[ekoh]

I just received my brand new 990FX Sabertooth, FX8120, and GSkill 1866 in the mail, fresh from RMA as i "Thought" i had issues with one of the last 3 pieces i RMA'd. I got them all back, I have been attempting to run Memtest 86 v4 from a bootable USB and it always freezes after 2-6 mins. I have tried disabling legacy USB as some people said to try, removing all USB devices, running from a bootable CD(which doesn't even boot). No errors just random freezes. The little underscore stays blinking but the menu and time locks up. I can run OCCT,Linkpack, P95, and HYPER PI all for the same amount of time and runs just fine in windows. I had this same issue with the same but different parts (RMA'd the first batch) and having it again. I have came to the conclusion that possibly memtest doesn't like my setup by nature of design, rather than by error. I am running old school memtest v3.5 no problems no issues. Has anyone else had or heard of issues with memtest86 v4?

I even have a diff video card in while my bad one is off to gigabyte for RMA (yes i know bad luck.)

I will say I only pulled 32sec on hyper pi 1M. Which seems rather low?

Any input would be appreciated!

[ekoh]

I'm an idiot. So far 86+ is running fine. Why I was running only 86 is beyond me. I'll report in the morning.

[larrymoencurly]

I've had many modules fail MemTest86 but pass MemTest86+, and MemTest86+ turned out to be wrong.

[ekoh]

Originally Posted by larrymoencurly

I've had many modules fail MemTest86 but pass MemTest86+, and MemTest86+ turned out to be wrong.

What really is the difference between the two? Is the + version supposed to be multithreaded or something?

edit update:

I've done a little research and found that 86 will give false positives on newer platforms as it's last build was in 2009 and it is not maintained anymore, though + is roughly a year out on it's last build it seems to be more accurate, but then again who knows. I think i have most of my bugs worked out though......for now.

[larrymoencurly]

Originally Posted by ekoh

What really is the difference between the two? Is the + version supposed to be multithreaded or something?

edit update:

I've done a little research and found that 86 will give false positives on newer platforms as it's last build was in 2009 and it is not maintained anymore, though + is roughly a year out on it's last build it seems to be more accurate, but then again who knows. I think i have most of my bugs worked out though......for now.

I don't know why MemTest86 and MemTest86+ have given me different results, despite both using the same testing algorithms. I originally thought it was because their different sizes caused different effects in lower memory, but I've seen MemTest86 report errors even at higher addresses. BTW I've never tested with more than 4GB total memory in any system. I've also gotten different results with Gold Memory 5.07 than with Gold Memory 6.92.

Just wondering, but how do you know that MemTest86 gives false positives unless you have something that's been proved to be better, like Ultra-X's RST diagnostics, or were allowed to test with special modules, like the ones RealWorldTech obtained from Micron for it's tests from 10+ years ago?

[dlang123]

Okay I have an ASUS M5A99X EVO AM3+ AMD 990X and 16gbs (4x4GBS) Geil memory, one RipJaws set, and one basic desktop memory. AMD FX-6100

If I put all 16gbs in and run Memtest 86+ it is fine, passes 100%

If I try Memtest 86 v4.0 I get fails, I tried pulling out 8 gigs, still failed.. tried just 4gbs still fails.

Everything seems to be fine in Windows, but I was using this third party software that aggregates your hard drives to be one giant drive, and on copys the CRC was failing, but when I used the drives normally the copys were perfect.

I put in a defect to him, and after researching some of my failed copies, they were all failing on the same bit in the same word, which lead us both to believe memory issues.

So I thought, "Well maybe memtest 86 v4.0 has a problem with these newer chipsets, thats why I get tons of errors on the run". Now the question here is, which one to believe? Would I believe that all 16gbs of these new sticks are bad? No. I would believe that maybe the motherboard/processor had a bad memory controller first, because this is actually my 2nd motherboard already, the first one I couldn't stop it from blue screening.

Thoughts or suggestions?

['Cuda340]

If your interested & want to try it out, Memtest86+ 5.00 Beta is available.......

Memtest86+ 5.00 Beta

[dlang123]

Sure thing.. downloading now...

Windows Memory Diagnostic is coming up clean too. Makes me think that the old Memtest86 just doesnt work on newer chipsets...

Will report back about memtest86+ 5

[dlang123]

Okay it has passed 100% with all 16gbs in memtest86+ beta 5... would you assume that it is good memory then? I am letting it run for a few more passes to verify.

[dlang123]

3 full passes on Memtest86+ Beta 5, 3 Full Memtest86+ 4.2, 2 Full Passes Windows Memory Diagnostic, and Nothing but fails with Memtest86 4.0

Seems like I can infer that Memtest86 4.0 is just not compatible? That with the + that passed memory is good?