How long to run memtest86?

orangat

Golden Member
Jun 7, 2004
1,579
0
0
How long or how many passes do people normally run memtest86 to ensure stability?

Also is there a windows utility which works as good as memtest86 since I'd prefer my dloads to be running.

Thanks
 

gamerj

Member
Dec 18, 2004
169
0
0
How long do you usually use your computer?

Such tests i use for an hour or so...i think the chances of errors decrease the longer the test is running...

Some people say 24 hours....i mean whatever u want...No errors after 30 min? ... again...chances of errors decrease when time expires...
 

orangat

Golden Member
Jun 7, 2004
1,579
0
0
I usually have a daily reboot but sometimes leave it running overnight. I could still ctrl-alt-del to kill the unstable progs, scanned my system and everythings clean.

Has anyone tried memtest by hcidesign? Its a windows prog.
 

Xed

Golden Member
Nov 15, 2003
1,452
0
71
I run it overnight while I sleep, if you don't get any errors by then it's a pretty safe bet your memory is fine
 

Steve

Lifer
May 2, 2004
15,945
11
81
I run through all nine tests once, and then I'm satisfied. Testing while in Windows can't be reliable as in a DOS-type environment due to side effects of things constantly running while Windows is loaded.
 

chinkgai

Diamond Member
Apr 4, 2001
3,904
0
71
memtest does not ensure stability, only prime 95 running for 24 hours or so will

i've had memtest run forever then fail prime after 30 min

i've had prime run for 17 hours before failing (sometimes 8 hours sometimes 11)

after i run prime for 24 hours, im still not done

then i loop 3dmark, some people choose 3dmark01, i choose 03 just cuz its more strenuous on my system than 01 in my opinion. i usually do this for a few hours before i stop it if there are no freezes or such

only then will i consider my computer prime stable
 

Navid

Diamond Member
Jul 26, 2004
5,053
0
0
If you fail prime but not memtest, it means that your CPU is the bottleneck. If you fail memtest, it means that your memory is unstable. Your CPU may be too!

But, this varies from system to system.
Just because you passed memtest and not prime95 does not mean that every system is going to be like that.

It is usually advised that you run memtest first overnight. If it passes, you then run prime95 overnight. The reason is that if you only run prime95 and fail, you won't know if it is your RAM or CPU.

You also need to check your PSU. That should be done first.

Your graphics card should be checked last.

You only move from one step to the next if what you have tested is stable.
If not, you have to fix it first.
 

Duvie

Elite Member
Feb 5, 2001
16,215
0
71
Originally posted by: sm8000
I run through all nine tests once, and then I'm satisfied. Testing while in Windows can't be reliable as in a DOS-type environment due to side effects of things constantly running while Windows is loaded.


1 pass or less then 30 minutes doesn't mean crap!!! That doesn't stress anything and so in return you get no reasonable assertion of stability........I see many BSOD in your near future if you are an ocer....
 

chinkgai

Diamond Member
Apr 4, 2001
3,904
0
71
Originally posted by: Navid
If you fail prime but not memtest, it means that your CPU is the bottleneck. If you fail memtest, it means that your memory is unstable. Your CPU may be too!

yes, in theory this is how it should work but in practice, it is often not the case.

when trying to find the max of my current memory, which is memtest stable overnight at 255 fsb for close to 24 hours before i stopped it, it failed prime after 30 min as i said earlier

one would assume that if it passes memtest and fails prime, then it would be the cpu

well not the case for me, on more than one occasion also, so this isnt an isolated case

my cpu is very very capable of running 255 fsb since i run it at 275 fsb all the time or even up to 300 at different settings, so it is NOT the bottleneck!

the memory i later found after various prime testing is only 250 fsb stable! which was found with prime, not memtest. memtest is a good starting point though, but i can almost surely say that if you are looking for the ceiling and find it with memtest, it will not be stable in prime even if ur cpu is capable. it will be close though from my experiences.

so my point is...memtest is only useful to find a good starting point, where almost always, you would have to work downwards to make it prime stable
 

Steve

Lifer
May 2, 2004
15,945
11
81
Originally posted by: Duvie
Originally posted by: sm8000
I run through all nine tests once, and then I'm satisfied. Testing while in Windows can't be reliable as in a DOS-type environment due to side effects of things constantly running while Windows is loaded.


1 pass or less then 30 minutes doesn't mean crap!!! That doesn't stress anything and so in return you get no reasonable assertion of stability........I see many BSOD in your near future if you are an ocer....

Personally I'm not an overclocker (or even a gamer), plus at work, the one pass of all nine tests is all we have time for in a corporate environment too. I wasn't aware that multiple consecutive passes of memtest would stress the RAM, and I'm not clear on what the benefit of that is?
 

chinkgai

Diamond Member
Apr 4, 2001
3,904
0
71
multiple consecutive passes is more beneficial because heat becomes a factor after longer periods of time

edit: where do you work?
 

Duvie

Elite Member
Feb 5, 2001
16,215
0
71
memtest does not ensure stability, only prime 95 running for 24 hours or so will

Was pretty much right except this statement....

On the contrary running prime95 alone does not insure stability (I think ensure is the drink).....prime95 does not effectively stress the memory, and if you use the blend test while it test the ram the most it is not enough and stresses the cpu and cache the least for which you should be tesing with prime95....The correct test 1 or 2 stress ram very little and thus calls for other means. Test 1 or 2 however are good for the must power consumption, and heat producing stress of them all....

I use prime95 01 because it is more cpu and system dependent then any of the 3. Now if you are ocing your Vid card 03 and 05 are far better to do that with....
 

Duvie

Elite Member
Feb 5, 2001
16,215
0
71
Originally posted by: sm8000
Originally posted by: Duvie
Originally posted by: sm8000
I run through all nine tests once, and then I'm satisfied. Testing while in Windows can't be reliable as in a DOS-type environment due to side effects of things constantly running while Windows is loaded.


1 pass or less then 30 minutes doesn't mean crap!!! That doesn't stress anything and so in return you get no reasonable assertion of stability........I see many BSOD in your near future if you are an ocer....

Personally I'm not an overclocker (or even a gamer), plus at work, the one pass of all nine tests is all we have time for in a corporate environment too. I wasn't aware that multiple consecutive passes of memtest would stress the RAM, and I'm not clear on what the benefit of that is?

For amd64 systems looping test 5 or 6 often times is more beneficial and usually where errors occur in the A64 chips....1 pass is to short to see if errors will occur. I run 10 passes of all test and then run 100 passes of 5 and 6 (which doesn't take long)....

1 pass of memtest means nothing. I can error on the second [ass frequently and thus would be one of those mysterious freezes, reboots or BSOD that may occur....


Edit: when I had a bad stick of GSkill prime95 would run 24 hours without error and memtest would pass 8-10 passes fine....However loopping test 5 found 6-8 errors every 20th pass and this was causes an occasional crash. Since I occasionally get on a kick of FH this is not exceptable since I dont want early end units or to have incorrect info given back...
 

Duvie

Elite Member
Feb 5, 2001
16,215
0
71
Originally posted by: chinkgai
you missed the part where i said i loop 3dmark too ;)


So!!! It is still not the most effective test of ram and like I said 03 doesn't stress the cpu and system as much as 01....03 is going to offload some of that stress to the gpu....

I like stressing gpu at sametime as stressing cpu to help build that case heat up, but I am telling you it is not the most optimal for finding issues with ram....

memtest is to ram what prime is to cpu....

in actuallity with AMD cpus more so then P4's superpi 32mb seems to be far better at routing out memory issues then prime95
 

Navid

Diamond Member
Jul 26, 2004
5,053
0
0
Originally posted by: sm8000
Personally I'm not an overclocker (or even a gamer), plus at work, the one pass of all nine tests is all we have time for in a corporate environment too. I wasn't aware that multiple consecutive passes of memtest would stress the RAM, and I'm not clear on what the benefit of that is?

It's all about stability margin.
It takes a system a few cycles before its temperature stabilizes.

When you overclock, you need margin to ensure that your system will not crash on a warm day or when you run something that is CPU or memory intensive.

If you do not overclock, you still need stability margin. But, you rely on the manufacturer for having qualified the products (memory and or the CPU) to operate with enough margin. If you get a decent memory, that is a reasonable assumption at rated speeds and timings (no overclock).

It is not a good idea to overclock for business anyway.
 

Steve

Lifer
May 2, 2004
15,945
11
81
We'd never overclock any work systems. Honestly, it just plain takes too long to repeatedly run memtest at work. That user needs a PC pronto! At work, I only ever run it in a case where the symptoms point to defective RAM.