Darn, so close! :)

[Arboleda]

Short story, I've arrived at an overclock (roughly 9 x 290) and 1:1 for memory (OCZ Plat rev. 2) that has seemed stable. I can compute SuperPi to 32M, MemTest86 runs for hours, etc.

 

However, Prime95 produced a rounding error 8 hours into the blend torture test. This kills me! At this point, what would you all do to achieve 24 hour stability? Would you drop 15Mhz off the FSB? Would you first try a memory ratio (9/10, etc)?

 

Believe, I've come to know the principal of modify values and test them, refine and test, refine and test. I've probably done so at least 50-75 times thus far. It's just that if now it takes 8 hours to land at my optimal settings, I thought I'd get advice about what granularity of FSB up and down I ought to try.

[Beer Moon]

Kick down a divider and see if that fixes it. Then kick up to the HTT setting and run at 8x or 6x on teh CPU and see if that fixes it to nail down whether RAM or CPU is responsible. Then adjust accordingly.

 

Take a fraction of the time it took you to get this far hehe.

[loc.o]

If its stable in whatever you do with your comp (which aint priming 24/7 i presume) i wouldnt bother.

[Fight Game]

I wouldn't drop 15 from the HTT. Since no errors in any other benchmark, and it takes this benchamrk 8+ hours, I would lower the HTT by 1 or 2.

Can you complete 3dmark01, 03, and/or 05? Memtest is an ok test for the memory to start, but some have passed memtest for hours, and can't complete any other benchmark.

 

If its stable in whatever you do with your comp (which aint priming 24/7 i presume) i wouldnt botter.

 

It may appear stable for some things, but what happens when you're in the middle of something important (like BF2 or HL2!) and it crashes. You'll never know why without doing all these tests. This method is ok for you, but I wouldn't suggest it to someone looking to stabilize their system. Some even prime for 24 hours, and don't see an error until they really put their video under some strain, and creates enough heat to overheat it and/or something else.

[loc.o]

It may appear stable for some things, but what happens when you're in the middle of something important (like BF2 or HL2!) and it crashes. You'll never know why without doing all these tests. This method is ok for you, but I wouldn't suggest it to someone looking to stabilize their system. Some even prime for 24 hours, and don't see an error until they really put their video under some strain, and creates enough heat to overheat it and/or something else.

 

"Whatever you do with it" includes 3Dmarks/games etc if you do them. He ran 8 hours without an error, that seems pretty stable to me. If 8 hours aint enough what is? Your absolutely right that its up to the actual user what can be considered stable and what not. Ive ran me comp for days on settings that wouldnt prime over 2 hours, but it was rocksolid in any bench or game I play (including BF2). Its all about personal pref. I dont get your graphic card example though, it isnt stressed in Prime and if it is the reason for a BSOD or whatever Prime would never notice it, regardless the time you let it run. In fact thats just what i meant..

[Arboleda]

Thanks for the perspectives, guys. Great academic debate on "how stable" is stable enough for everyday use. To loc.o's point, I'm a software developer which in reality means idle CPU most of the time (as I'm coding) and then spurts as I compile the whole solution, load various programs, test a few components, etc (all under a few minutes worth of stuff). It's just that I want it dang fast when I do that stuff

 

To Fight Game's point, one of my goals for this rig was just rock solid stability with the overclock. Even though I do some IT work on the side and am perfectly used to the imperfections in hardware, OS's, and applications - there's still nothing like the feeling of "hey, I tested my rig for 24 hours of torture testing - it's rock solid and massively overclocked!".

 

Thanks for sharing both perspectives.