What's the point of synthetic 'heat virus' stress tests?

I know everyone recommends stress testing one's OC configuration with some synthetic application like prime95, AIDA64 or whatever, including the savvy guys. But I'd like to actually learn on the actual point of it.

Fact:
The stress/temperature imposed during synthetic testing is unachievable in many real scenarios (especially gaming!*).
*ArmA 3 example: Cores 1-3 50-80% utilization, Cores 4-5 ~30% utilization, Core 6 almost not utilized, plus most probably only partial utilization of CPU feature sets.

Problem:
So why would I bother hammering my CPU under conditions it will never experience?


The only thing that I could think of is:
'If the particular OC config gives errors in Prime95 it means that something bad is going on in the CPU and this bad thing will continue to be going on at that OC config even if CPU load is at 60-80%.'

Disclaimer:
I OC the rig exclusively to play ArmA. Once I finish my session I load up a power-saving oriented BIOS profile.

Hello Bucic, there are a few reasons you want to run very stressful tests on your system.

Stability is a fickle thing to achieve, and pushing your system hard makes obtaining it substantially harder. There is no test, or sets of tests, you can run to know you system is absolutely stable. By running stress tests, you are showing the probability of instability is lower, the more stressful the test is and the longer you run it. How stable you need the system depends on what you are doing. For example, I do a lot of programming, so system instability is A Very Bad Thing for me. Whereas if you only do gaming, you may be able to tolerate (or not even notice) instability, but you might notice the increase in processor speed.

Additionally, by pushing your system extremely hard, you are testing the worst case scenario it will ever see. Not a single one of my applications will stress as hard as something like linpack or prime95. So, if my system is stable under those conditions, it should be able to handle "simpler" tasks, such as running a virtual machine, having two email clients open, countless browsers and tabs, Visual Studio, foobar, Excel (etc) simultaneously.

I also don't see people mentioning when you should run stress tests. I frequently see mention of running tests overnight on the forums, which is a great way to get a lot of time to run a test. However, the environmental factors are different at night, which will affect the stress test. Temperatures drop at night, which aids stability. If the computer sits near a source of heat (window, radiator, etc), it may crash during the day because the ambient temperature rises.

Each person has their own situation and ideas, so answers to your original question vary; the only correct answer is "it depends".

My take is very similar, if you're a gamer and the odd crash isn't going to affect you too much then maybe you don't need it to be "100% stable" . I also see why this is important to you since your mobo is a bit underpowered for that CPU you're running.
Your not likely running any shared computing like F@H which will error out even after a P95 stability pass. Or recoding video where errorr in calculation will ruin the final product and hours of work need to be re-done.
Since I use my PC for more intense applications , my testing consists of 2 hrs of P95 blend then if it passes I start heaven benchmark in a loop for another two hours. I need this stability since some applications will load the CPU and GFX cards to nearly 100% for hours on end. This test not only shows stability but also heat dissipation from the case. With an 8core FX and 2 GTX 580 there's a lot of heat produced when under load and if it builds up it "will" affect your stability. I spend a lot of time with stability and airflow so I'm fairly confident that in any circumstance my PC will not crash but for me I would lose more than just my last save point.
I would also like to add that Windows is susceptible to crashes and eventually you'll need to reinstall it if you continue to crash.

Yes, I would go as far as saying no application I know of is going to put as much load upon CPU as prime95.

"Additionally, by pushing your system extremely hard, you are testing the worst case scenario it will ever see."
That's the thing, it's not true. It will not see such a scenario

"you don't need it to be "100% stable" "
Is this statement even correct? I mean, I have never experienced a single crash in ArmA 3 with the OC config I use. It's 100% to me I know, I know, you meant it differently.

Anyway, I just wanted to make whether the statement in the OP in italics is not true. I wouldn't want to wreck my CPU within a year.

Stability is subjective to the individual and the tasks one wants to perform.
What is stable for your day to day use and gaming may not be stable enough for those who might want to launch a missile or fold.

These guys are giving you a baseline test (2 hrs. of P95). It's up to you whether you want to skew from that baseline.

Bucic said:

"Additionally, by pushing your system extremely hard, you are testing the worst case scenario it will ever see."
That's the thing, it's not true. It will not see such a scenario

Click to expand...

What you interpreted from my post wasn't my point. If your stability test pushes your system the hardest and it "seems stable", you shouldn't have issues under normal usage. The point is to find where instability occurs. If it errors 24 hours into a Prime95/linpack test, you may not care because you don't push the system anywhere near that hard.

Bucic said:

"you don't need it to be "100% stable" "
Is this statement even correct? I mean, I have never experienced a single crash in ArmA 3 with the OC config I use. It's 100% to me I know, I know, you meant it differently.

Click to expand...

The problem with this thinking is you can't guarantee an overclock is "100% stable" without running the chip indefinitely, which isn't something most people want to do. You can only test to see how unstable it is.

Bucic said:

Anyway, I just wanted to make whether the statement in the OP in italics is not true. I wouldn't want to wreck my CPU within a year.

Click to expand...

The only thing that I could think of is:
'If the particular OC config gives errors in Prime95 it means that something bad is going on in the CPU and this bad thing will continue to be going on at that OC config even if CPU load is at 60-80%.'

Click to expand...

If a particular overclock configuration produces errors in Prime95, the system is unstable under those conditions. The system may seem perfectly stable outside of the tests and may never crash for you. It all comes down to what you need, which is why you are getting different answers. As I mentioned above, I need my system to be as stable as possible, since I'm programming. You may not need that level of stability if you are gaming, because the game may not have a problem with a few computation errors or you may not care if there are rare issues.

There is no right answer except "it depends".

There is no right answer except "it depends".

Click to expand...

QFT!

I feel part of the issue is when someone comes to the forum wanting to OC they will usually hear that the majority of more experienced OCers consider a system stable when it passes 2 hours of Prime Blend. There are also specific monitoring programs that we recommend. One reason why these programs are recommended is because it is very difficult for someone to help guide another person with their overclocking if we are not all on the same page. It is very helpful, especially when really trying to push ones system if we know exactly where we are at. So not only do we recommend 2 hours of Prime Blend to see if it's "stable", it's also so it makes it easier for us to help someone really push further. Are these programs the best to use not necessarily but again it is helpful when we all know what we are seeing.

On the points above, I use my rig Oced for my work and the programs I use load the cpu up to 100% for quick bursts all day long for about 8 hours of the day. If my rig were to crash it could be very costly for me. I know that if I can pass 2 hours prime blend I can use my rig for everything and anything I throw at it and it will not crash. Is it an infallible no but it works for what I use it for, so therefore it all depends.

thideras said:

There is no right answer except "it depends".

Click to expand...

To me you've given a definite answer yourself.

thideras said:

If a particular overclock configuration produces errors in Prime95, the system is unstable under those conditions.

Click to expand...

Thank you guys. I just wanted to confirm there isn't any type of creeping damage while playing a game on OC that is unstable in p95.

Also, the re core utilization in ArmA 3. Here's a correction. It's actually 0% for cores 5 and 6. To show discrepancy between A3 and p95 in terms of temperatures:
A3: ~59 deg C max in ArmA3Mark Altis benchmark
p95 small FFT's: 75 deg C within 10 seconds

I've terminated the p95 test immediately and I'm not going back At least not for ArmA 3.

I just wanted to confirm there isn't any type of creeping damage while playing a game on OC that is unstable in p95.

Click to expand...

The only creeping damage will be to you OS. Sometime things just won't seem to work like they're supposed to. The first question when people experience issues is " Do you have the system OC'd" There's a reason for that. It can slowly break down your OS.

Johan45 said:

The only creeping damage will be to you OS. Sometime things just won't seem to work like they're supposed to. The first question when people experience issues is " Do you have the system OC'd" There's a reason for that. It can slowly break down your OS.

Click to expand...

Yup, I know of this particular possibility.

Anyway, thank you guys for your input!

I'd like to weigh in on the benefits that I can see.

Understanding your system better.

to torture testing a system safely, one to get to know what their components are capible of. One needs to get a baseline. One has to know what temps and voltages are typical for their system.

By knowing these things, and knowing where to find them, you can become a better system builder. As you run into instability issues to continue on the quest for a more stable system, you are forced to adapt and continue to learn.

Stress testing rapidly reveals instability that might not rear its ugly head for a long time or at particular temps. Wouldn't you rather know your system was stable way before the fancy overclock you placed on it failed?

Experience becomes expertise

Since starting my OCing journey I was able to "fix" my little cousin's computer because he had messed around in the BIOS. He was obviously running his RAM too tightly for the clock speed, but by simply resetting them to baseline he was able to return to a clean setup.

After experiencing my own RAM instability from my own pushing of my own sticks, I could tell that when the system started calling on a lot of RAM it would crash, and looking up the native timing made it a quick and easy fix. Now, I'm not saying that it was the only thing wrong with that set-up, but it fixed his issues and now he's stopped talking about buying a new computer... which is too bad really.

Lets say your a serious gamer, or even a casual gamer... there's nothing WORSE than your system going belly up during a critical moment when your team needed you... and you crashed because of something that with a little knowledge and testing you could have prevented.

My little cousin had crash after crash in League that made the experience horrible for all of us. If he had given me access sooner, or if I had known how to fix it sooner, maybe I could have done something and we would have avoided a lot of bad gaming situations.

Anyways, that's a newbie's take. Take it for what it's worth