Sign in with
Sign up | Sign in
Your question
Solved

Memtest86 - first errors, now no errors..

Tags:
Last response: in Memory
Share
November 19, 2013 12:29:34 AM

This morning I continued running some synthetic benchmarks with different GPU/CPU configurations for a build that I'm working on. Everything was fine last night but I swapped my 3570 CPU out with a 2500 (both stock chips, no K models). I ran a few benches fine but while I was running "Heaven," the app started to crash. I restarted, retried, reinstalled nVidia drivers - then the driver install kept crashing as well and then I restarted and couldn't boot into windows. I did a system restore and still couldn't install my nVidia drivers. I then went to do a clean install on windows and that's when I started getting BSODs. I ran chkdsk to test my SSD and found no faults so I moved on to Memtest86+ to test the RAM.

In about 30mins of running Memtest86 I'd come up with about 3000 errors with one stick, so I swapped it out for the other. I immediately got 21 errors with the second stick but none after that. I went out for a bit and ran the second stick through Memtest and got 0 errors in 2 passes, so I put in the first stick again and also got 0errors in two passes.

Obviously this is ruining my day and seriously hampering my productivity but I'm not experienced with RAM failures or memtest86 and I can't seem to find any information online with results similar to mine (my eyes hurt now and so it's possible that I'm getting too lazy to search as best I can).

From what I can gather, if memtest comes up with a single error, seek RMA. That said, make at least 4 passes to confirm the accuracy of memtest86. This I find confusing, as do I find it perplexing that my RAM was so clearly error prone in the first two runs, but hasn't shown any errors since.

Can anyone shed any light on this for me please? I'm not that keen to sit here for 3 days running memtest on my RAM if I can avoid it, but I also don't feel comfortable building a PC with parts that should and could be replaced... Should I ignore the first results, or keep running memtest for another couple of days, or immediately seek RMA?

More about : memtest86 errors errors

a b } Memory
November 19, 2013 1:46:57 PM

Frustrating, indeed. One thing about memtest (or any other memory testing pgm) is that running it on a cool machine can give different results that on a warm machine. That's why they always recommend extended runs to be sure. However, I'm usually in a hurry and when I buy new RAM I just run it for one pass. So far, so good. But I realize that doesn't help you.
But if your sticks showed errors at all, it would be a good bet that they are defective. Just not all the time. Of course that is assuming the voltage controller/memory controller is providing the correct voltage all the time.

If these are all new parts, I think I'd RMA the memory with the explanation that they failed the memtest run. If the same issue occurs with the new memory, then you'll have a fair idea the issue is elsewhere.
m
0
l
a c 131 } Memory
November 19, 2013 2:05:46 PM

What speed is your memory running at? its possible your 2500 doesn't have a strong enough memory controller to run at what you were aimed at, and at some point between sticks reverted back to 1333 mhz? Ivy definatly has a better memory controller than sandy did. You will almost always get better memory results from ivy.
m
0
l
Related resources
a c 2198 } Memory
November 19, 2013 2:08:01 PM

What mobo/CPU? IF Intel and 1600 or above are these set up under XMP? If DRAM was installed and left at mobo defaults - it may well not be running at optimum (either spec or SPD optimum and you may or may not encounter errors)
m
0
l
a b } Memory
November 19, 2013 2:30:22 PM

I'm running 1866MHZ memory at 1600MHZ on my SB with no issues.
m
0
l
a c 131 } Memory
November 19, 2013 2:34:58 PM

Yes but for all we know he's trying to run it at 2133 full of 8gb sticks. thats why I asked the question :)  If the problem doesn't present itself with the Ivy processor you could have damaged the IMC at some point on the Sandy proc.
m
0
l
a c 2198 } Memory
November 19, 2013 3:49:54 PM

Supahos said:
Yes but for all we know he's trying to run it at 2133 full of 8gb sticks. thats why I asked the question :)  If the problem doesn't present itself with the Ivy processor you could have damaged the IMC at some point on the Sandy proc.

_______________________
+1 The info is needed to help determine what is best possible
m
0
l
November 19, 2013 10:22:20 PM

Running on Asrock Z68 Extreme4 Gen 3 with an i5 when I got the initial fault, XMP 1600Mhz.

Stripped down to Memtest86 on USB, single sticks and no HDD connected for the actual memory tests. I swapped out the 2500 for a 3470 last night and immediately got 20+ errors on each stick (individually), so I'm gonna RMA them...

I've been running all of my other RAM through memtest86 since then, and had another stick do 6 passes with 32 errors but in a different motherboard (2500/ga-h77n-wifi) and i've had several sticks do 4+ passes with no errors, so I'm happy to conclude that it's not a fault with the memory controller or DIMMS slot. All sticks are corsair 4GB running XMP 1600Mhz.

I did have one 2GB stick of corsair value select that seemed to freeze Memtest86 at 26% on both motherboards (3470 still in the asrock, 2500 in the H77). I say "freeze" because my cpu cooler sounds like the CPU then goes under load but everything on screen locks up, except the little red "+" symbol in the top left corner, which continued to flash. Even though this flashes and no actual errors are displayed on screen, I'm guessing that that stick can just go in the bin...

Is that unusual for memtest86 or just a genuine sign that the RAM is garbage?

I still have 6 sticks to test, so thanks very much for your input people, it's the first time I've had to run this software so it's always helpful to hear how it all works (ie, what else to check and what can be at fault)

As a side question, I'm going to pick some bits up today and I have some pieces laying about (I run 3 PC's myself and seem to fix everyone's PCs around me as well) - I was thinking of getting a G1620 and putting it in one of my spare gigabyte z77 boards and leaving it on the bench just to run memtest86 when I need it. The G1620 is only $45 up the road, or a G2030 is $60. Would they be sufficient to test any DDR3 RAM available, or will I be better off using higher bracket CPUs?

m
0
l

Best solution

a c 131 } Memory
November 20, 2013 6:32:24 AM

I've not had it lock up so I can't speak to that. But I will say I had a stick a while back that routinely threw 5~10 errors per pass I never RMAed the stick and it worked with maybe one blue screen every 2 months for 3 years (didn't test it till a few years after purchase) If its showing errors and a slight voltage bump doesn't make them go away, then I would definatly say RMA them.


as fo ryour side question If i rememgber correctly I think one of those processors can run 1600 mhz one cant. I would say for any real memory testing you would want a processor capable of at least 1600 mhz.
Share
November 22, 2013 4:16:07 PM

Supahos said:
I've not had it lock up so I can't speak to that. But I will say I had a stick a while back that routinely threw 5~10 errors per pass I never RMAed the stick and it worked with maybe one blue screen every 2 months for 3 years (didn't test it till a few years after purchase) If its showing errors and a slight voltage bump doesn't make them go away, then I would definatly say RMA them.


as fo ryour side question If i rememgber correctly I think one of those processors can run 1600 mhz one cant. I would say for any real memory testing you would want a processor capable of at least 1600 mhz.



Thank you kindly!

m
0
l
a b } Memory
November 22, 2013 7:08:13 PM

Can you down clock the memory to the SB i5's memory controller speed; 1333MHZ? Then try again with memtest.
m
0
l
November 22, 2013 7:22:05 PM

Aside, it wouldn't hurt to get a can of compressed air and blow out the dimm slots. Sounds like re-seating the dimm made the memory errors change/ go away.
m
0
l
a c 2198 } Memory
November 22, 2013 7:34:34 PM

What DRAM are you trying to run - the model #, the Z68s for the most part were left hanging, they came after the P67 and jsut before the Z77 and never really fully matured, most BIOS s are spotty when it comes to DRAM - if you give the DRAM model number might well be able to stabilize the sticks at spec freq
m
0
l
!