Originally Posted by sdlvx
I appreciate you trying to deduce the problem, but you don't need to deduce when you have evidence.
Through google search results, we have mounds and mounds of evidence that shows that Prime95 has problems with new architectures, even at stock settings, and those problems are fixed with an update to Prime95. I realize that patterns don't always repeat themselves, but the possibility of Prime95 needing changes to its code in order to run properly on AMD Piledriver are high given the history of Prime95 and new architectures.
But that is the question, whether it requires any changes for Piledriver at all or not.
It appears the developers are aware of Piledriver and are discussing optimizing for it.
However, there is no mention of any known problem for the current version.
AFAIK Prime95 is closed source.
You run a stability test to see if your overclock is 100% stable.
In order to do that, you need a tool that you have 100% confidence in that it will tell you when your system is unstable.
Prime95 is definitely not that tool. It has a history of not working with new architectures for unknown reasons without overclocks being applied.
Well, the change log and the source code would surely show what those "unknown reasons" could have been.
Let's not confuse different categories of problems.
If a test program fails systematically without OC on a new chip, then it's easy to dismiss it as useless.
And I would have done that myself if that was the case, believe me.
However, I certainly don't see anyone saying Prime95 fails systematically on Piledriver without OC.
Another posted said somebody had it fail on PD without OC, but I couldn't find any reference to that instance myself.
My own run at stock settings is going 8 hours now and no errors - that's 1 hour longer than I have had it going with any OC setting, and I'm keeping it overnight.
Now, we are fighting over if a benchmark that has a history of not being stable on new architectures until updated should be used on to test stability on a brand new architecture while that benchmark has had no updates to support that processor?
There are lots of tools that do the same thing as Prime95. We have OCCT, IBT, AMD Overdrive Tests, etc.
Not every chip necessarily requires an update. Most of the Prime95 updates have been optimizations for new instructions set as they become available on new chips, not t fixes for the inability to run on new chips, though there have been a few of those certainly.
It seems you and a few others choose to shoot the messenger (Prime95) when it tells you that your OC system is unstable. It just seems unwise to me to ignore it unless or until the reason is known. I prefer err on the side of caution. And since my chip appears to run Prime95 just fine without OC, but not with OC, I can only assume the OC is what breaks the program, ie. the OC is not stable.