Correct way to test overclocking?

[iloqin]

Noob, here sorry if this is a repeated post, I wish there was a search this sub forum option only or in this case, a sticky post on top about the correct way to test overclocks. In any case, I do know that there are lists of free software here that are useful when overclocking, I plan on grabbing them. Things like CPU-Z, prime95, etc. I do know about running prime95 at the END when thinking you're at the stable point from anywhere between 2hrs-24hrs.

The question I have is how long and what test should I run if I am already planning to go back in and push it some more?

This is my idea of testing(probably wrong, but I want to make sure):
1) Up the CPU FSB by 2htz
2) Save
3) Reboot
4) Run Test... (what exactly does this mean? Run Prime95 for a few minutes/24hrs?)

Then rinse and repeat?

Overclock:
AMDAthlon 440 3.0 Rana
Coolmaster Hyper 212+.
ASRock M3A770DE AM3 AMD 770 ATX AMD Motherboard

Anyone know a ballpark figure of where I can push this? Tryin' to stay under 50C if possible, if there is more OCing potential under 60C then maybe I'd shoot for that temp, under load that is.

 

[fisshy]

Increase the CPU FSB by 10MHZ just so you don't die of boredom

Save

Reboot

Log Into Windows



Then repeat...

once you have reached a target Run Prime 95 On small FETs For 2 Hour and then on blended for about 4 hours

Good Luck AMD guy

 

[wolfram23]

I found Prime95 gave errors within 5 iterations if my voltages were a lot too low, but otherwise if you're close it takes hours to find an error. For a lot faster testing grab Intel Burn Test which runs a Linepack test, it can find an error in a few minutes instead of hours. Once stable there for 20 iterations, run Prime95 for 12-24 hrs and if it passes you're all set on the CPU and RAM end (although if the RAM is OCed at all, a Memtest86+ should be run). Prime95 run times are very subjective tho, I haven't done over 3 hours yet but I also am still tweaking a lot to find that sweet spot, as well as to learn what everything does.

 

[jsc]

There have been a lot of discussions here about how long to run what program.

I use Prime95. I run my final tests for 24 hours. I used to think 12 hours was enough. But once, I was testing one of my systems and I just let it run over night. Next morning, I discovered that it failed P95 at 14 hours one night and at 18 hours the next.

 

[EXT64]

For the intermediate steps (which I assume was the original question) I usually do 10 min large FFT. If it passes that, it is not likely to fail to post or bluescreen on the next notch (it is at least moderately stable). Then once it fails, raise voltage/back off. At desired speed, use prime for 12-24 hrs or some linpack test (like linX) for some time (not sure what to recommend as experiences vary, I usually did several hours),