Stability testing To PRIME or not to PRIME that is the question!

[Justintime]

To each their own but at the end of the day Intel would not ship a CPU if it was not stable in Prime95, processors are supposed to be stable under all conditions. It's only a recent phenomenon due to out of control TDP levels (high end GPU's, Bulldozer/Piledriver) that throttling has become a 'feature' in stressful conditions to protect the hardware.

Most people take the anti-Prime stance because it tells them something they don't want to hear. You should use a variety of tools not just one or two, a stable processor shouldn't fail in anything software bugs aside.

This. The "my cpu fails prime but is game stable" crowd loves this. Proper overclock should handle EVERYTHING a stock cpu can. Also a lot of people use massively outdated versions or wrong (32 bit on a 64 bit os etc.) and don't properly test modern cpus, hence passing prime but failing gaming etc..

 

[Make]

This. The "my cpu fails prime but is game stable" crowd loves this. Proper overclock should handle EVERYTHING a stock cpu can. Also a lot of people use massively outdated versions or wrong (32 bit on a 64 bit os etc.) and don't properly test modern cpus, hence passing prime but failing gaming etc..

But if a person ONLY ever does gaming, why would they need stability for anything else?

You can replace *gaming* with anything there by the way. My temps go way too high on IBT, yet I don't use IBT daily, weekly, monthly, or ever. So it's not stable for IBT but it is for everything else I do.

By the way how the heck do you know your PC has done a "maths" error if you get a corrupt driver for example? It can be anything from memory to ssd to sata-controller corrupting it..

 

[Doomedspeed]

But if a person ONLY ever does gaming, why would they need stability for anything else?

You can replace *gaming* with anything there by the way. My temps go way too high on IBT, yet I don't use IBT daily, weekly, monthly, or ever. So it's not stable for IBT but it is for everything else I do.

I 100% see your point but the point is you don't play the same games all the time and newer games will have newer CPU requirements.

Say you set your Overclock up to play MW3 and the CPU was only used 60% (with you un-prime tested overclock) then you fire up battlefeild 4 and you get 90% usage and a BSOD...

So you'd need to fix your overclock before you can play the game, i'd find that pretty frustrating.

You can also replace "gaming" and the "game names" with any other scenario..

 

[swiftmetal]

Don't worry, your brand new car has already had the nuts thrashed off it at the factory ( I used to do it)

I really want your job lol

Just make sure you do tests with your GPU running as well, otherwise you might find that you get overheating even if the CPU is prime stable.

This is a very good point that I totally overlooked seen as your temps look stable in prime and then add some gpus to the mix and bang your case temps sore or even if you are water cooling your coolant temps could go up a lot too>

AGAINST:

For some people "stable enough for what I do" is enough, playing [game] 250MHz faster than you could while completely stable and getting the weekly blue screen in exchange is fine for some people and I know a bunch of them.

I am totally guilty of this as this is why I spend money on my system to plays games and the faster and the higher quality the better.

 

[swiftmetal]

Prime No need!!!

Encoding, Gaming, Editing, win rar, browsing, folding or whatever you use your PC for!!! Do need!!!

Nuff said really

I 100% see your point but the point is you don't play the same games all the time and newer games will have newer CPU requirements.

Say you set your Overclock up to play MW3 and the CPU was only used 60% (with you un-prime tested overclock) then you fire up battlefeild 4 and you get 90% usage and a BSOD...

So you'd need to fix your overclock before you can play the game, i'd find that pretty frustrating.

You can also replace "gaming" and the "game names" with any other scenario..

I do appreciate your point and do get where you are coming from but if you want gaming stability why no just put any of the new versions of 3D Mark on loop and that should give you a good idea of gaming stability.

 

[Doomedspeed]

I do appreciate your point and do get where you are coming from but if you want gaming stability why no just put any of the new versions of 3D Mark on loop and that should give you a good idea of gaming stability.

But thats not natural use either?

That was youre point wasn't it?

 

[swiftmetal]

It is a test that is more suited to gaming and it was just a suggestion for people that must prove to them selves that their setup is stable within games before actually gaming. It has more realistic temps and loads levels and stresses bot cpu and gpu to "gaming levels" and is a better indication that overclocks in cpu and gpu are gaming stable.

Personally I just oc then go on a game if it crashes so bet it mess rinse and repeat.

 

[Justintime]

But if a person ONLY ever does gaming, why would they need stability for anything else?

.

If they don't mind every now and then a game crashing, good luck to them. Because it dosen't crash on say Doom 3, dosen't mean it won't crash on GTA 4 as that may do something or try to calculate something and get it wrong then crash, whereas D3 did'nt run that routine.

 

[Scougar]

FOR. It's a quick and easy stability check. If it's failing quickly in Prime, then go back to the drawing board. Every time I've failed prime, it's failed in game, it just takes longer in game, or I get bad side affects from a bad overclock that prime identifies quickly.

It's not the be all and end all, but it's a quick indicator.

People doing benches are probably not going to bother with Prime as they just want the highest OC, and will keep trying that OC at different settings until they pass whichever bench it is they want to satisfy, they aren't particularly interested in stability beyond the bench, only that the scores are greater.

If you are having crashes with a game, a quick way to potentially identify an overclock issue is to run a prime session (or similar). This might pick up a problem, rather than you wasting your time trying to find a config problem, or a fix you need.

 

[almighty15]

50 runs of AVX Intel Burn Test with all RAM tested = Job done..

No need for hours upon hours of Prime

 

[PCZ]

BOTH.

FOR:

If the system isn't stable, then it isn't stable, it's that simple. A lot of people forget that Prime is a real world application not a synthetic test like IBT, if it fails in Prime then it may well fail in F@H or BOINC or Bitcoin. Also just because it is stable in Prime doesn't mean it will be stable in everything.

Personally I prefer to tune my system to be as stable and as fast as possible, without sacrificing either.


AGAINST:

For some people "stable enough for what I do" is enough, playing [game] 250MHz faster than you could while completely stable and getting the weekly blue screen in exchange is fine for some people and I know a bunch of them.

A lot of people don't realise that P95 is a real world application.
Something I used to use a lot.
Those forum members involved in DC (Distributed Computing) will know what it is for.
Those with a lot of BOINC projects attached will most likely have done work for PrimeGrid which uses the P95 (GW NUN) math libraries for the majority of their projects.

Before using a PC for any of the DC projects it was common practice to run P95 to prove some level of stability before doing real work.
Returning bad results helps no one was the mantra.

Having a P95 stable computer meant that it was also very stable at normal everyday tasks gaming included.

Things have changed a bit since the AVX CPU's become available.
Using the AVX instructions speeds up number crunching a lot and consequently DC computing and encoding APPS are using it a lot.

Games do not make much use of AVX and testing with AVX enabled P95 may not test the integer and standard FPU instructions that games are using.
P95 used to, but not any more.

Generally the AVX instructions will start to fail at lower clocks than the traditional X86 instructions but that is not always the case.

P95 can still be useful to a gamer as it will quickly weed out very unstable CPU and memory clocks.
No need for long punishing tests though unless you are going to get involved in DC or are going lone wolf to find the worlds biggest prime

Gamers can use looped benches such as RE6, to test stability.
A favourite of mine as it is sensitive to instability in the RAM or CPU.

 

[Glanza]

I run prime for 30min after building to get a base idea of temps.

After that if no issue I tend not to use it anymore and just do my normal daily routine on the PC and test that way i.e gaming for several hours at a time.

 

[Sin_Chase]

You should us a combination of activities to stablity test.

Throw everything you can at your overclock to stress it. For example - I could have my Q6600 at 3.6Ghz stable in ABSOLUTELY EVERYTHING but Battlefield BC2 would crash it within 10 minutes.

Thinking any one tool will give you a stability rating of 100% is foolish.

Use Prime, IBT etc for looking at temps and your cooling capacity. It might reveal an instability a higher temps. Gaming might reveal a real world usage instability, perhaps as a result of the GPU being thrown into the mix - more heat, more power draw.

You wont know for sure until you settle on a frequency and just use it. I never ever ever leave my system overnight running some test....what's the point. Change on the fly, as and when you experience issues.

 

[GinG]

Prime No need!!!

Encoding, Gaming, Editing, win rar, browsing, folding or whatever you use your PC for!!! Do need!!!

Do you prime test your systems sold on overclockers??

 

[mmj_uk]

I'd feel ripped off in buying one if they didn't.

If Intel and AMD sold a processor which failed running Prime95 at stock it would classify as being faulty.

That said no single test is 'definitive', a stable processor shouldn't crash/fail at running anything (software bugs aside). Picking and choosing which tests you prefer and ignoring others is just a way to claim higher overclocks as being stable when they aren't.

 

[Make]

Clocks might be stable on everything else except temps.

Prime produces high temps on clocks that might be just fine on any other application.

Only difference is, do you want to waste money on a cooler that can keep prime going for 8 hours on an otherwise stable PC, or not.

 

[PCZ]

Do you prime test your systems sold on overclockers??

A very good question

I would think that if you are going to sell a guaranteed OC that it would be capable of maintaining that clock on all tasks.

I guess in the main the Overclocked bundles are sold to gamers.
However If someone bought one and started running DC a lot, I wonder how that would work out.

 

[Justintime]

A very good question

I would think that if you are going to sell a guaranteed OC that it would be capable of maintaining that clock on all tasks.

I guess in the main the Overclocked bundles are sold to gamers.
However If someone bought one and started running DC a lot, I wonder how that would work out.

I don't know if it was this guys builds or not, but i have had to redo 3 overclockers 'overclocked' pcs as they were ultimately not stable (crashing explorer, bsod on games or even idle etc..). The last was about 4 months back. One was on my own recommendation for a mate who lived too far for me to build one lol. Far as i remember it was vcore and VTT being the main culprit.

Clocks might be stable on everything else except temps.

Prime produces high temps on clocks that might be just fine on any other application.

Only difference is, do you want to waste money on a cooler that can keep prime going for 8 hours on an otherwise stable PC, or not.

Prime is not an 'unrealistic' load as some make it out to be, there ARE programs out there that can get the cpu just as hot.

I'd feel ripped off in buying one if they didn't.

If Intel and AMD sold a processor which failed running Prime95 at stock it would classify as being faulty.

That said no single test is 'definitive', a stable processor shouldn't crash/fail at running anything (software bugs aside). Picking and choosing which tests you prefer and ignoring others is just a way to claim higher overclocks as being stable when they aren't.

Nicely put.

 

[Telecaster]

How does a CPU pass Prime95 and IBT but BSOD whilst doing nothing stressful?

 

[Justintime]

How does a CPU pass Prime95 and IBT but BSOD whilst doing nothing stressful?

Usually the load line calibration is set too high so when idle vdroop drops the vcore past a stable limit for the non loaded frequency.