True or false: A working CPU should not crash in a stress test

Jus136

Distinguished
Nov 12, 2008
47
0
18,530
There are plenty other factors that go into a stress test: what memory you have, power supply, voltages, heatsink fans, etc. Give us the full picture of your problem and we can help you better.
 

cadder

Distinguished
Nov 17, 2008
1,711
1
19,865
Technically it depends on what your stress test is vs. what your average use of the CPU is, and how much safety factor you want. If your cpu can survive intense stress testing for hours and hours, you can be sure that if something crashes while you are running the machine it wasn't caused by your overclocking.

We talk about running a stress test for hours, but in my case running CAD, my CPU is typically only at max. for less than a minute at a time. In between those intense minutes it is sitting and waiting on mouse/keyboard input from me. But when I need it to work hard, I really need it to work hard. If it will work hard for 1 minute, but maybe crash if it had to do it for 5 minutes, I might never know the difference.
 

roofus

Distinguished
Jul 4, 2008
1,392
0
19,290
crashing could mean alot of things but not necessarily a bad cpu. incorrect voltages, bad thermals, memory timings, etc. not cut and dry good or bad.
 
True. It should not crash in a stress test.

If it does, it is not actually stable, and even when it seems to be running, it might drop a bit and return the wrong result or something similar. A working, stable CPU should be able to run even something incredibly stressful like Linpack or Prime95 almost indefinitely without trouble.
 

Jus136

Distinguished
Nov 12, 2008
47
0
18,530


That's assuming that everything else in the system is working flawlessly. If that's the case, then I'd say true, but the OP didn't provide enough information to let us know that.
 

roofus

Distinguished
Jul 4, 2008
1,392
0
19,290
yea i wouldn't automatically call it a bad cpu if it fails a stress test. it could be a bad setting, under powered, too hot, etc. not enough information.
 

SteveSmith1980

Distinguished
Jan 20, 2009
64
0
18,630
I tried two types of tests. First I opened up about 15 Youtube videos at once. Second, I used a program called CPUStabTest. Both had my CPU work at 100%. and both made my computer crash, and my power supply even fried once! Also, when it crashes, my computer doesn't power for about 20 minutes. It's very scary.

It's an ES chip, and I returned it... I don't think that's normal behavior for a good-working processor.
 

pr2thej

Distinguished
Sep 25, 2008
1,352
0
19,290
Well, opening 15 youtube vids isnt normal behaviour either :D im not familiar with the other stress test.
I wonder how Prime 95 works though, just it just take the CPU to 100% bang on and keep that up for the duration or does it try to exceed 100% ?

For example (im sure the number of calcs is wrong but just an example):

2Gz single core processing 1000 calcs / sec = 100% CPU load
4Gz single core processing 2000 calcs / sec = 100% CPU load

4Gz single core processing 2100 calcs / sec = 105% CPU load

So...does Prime95 try to push the CPU over 100% like xxxx number of youtube vids would?


Anyway yea dodgly PSU syndrome + ES = crashes. Whats your rig gona look like with a new chip, and whats your PSU actually outputting on each rail? CPU-z can tell you that.
 
A more accurate question should be, "T or F, A working system ..."

I stress test with Prime95. My personal goal is a 24 hour test run. When I was working up the settings for my Q6600, I thought I was done until it crashed 14 hours and 18 hours into what I had hoped were the final tests. Finally had it running at 3.6 GHz.
 

Malovane

Distinguished
Jun 17, 2008
177
0
18,680
As said, stress testing pushes a lot of components hard... not just the CPU. CPU's are usually tested fairly well at the factory... chopped and binned for certain characteristics, then tested again once fully assembled and ready to ship. It's rare to purchase a malfunctioning CPU, though it isn't hard to damage the chip upon installation, as they can get a zap from electrostatic discharge off a person, or get a voltage shock from entering a not-totally-discharged motherboard (some people even go so far as to remove the CMOS batteries before installing the CPU). Even so, CPU's are often fairly resistant to such abuses. More often than not, other components are to blame for the failure of the system during a stress test.

Memory settings (voltage and timings) can be set up improperly, so that while it can do simple things like load windows, they will fail on heavy tasks. Even if the voltage and timings are set up to manufacturer specifications, the sticks you buy may not meet those specifications themselves. Hence the need for stress testing, so you can retard the timings or increase the voltage if needed (or just return the junk sticks). It also works the other way as well... where you can tighten the timings and increase the speed as long as it keeps doing fine under stress testing.

Sometimes motherboards themselves have issues that would cause instability with the stress test. It could be an issue with the northbridge, the voltage regulators, and even improperly configured BIOS's can be cause for a crash in a stress test. A poor quality (and/or highly taxed) power supply could also be problematic.. as they can have pretty bad fluctuations in voltage which can lead to instability.

Your power supply failed on a stress test. Let's start there. What power supply were you using that blew, and what supply are you using now? What other components do you have in your system?
 


A working *system* should not crash in a stress test. A CPU stress test that runs the CPU at full load for an extended period of time only fails if the CPU screws up, but the CPU itself may not actually be the problem as something else may be causing it to screw up calculations:

- Your CPU may be inadequately cooled due to an inadequately-sized or dirty heatsink, poorly-ventilated case, inappropriately slow heatsink fan speed, or the CPU-thermal compound-heatsink interface may be messed up (too much/little thermal compound, poor fit, rough heatsink base.)

- Your power supply or motherboard may have too much voltage droop under full CPU load, causing the CPU voltage to dip too low and the CPU to screw up calculations.

- Your RAM may be faulty or set with inappropriate voltage or timings, causing data corruption and the CPU to do math using faulty data.

- Your chipset may be inadequately cooled or have its voltage set incorrectly so it introduces errors when transmitting data from RAM to the CPU if your system has a frontside bus. It could also be faulty as well.

So what you should do is try to track down the problem.
- Memory issues: run Memtest86+ for several passes and make sure there are no errors.
- Power supply issues: use a hardware monitoring tool like SpeedFan to monitor PSU voltages and CPU core voltage and make sure they are in appropriate ranges.
- Temperature issues: use the same hardware-monitoring software to ensure that all temperatures are within normal limits.

If all that looks okay and you still get problems, then it is probably time to start swapping out parts to see what is bad.