Prime runs twice as long before failure

[trexpesto]

I am testing my OC with Prime. Specific settings are not relevant to my question, don't go there.

With a change in settings, it runs twice as long.
Assuming all else is invariant, does that mean it is twice as stable? Or do I have to repeat the tests to increase statistical certainty?

Same with other stress tests?

 

[GuitarDaddy]

nope, just means its unstable

unstable x 2 = unstable

 

[TrevorRC]

Guitar, ALL systems, eventually, given enough time, will give out erronous data.

No, it does not mean it's 'twice as stable'--if you run it again, it may take half that time again to churn out an error.

It does not run for a certain amount of time, building up stress, 'till the breaking point.

Rather, it runs around a track over... and over.. and over again, and waits for a slip up to happen.
Sometimes an athelete slips up early in the game.
Sometimes it slips up late.

Slipping up early is a good indication of instability, but ALL systems produce errors eventually. No errors for 24 hours is a good way to check for stability, nonetheless.
If you run Prime for 18 hours and it fails, you can run it again and it won't fail for a full 24. Just statistics. Anomalies always show up.

--Trevor

Edit: Same reason why servers use Raid 1/Variants.
Hard drives, though at stock factory settings, EVENTUALLY fail. It can happen at the beginning, or 2 years down the road. But eventually, they WILL fail. Hence why you use two [Basically, repeat everything I said up there and tweak it a tad bit.]

Edit again:

Google Six Sigma.

 

[DaNorthface]

i would agree with guitar.. it's just unstable... unstable systems are useless. yes you would have to run the test a few more times to get better data

 

[Boogak]

That would be an erroneous assumption. If it took 4 seconds to error out instead of 2 seconds, would you call that twice as stable? :Q

 

[wrangler]

Yea, how long is twice as long? Are we talking 3 hours to 6 hours, 6 hours to 12 hours.......I think if we are talking hours of operation, you could say that it is a measure more stable than it was, if the actual time to failure with each configuration is reasonably repeatable. I don't think, however, that anyone here is going to agree that a doubling of time to failure with this kind of test equals a finite number ("twice as stable", 100% more stable etc. etc.) indicating a certain level of stability. It's like asking for a 3dmark type score out of Prime 95.

 

[A554SS1N]

Hmm, it's still 'unstable' in theory if it fails but if it's stable for a long enough time (24 hours?) it's stable enough for most uses, just not Prime stable.

 

[trexpesto]

2 hours versus 1 hour

 

[allies]

^Unstable. Try to shoot for at least 18 hours of stability.

 

[trexpesto]

TrevorRC wins a everyone else read his/her post, #3

 

[bjc112]

Originally posted by: TrevorRC
Guitar, ALL systems, eventually, given enough time, will give out erronous data.

No, it does not mean it's 'twice as stable'--if you run it again, it may take half that time again to churn out an error.

It does not run for a certain amount of time, building up stress, 'till the breaking point.

Rather, it runs around a track over... and over.. and over again, and waits for a slip up to happen.
Sometimes an athelete slips up early in the game.
Sometimes it slips up late.

Slipping up early is a good indication of instability, but ALL systems produce errors eventually. No errors for 24 hours is a good way to check for stability, nonetheless.
If you run Prime for 18 hours and it fails, you can run it again and it won't fail for a full 24. Just statistics. Anomalies always show up.

--Trevor

Edit: Same reason why servers use Raid 1/Variants.
Hard drives, though at stock factory settings, EVENTUALLY fail. It can happen at the beginning, or 2 years down the road. But eventually, they WILL fail. Hence why you use two [Basically, repeat everything I said up there and tweak it a tad bit.]

Edit again:

Google Six Sigma.

I don't agree with all of that..

If I have my system on the border of stability and it crashed @ say 3 hours when I am pushing too hard, it will not then run for 8 hours the next run.. It was crash around the same time..

No one is saying you should run prime95 ofr 200 days and several thousand hours, but if it can't go 24 hours.. Its TECHNICALLY, not stable.

 

[Jojo1971]

Originally posted by: GuitarDaddy
nope, just means its unstable

unstable x 2 = unstable

:thumbsup:

 

[aka1nas]

Originally posted by: bjc112

Originally posted by: TrevorRC
Guitar, ALL systems, eventually, given enough time, will give out erronous data.

No, it does not mean it's 'twice as stable'--if you run it again, it may take half that time again to churn out an error.

It does not run for a certain amount of time, building up stress, 'till the breaking point.

Rather, it runs around a track over... and over.. and over again, and waits for a slip up to happen.
Sometimes an athelete slips up early in the game.
Sometimes it slips up late.

Slipping up early is a good indication of instability, but ALL systems produce errors eventually. No errors for 24 hours is a good way to check for stability, nonetheless.
If you run Prime for 18 hours and it fails, you can run it again and it won't fail for a full 24. Just statistics. Anomalies always show up.

--Trevor

Edit: Same reason why servers use Raid 1/Variants.
Hard drives, though at stock factory settings, EVENTUALLY fail. It can happen at the beginning, or 2 years down the road. But eventually, they WILL fail. Hence why you use two [Basically, repeat everything I said up there and tweak it a tad bit.]

Edit again:

Google Six Sigma.

I don't agree with all of that..

If I have my system on the border of stability and it crashed @ say 3 hours when I am pushing too hard, it will not then run for 8 hours the next run.. It was crash around the same time..

No one is saying you should run prime95 ofr 200 days and several thousand hours, but if it can't go 24 hours.. Its TECHNICALLY, not stable.

Or, more quanitifably, most would probably consider a desktop PC stable with something under 5 nines of reliability(i.e. 99.999% uptime), but probably more than 3 nines (99.9% uptime).

 

[furballi]

Stability is a sliding scale. A change in case temp and room temp will affect stability. There are many other factors that will also affect system stability. A general rule is to find the point of failure in Prime95, then reduce the CPU core speed by at least 50MHz. I often use 75 to 100MHz. Never had any issue.

 

[Kakumba]

Well, I have never seen a stable computer, and never will. Stable means it never crashes. EVER. And I defy anyone with a 2 or 3 year old overclocked comp to say honestly that it has NEVER crashed..... Aside from those first times when you are doing the overclock. That doesnt count.

That said, if it makes it through 24 hours, it "stable enough that I will use it".

Finally, in response to the question, no.