Prime inconsistent instability failing or frustrating???

[-=apexi=-]

Hi I'm new to the forums.

This is my first custom build and I've got incredibly immersed into the world of overclocking. It makes no sense whatsoever, but I found I couldn't stop from the challenge of o/c these things: its like the challenge of solving a puzzle!

All was fun and exciting until I heard about this thing called Prime and everyone talking about "Prime Stable." I made the mistake of downloading it and my o/c adventure turned upside down. Like many others, my system could be totally stable with games and all the other benchmark/stability tests out there. But start up prime and it fails in seconds or minutes, and definitely within an hour.

I tried everything, every variable. I spent weeks specing out my system with quality components: Fortron PSU, DFI Lanparty NF4 Ultra-D mobo, Crucial Ballistix DDR500 sticks, liquid cooling, etc...the works.

I filled up a whole excel spreadsheet of various vcore, vdimm, memory timing combinations in my relentless pursuits to get my system prime stable. After trying everything over the course of days or was it weeks, I started surfing forums about prime95 problems.

I found the camps were split 50/50 on the believers in prime. I was about to join the camp that feels that prime is overkill, unnecessary, buggy, flaky, etc. But in the final night of my final attempt, I finally found prime stability over 12.5 hours!

So to those walking this path now, please learn from my pains:
I know you've researched the basics on overclocking since there are so many guides out there, but if you are having problems getting your system stable, remember this:

1. NOT ALL CHIPS ARE SAME so do NOT compare YOUR chip to those of a different week/stepping !! I got a week 22 venice 3200 very recently. I was reading all the forums about guys hitting 2.7/2.8 ghz on stock volts at 1.4/1.5V. My chip's upper limit is 2.6ghz, and that requires 1.68V for prime stability!! Remember this FACT and save yourself endless grief.

2. Don't be afraid to increase voltage as necessary. From the example above, I was expecting to hit 2.5 at near stock volts. Of course, I didn't dare venture beyond the 1.5V limit for a long time, trying everything else in vain, because "everyone else was able to do it on 1.4V, etc" It actually took my chip 1.52-1.54V to get my chip prime stable.

3. Watch your temps! The NF4 chip on my DFI board was cooking! I moved the waterblock on my GPU to the NF4...now the thing is cooool...everything is happy.

4. Get a GOOD power supply. Spend the cash, get GOOD cooling, liquid if possible, and keep an eye on your temps as this is directly related to component life. Keep it running 40-50C overclocked and all will be good.

5. Before you throw in the towel on Prime, try the above and forget all the online stories of my system does 7.5 ghz on 1.3V at 25C, etc. Every system is different, every o/c is a unique challenge...that's part of the fun!


http://members.shaw.ca/apexi/oc.jpg

[CRUSH-ER]

Good tips for noobs, but most of us already know this. :)

[ccokeman]

An all to familiar story for me. The first week or so after a hardware upgrade is all long nights trying to get the max stable clock speed. Screen shots are nice but reliablity is even better for me.

[carmelo]

People should use a variety of software. I use S&M and OCCT for my main over clocking. They tend to pull up the errors prime would normally pull up quicker. Than at the very end when I find the speed I would like to stay at, ill run prime for a bit... it’s overrated in what it does. It’s a proven and tested proven, but since it is no one is willing to give other programs a chance. Prime is a good at what it does, but others do the same thing in my book.

6. Just because your prime stable doesn’t mean your computer will never crash. Programs will always have errors, and windows will always be buggy. So you still may crash with 1.89v and prime stable.

7. If you’re not heavily using your machine, you may not have to be prime stable if you never use your comp to the fullest extent than maybe a lower vcore would be ok. You still may be at risks for unreliability but if your not stressing out your comp you MAY be ok.

8. Try a burn in process. A lot of people have tried a type of burn in and found success with it (including my self). In the end it may save you on having to run a higher vcore. I can run my vcore at 1.6 @ 2.6mhz, unlike before where I needed at least 1.62+ to be stable.

[-=apexi=-]

Yeah...it's too bad nobody else really uses any of the other (perhaps better) programs out their for stability testing than Prime95. It does take very long when your system is NEAR stable to see if it's really stable. Nothing sucks more than to see Prime crash after 45 mins, or 1.5 hours after you gone to sleep.

At least it finds relatively UNSTABLE systems very quickly...within less than a minute or 10-15 minutes at most.

One thing I found really helps, is to run Prime95 AND 3DMark05 at the same time for at least 1 cycle of 3DMark05, it should only take 10mins or so. If Prime is still running after this, chances are real good it will be prime stable for the long haul.


http://members.shaw.ca/apexi/oc.jpg