Sign-in / Sign-up
Your question
Solved

Memtest86+ counts 5 errors for 10 hours running

Tags:
  • Memory
Last response: in Memory
December 1, 2013 2:40:59 PM

Hi guys,

As in the title, my Corsair Vengeance LP Blue 16 Go (4x4) 1600 CL9 fails memtest86+, as in the screenshot.

Motherboard : Gigabyte GA-990XA-UD3 ; CPU : AMD FX-8320 BE.

PRIME95 freezes in less than 5 minutes, hard reboot needed. Everything else works fine, included FurMark, PCMARK 7, 3DMARK 11 and CPU-Z reports everything as expected.

I've tried setting them to CAS 11-11-11-29, same thing, I've tried running them at 1.65v instead of 1.5 guaranteed by Corsair...

Please help me !

More about : memtest86 counts errors hours running

a c 143 } Memory
December 1, 2013 2:50:25 PM

Replace the stick with the error. If they were a matched set, be sure to replace the stick with one that has the exact same specs. Timing, speed, brand, model number, etc.
m
0
l
December 1, 2013 2:52:30 PM

clutchc said:
Replace the stick with the error. If they were a matched set, be sure to replace the stick with one that has the exact same specs. Timing, speed, brand, model number, etc.


I've updated my request, can you check please ?
m
0
l
Related resources
a c 5390 } Memory
December 1, 2013 2:53:23 PM

On AMD might try a slight voltage increase, say + 0.06 so if DRAM is rated 1.5 try the DRAM voltage at 1.56 and try that
m
0
l
December 1, 2013 2:55:09 PM

Tradesman1 said:
On AMD might try a slight voltage increase, say + 0.06 so if DRAM is rated 1.5 try the DRAM voltage at 1.56 and try that


Already tried up to 1.65 and getting that screen in the morning.

Isn't it dangerous for the memory sticks to go higher in voltage ?
m
0
l
a c 143 } Memory
December 1, 2013 3:03:05 PM

1.65 is more than the memory controller can safely handle. I would leave it at 1.5V and try prime95 again. On your pic, it appears you only have an error show up on the 8th pass. But it looks like you are testing with all the sticks in at the same time. You should test each stick individually.

Are you O/C'd when you run Prime95?
m
0
l
a c 5390 } Memory
December 1, 2013 3:04:38 PM

Wouldn't go to 1.65 try the 1.56 and raise the CPU/NB at 1.2
m
0
l
December 1, 2013 3:23:03 PM

clutchc said:
1.65 is more than the memory controller can safely handle. I would leave it at 1.5V and try prime95 again. On your pic, it appears you only have an error show up on the 8th pass. But it looks like you are testing with all the sticks in at the same time. You should test each stick individually.

Are you O/C'd when you run Prime95?


Is there a way to test each stick individually without unplugging the other three ?

Nope everything was set up at normal speed, and the memory voltage was raised to 1.65.
m
0
l
December 1, 2013 3:25:57 PM

Tradesman1 said:
Wouldn't go to 1.65 try the 1.56 and raise the CPU/NB at 1.2


How long should I run memtest86+ on each 4 Go memory stick to be certain ?

Do you believe that too high voltages could have destroyed the memory even if it doesn't overheat ?
m
0
l
a c 143 } Memory
December 1, 2013 3:31:53 PM

LorenzoDaLione said:
Tradesman1 said:
Wouldn't go to 1.65 try the 1.56 and raise the CPU/NB at 1.2


How long should I run memtest86+ on each 4 Go memory stick to be certain ?

Do you believe that too high voltages could have destroyed the memory even if it doesn't overheat ?


You have to remove the other sticks, of course... if you want to test for the bad stick.

Memory overheating doesn't necessarily show up due to over volting it. But nothing kills electronic components faster than applying to high a voltage.

Most of the time, memtest will show an error after one completed pass. I've never had to run it overnight or anything like that to find a bad stick. But if you are running memtest 10 hours with the memory voltage set at 1.65, you aren't doing the sticks any good. may in fact be causing the errors.
m
0
l
December 1, 2013 3:49:40 PM

clutchc said:
LorenzoDaLione said:
Tradesman1 said:
Wouldn't go to 1.65 try the 1.56 and raise the CPU/NB at 1.2


How long should I run memtest86+ on each 4 Go memory stick to be certain ?

Do you believe that too high voltages could have destroyed the memory even if it doesn't overheat ?


You have to remove the other sticks, of course... if you want to test for the bad stick.

Memory overheating doesn't necessarily show up due to over volting it. But nothing kills electronic components faster than applying to high a voltage.

Most of the time, memtest will show an error after one completed pass. I've never had to run it overnight or anything like that to find a bad stick. But if you are running memtest 10 hours with the memory voltage set at 1.65, you aren't doing the sticks any good. may in fact be causing the errors.


I believe there were 2 errors in test 8 at every pass, shouldn't it be more random if too high voltage was the cause ?

Could it mean that there is only one stick damaged, so errors happen always exactly at the same test ?

What does the "bits in error" part of the report mean ?

Thank you for responding so fast by the way.
m
0
l

Best solution

a c 5390 } Memory
December 1, 2013 4:27:01 PM

It may well be the app itself, Memtest has been know to give false errors in passes 7-8, have you had problems with them at their stock settings?
Share
December 2, 2013 3:36:23 AM

Well yeah there's the Prime95 problem.

I've been testing each stick with prime95, and two of them caused different crashes : One froze the computer like before, and the second made it hard reboot without a warning.

So i've tested the two other ones with memtest to make sure, and thay have passed 4 times without an error, only memtest is showing those as DDR3-805, timings 4-5-5-15. Is this normal? If these are the real frequences and timings, would it make them pass when they actually defect in the right frequencies ?

I haven't changed a thing in the bios though.

I'm testing right now the stick that made my computer hard reboot during Prime95 yesterday, so far going through test 10 without a single error and showing the same bullshit frequencies and timings as the other two.
m
0
l