, i took a screenshot and the computer crashed when i pasted it into paint. another time, i did a 5 hour prime test on a bit lower FSB, and when that finished the computer was at a crawl.. the mouse wasnt very responsive, applications took forever to open, etc.
 
Overclocking leads to heat
Heat leads to device failure
Bad Star Wars jokes lead to getting kicked in the nuts...how's that for a mixed reference.

Your OC isn't stable, could be lotsa things, but I'm betting that the dead fan dun help anything. At that, if you don't have it locked you might be inadvertantly overclocking an uncooled vid card while under load.

I'd say it's safe to say that for whatever combination of reasons, your computer isn't up for an OC right now...
 
ashmedai said:
Overclocking leads to heat
Heat leads to device failure
Bad Star Wars jokes lead to getting kicked in the nuts...how's that for a mixed reference.

Your OC isn't stable, could be lotsa things, but I'm betting that the dead fan dun help anything. At that, if you don't have it locked you might be inadvertantly overclocking an uncooled vid card while under load.

I'd say it's safe to say that for whatever combination of reasons, your computer isn't up for an OC right now...
Click to expand...
<yoda>
Hrm -hm right are you, properly cool he did not
</yoda>
Hell it could simple be your hardware doesn't like OCing, start over and remove everything and replace any broken fans ( :rolleyes: ) and start with a clean install of Win.

If your video card is fine (fine being relative and all) at stock speeds and then you increase the FSB pumping out more heat then guess what? If your using a 3d benchmark then it's going to fail cause all that extra heat is just going to overwhelm your broken video card.

Quite frankly it sounds like you shouldn't really be OCing that system anyways and it's clearly not in any shape to be OCed.
 
erorr404 said:
i oc'ed my P4's FSB from 133 MHz to 147 MHz.
i ran prime95 blend test for 10 hours and S&M on max load and max time twice, and didnt get a single error or warning from them. i wake up this morning and take a screenshot of them to keep a record of my stable 147 MHz FSB overclock, and when i paste the it into Paint the system freezes. obviously it was very unstable at that setting, but the programs i ran said nothing.
can someone please explain this to me? these programs are supposed to test CPU stability, are they not?
Click to expand...

Well for starters you need to run Prime95 using In-place large FFT's instead. Blend doesn't put the system under that heavy of a load and tests primarily alot of RAM. You can also bump the priority level of Prime95 from 1 to 10 to put the system under even more stress.

Run Prime95 using In-place large FFT's to test CPU and overall system stability and then Memtest86 to test your RAM.
 
I always run Memtest86 first.
Then I run run Prime95. Useing same number of copies as to processors.
Then if thats stable add 3dMark on top of Prime.
If it survives that then not a lot of things you can throw at the system will cause it to crash.

Luck.......:D
 
thanks for the good advice Tigerbiten, ill throw a bunch of stuff at it when i overclock to make sure its stable.

burningrave, i actually have tried In-place large FFT's before, when FSB was overclocked to around 140 MHz (forgot exactly how much), and i still didnt get any errors, while 3dmark crashes at 139 MHz and above.

my RAM should be fine.. i tested it in memtest and S&M already and it gave no errors when running at the current timings. besides, the RAM is underclocked by 30 MHz just to make sure it doesnt give me any errors when im messing with CPU.

crimandevil, i think the cooling on all of my components is fine (except possibly GPU cause of the fan). the CPU is 30 C on idle and the highest ive ever seen it on load (since cleaning the dust out of my case a couple of weeks ago) is 48 C. case temp is 25 C.

i think there are only two possible things that can be happening:
1. S&M and Prime95 didnt detect my unstable CPU.
2. the CPU was stable at those speeds, but something else was not.
that something else couldnt be the RAM since its underclocked and totally stable (memtest86 and S&M tested). everything points to the video card with the broken fan.


i think asmedai is most likely right. i dont think the AGP speed is locked, and i dont think it is lockable on this motherboard.
is the "AGP/PCI Frequency" supposed to be the one i should lock? the available settings are 66/33, 68/34, 70/35, 74/37, 76/38, 78/39, and 80/40. there is no lock setting. it currently (and by default) set to 66/33. is this the safest setting?
i will go download a video card overclocking utility right now to see if the video card clocks go up when i raise the FSB clock.


EDIT: actually i just did some research on my motherboard and i realized that he AGP/PCI freq is locked when i change FSB. i just went into BIOS and when i change FSB speed, the AGP/PCI frequency remains at 66/33. so i guess it is locked.
so that eliminates the possibility of the video card ruining causing 3dmark01 to crash.. since at stock FSB it doesnt crash in 3dmark but with high FSB is does crash and the video card remains unchanged when the FSB changes. therefore, it must be that the programs just dont catch the instability for some reason :confused:.
im out of ideas.
 
You must log in or register to reply here.
Back
Top