Memory problems

[Felipe Muniz]

I'm experiencing crashes on random applications like games, Firefox, TeamSpeak, a few times a day - but I don't get any blue screen. Yesterday I decided run Prime95 test, and I noticed that calculations on Large FFT and Blend was sometimes crashing in 2 seconds, sometimes 10 minutes. First crash in one core, a few minutes later other core, until one hour later the last core crash.

 

Prime95: "Rounding was 0.5 expected less than 0.4"

 

I noticed that if I use the two sticks on dual channel, the crashes get more intense, I don't know why. Using the two sticks on single channel, stays more stable, but still crash on Prime95 after one hour or less.

 

Well, I'm using two memory sticks, 4GB each, 1333 DDR3 9-9-9-24 @ 1.5v and after changing slots etc, I decided to test just one stick per try, and I find out that one stick is OK and the other is "BAD", giving this errors on Prime95. I left the "Good" stick testing on Blend mode over night and get 0 errors on all cores.

 

Now, I'm playing with the "Bad" stick here. On default settings (1333MHZ 9-9-9-24 @ 1.5v) it crashes as expected. So then I changed this to 1066, 11-11-11-28 @ 1.8v (i think that is a high voltage)  -  and it's running prime95 for a least 2 hours now and until now 0 errors.

 

Why this? This make any sense to you guys? What's going on here?

[ShiroArikawa]

"Looks at title thinks: ginkgo biloba..."

[MageTank]

I'm experiencing crashes on random applications like games, Firefox, TeamSpeak, a few times a day - but I don't get any blue screen. Yesterday I decided run Prime95 test, and I noticed that calculations on Large FFT and Blend was sometimes crashing in 2 seconds, sometimes 10 minutes. First crash in one core, a few minutes later other core, until one hour later the last core crash.

 

Prime95: "Rounding was 0.5 expected less than 0.4"

 

I noticed that if I use the two sticks on dual channel, the crashes get more intense, I don't know why. Using the two sticks on single channel, stays more stable, but still crash on Prime95 after one hour or less.

 

Well, I'm using two memory sticks, 4GB each, 1333 DDR3 9-9-9-24 @ 1.5v and after changing slots etc, I decided to test just one stick per try, and I find out that one stick is OK and the other is "BAD", giving this errors on Prime95. I left the "Good" stick testing on Blend mode over night and get 0 errors on all cores.

 

Now, I'm playing with the "Bad" stick here. On default settings (1333MHZ 9-9-9-24 @ 1.5v) it crashes as expected. So then I changed this to 1066, 11-11-11-28 @ 1.8v (i think that is a high voltage)  -  and it's running prime95 for a least 2 hours now and until now 0 errors.

 

Why this? This make any sense to you guys? What's going on here?

Manually dial in your timings. Do not leave them on auto. Do the same with your vDIMM and VCCIO/VCCSA voltages. Rounding errors can be caused by too much VCCIO/VCCSA voltage, or bad training from the board. If you have an option called MRC Fast Boot under ram, turn it off. Since the problem seems to become more intense under dual channel, You can look into loosening your tertiary timings. If you can tell me your motherboard model, I can tell you which specific timings to adjust.

[Felipe Muniz]

Manually dial in your timings. Do not leave them on auto. Do the same with your vDIMM and VCCIO/VCCSA voltages. Rounding errors can be caused by too much VCCIO/VCCSA voltage, or bad training from the board. If you have an option called MRC Fast Boot under ram, turn it off. Since the problem seems to become more intense under dual channel, You can look into loosening your tertiary timings. If you can tell me your motherboard model, I can tell you which specific timings to adjust.

 It's a ASUS P5P43TD, I'm using with a Q8300.

 

 

Update: I just restored the settings but leaving it on 1.8V and the prime95 crashed, so as you said it's a timing problem, right?