Memtest86 a poor indicator of memory stability?

Soldato
Joined
2 Dec 2005
Posts
5,514
Location
Herts
Hi all,

Opty 148, 2x512 kingston value ram, dfi lanparty expert
I was just playing with the crappy ram mentioned above. With the htt at 255, I thought I'd see if it was stable at a 9/10 divider (yeh I know, unlikely conisdering its value pc3200). Anyway, booted memtest86 from cd and left it for an hour with no errors. Loaded up windows and it wasnt stable (flickering screen, failed superpi and p95 etc). Tried looser timings and more volts, memtest still thought it was stable but windows clearly wasnt.

Wierd that memtest, which is meant to be a reliable indicator of stability didnt detect any errors. Any thoughts?
 
It can only be the CPU. memtest86+ is one of the best testers about. It's either the memory controller or the CPU.
 
aiii MemTest86+ iseven used by the pro's (Crucial Labs etc). The obvious solution is to keep overclock (memory) and drop CPU multi to have it at stock speeds, see what happens then.
 
Memtest is a reasnoable tester but it is no substitute for prime etc. I still use it every time I overclock because by passing memtest it is highly unlikely you will corrupt your OS etc. But passing memtest is no guarentee that your RAM is stable, only the use of a variety of torture tests (p95 blend etc) can do that.
 
Agree, no-one torture testing app can be used to test for 100% stability. A mixture of memtest, prime95, superpi, 3dmarks, sandra etc are usually the most commonly used.
 
What about S&M as an good indication of stability?

I've had 2x1GB sticks of Ballistix PC4000 pass over 10 passes of test 5 & 10 passes of test 8 in memtest86, pass 24 hours Prime 95, yet fail the memory test in S&M.
 
Memtets needs to be run for a good few hours on test 5 or 8 then ill only consider it ok.,

Ive even had it pass around 20 times then fail. A few hrs at least IMO. ;)
 
Oh yeah i agree that memtest isnt the final ok for stability as i can pass memtest yet still fail prime.

The same with Super PI 32Mb i can pass that with my CPU yet still fail prime in seconds. I use them all and test with each and have Prime 12Hr+ as my final OC.

Memtest is fine to get a idea before running windows and risk corruption.
 
4Qman said:
Oh yeah i agree that memtest isnt the final ok for stability as i can pass memtest yet still fail prime.
I can fail Memtest and pass Prime95 *blend* :)

Will have to check out this S&M though, gimme a link pls Google is giving me whips and stuff :p
 
Failing memory testers when memtest has passed is almost certainly a memory controller error and not the RAM itself.
 
YooEntSinMeROYT said:
Can that be remedied by giving the CPU a bit more voltage?
It can be. Depends on the strength of the controller but usually a bit more vcore sorts it.
 
Isteresting guys, thanks
smids said:
It can be. Depends on the strength of the controller but usually a bit more vcore sorts it.
Btw, what is a sensible vcore for a 148 with storm waterblock and 120.2? I thought about 1.6 was the max for good water
 
Back
Top Bottom