Jump to content
Sign in to follow this  
raphaelcvaz

Prime: Small FFTs or Blend, to check 100% stability ?

Recommended Posts

Use 1024-4096 fft, in place, 20mins per fft 1st.

After you get 8hrs-12hrs stable, try changing in place to 384 megs if you got 512megs of ram.

 

This could fail, badly.

If it fails try again and again, or do in place yet again and keep hammering it.

Eventually you'll be 100% stable.

Double check with s&m.

 

You don't really need to get out of in place mode, just use s&m to check the other stick after breakin/priming...

 

Good luck.

 

Small fft's, you only need those if you're tying to break in the cpu, and the cpu can't handle a large fft without huge probs.

You keep at it, and when one fft is stable you move on to a larger value, until you can run the norm 1024-4096 fft.

 

Don't bother with that crap blend mode, do custom all the way.

Share this post


Link to post
Share on other sites

run small so the CPU NEVER gets any sort of break, and memtest since it tests memory 100x better than high FFT prime(which is what blend switches between - low fft for cpu, and high for ram) does.

 

small fft failed at 8.5 hrs for me, 2400 MHz 1.45v, while blend was able to go on for almost a day.

Share this post


Link to post
Share on other sites
Guest Crusader

Apart from the several tests already mentioned you also can (should) use real games and 3dmark for testing stability. A few hrs playing Quake IV and HL2 and 30 loops! of 3dmark 01 and a few 03 and 05 tests qualifies as pretty damn stable for me.

 

What's more in the past I have been stable for several hrs in Prime only to freeze after a few minutes in 3dmark and/or Quake IV ... so that's why I now use a combination of both to test stability.

Share this post


Link to post
Share on other sites

I'm going to have to disagree..... I've had small fft's 24 hour stable, and still crashing out of games, so I've re-tested with large FFTs and it fails after about 6 hours. Test again with same MHz and volts, small fft’s no errors after 24h again and still crashing out of games. Conclusion unstable OC, backed off 100MHz, this time large fft's no errors after 24 hours, and no crashing out of games!!!!!

 

So now when I validate an OC as stable it must pass 24 hours of prime95 torture test Large FFTs and at least 8 hours of memtest86 full range of standard tests. Never once have I had an OC crash out of a game when using this combo!!!

 

I guess that why they put in brackets after Large FFTs (Maximum Heat and Power Consumption)

Share this post


Link to post
Share on other sites

blend blend blend! (or custom using a blend of sizes)

 

I disagree completely with the advice in post #2 (about using only 1024 and above)

 

Different sizes will fail under different conditions. I once had an overclock that was marginal, and would always and only fail at a 320K length. So I got the wrong impression that I should only test on 320K, to reduce the time it took to see if a combo was stable.

 

Well, guess what, I finally got it stable at 320K, and then one day tried blend again and failed. The fact was it simply needed more vcore. I could get it 99.9% stable very easily, but when you are failing Prime only in the 12-36 hour range, it is frustrating and you tend to look for weird explanations like memory settings etc. In the end my X2 simply needs a lot more vcore to be 100% stable. (3-7 day stable!)

 

The whole point of blend is that is gives you variety. The small blends are good because they run 1-2C hotter, which helps test for heat problems. You are testing for stability, which means you want a wide variety - otherwise you are passing a very limited test which might miss a major problem.

 

For example, I am 3 day Prime 95 stable at 2672MHz, but only on cool days - once ambient gets over 30C, and my temps hit 47/50/46C or so, I get failures - and at small sizes where the heat is really generated. Backing off to 2662Mhz means I can cope with a lot higher temps. But if I only tested on 1024 and above I wouldn't have found this weakness.

 

Anyway, I hope I have made my point. Variety is a good thing in testing.

Share this post


Link to post
Share on other sites

Prime = heat lol :.

Prime is'nt crap when it comes to heat.

Try s&m or a game.

 

Memtest shoudl only be used for testing somehting quick, move on to windows to finilized testing, as memtest is'nt anything stressful.

Neither is prime really but that's what the topic was about.

Share this post


Link to post
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

Loading...
Sign in to follow this  

×
×
  • Create New...