Memtest86+ Guide/How To's

[RollingThunder]

You might try this with cd's but I don't know if it will work.

Shadow,

It works with CDs, whatever you choose to boot from. In my case I still use a floppy or a CD.

 

[dmo580]

5 years ago when I built my AMD system, I ran tests 5 & 8 on loop and checked for 24 hour stability.

Today I'm on an i7 with triple channel. Is it still test #5 & 8 or should I run all? Anyway I'm on the 24 hr burn-in, so I have some time for y'all to respond to let me know how to proceed with my OC tests.

Thanks in advance!

 

[g0dM@n]

5 years ago when I built my AMD system, I ran tests 5 & 8 on loop and checked for 24 hour stability.

Today I'm on an i7 with triple channel. Is it still test #5 & 8 or should I run all? Anyway I'm on the 24 hr burn-in, so I have some time for y'all to respond to let me know how to proceed with my OC tests.

Thanks in advance!

That's a very good question. If you don't get an answer here and you have time, this is what I would do:

• Run your memory slightly unstable - overclock or tighten a timing just a little, and boot to memtest
• Run memtest and see which test kicks back the most errors
• Record the test #s that kicked back the most errors, and test against overclocking

My guess it would still be #5 and #8. I think those were the two I used to test back in the UTT-BH5/BH5/TCCD days of DDR. Oh, and let's not forget my first great DDR sticks, the Micron -5B C chips.

 

[johan851]

That's a very good question. If you don't get an answer here and you have time, this is what I would do:

• Run your memory slightly unstable - overclock or tighten a timing just a little, and boot to memtest
• Run memtest and see which test kicks back the most errors
• Record the test #s that kicked back the most errors, and test against overclocking

My guess it would still be #5 and #8. I think those were the two I used to test back in the UTT-BH5/BH5/TCCD days of DDR. Oh, and let's not forget my first great DDR sticks, the Micron -5B C chips.

Ah yes, BH-6 and BH-5 and TCCD. Reminds of the nForce2 overclocking days. That was really fun.

Sounds like a solid testing methodology. I always found more errors with test #6, if I recall correctly. Still, if you're using Memtest for memory stability testing, it's a good idea to isolate a test or two to sort of "ballpark" an overclock. Once you think you have it dialed in, I'd still recommend running the whole suite of tests a couple of times to be sure.

 

[g0dM@n]

Ah yes, BH-6 and BH-5 and TCCD. Reminds of the nForce2 overclocking days. That was really fun.

Sounds like a solid testing methodology. I always found more errors with test #6, if I recall correctly. Still, if you're using Memtest for memory stability testing, it's a good idea to isolate a test or two to sort of "ballpark" an overclock. Once you think you have it dialed in, I'd still recommend running the whole suite of tests a couple of times to be sure.

I definitely agree... actually, with everything you said.

 

[dmo580]

Ah yes, BH-6 and BH-5 and TCCD. Reminds of the nForce2 overclocking days. That was really fun.

Sounds like a solid testing methodology. I always found more errors with test #6, if I recall correctly. Still, if you're using Memtest for memory stability testing, it's a good idea to isolate a test or two to sort of "ballpark" an overclock. Once you think you have it dialed in, I'd still recommend running the whole suite of tests a couple of times to be sure.

Those were the days I came from. My testing methods come from 2005 essentially. The last time I built a rig I was able to use my 3dfx Voodoo 3 (PCI) graphics card as a video card since I wanted it separate from my main system while testing. That was when I was moving AGP => PCI-e, but I didn't have a PCI-e video card yet (waiting for 7800GT).

Well this time I'm doing something similar where my main system and its 4850 need to run, but apparently the Voodoo 3 is too ancient for Windows 7... LOL.

Anyway, I took that advice and I booted my memory close to the limits. Maybe I'll get errors, maybe not, but when I return home today hopefully my computer isn't a steaming pile of ash. I'll see how it is.

A simple google search of Memtest #5 or #8 reveals forum posts from back when I built my comp. Doesn't seem like anyone mentions them anymore. Then again I don't mind looping all tests either. Just whatever helps me detect instability!

 

[RollingThunder]

Well I've been getting BSOD's (with different error messages) for weeks now, tried all I can think of and stuff people have suggested. The new trick for my pc is just to go to a blank black screen while idle.
Ran memtest after a BSOD a few weeks ago...showed a whole list of non-stop errors straight away. Turned pc off and ran scan again....no errors...tried different slot combos etc... Messed around with RAM timings a bit (Crucial Ballistix pc8500 2x1Gb) and settled for 1066mhz at the claimed 5-5-5-15...all seem ok-ish until today when the black screen and restart was forced, oh and a blue screen later just for good measure.

Anyway ran memtest again and a whole list of errors straight away, but yet again after a while when I run another scan, no errors. Unless.....and this is the part I don't understand...unless I select the option in memtest under 'Memory Sizing' - BIOS-All.
Then it goes nuts with errors again, but I'm not sure if it's valid since I've been testing 1Gb sticks at a time, yet it states testing of between 3840mb - 4096mb when this option is selected.

I've attached a picture so if anyone could shed some light it would be greatly appreciated.

Memtest86 from a boot disk isn't the final solution, it's only a preliminary test. There's more to be done:

The three steps I was taught and do without question:

1. Memtest86 from a boot disk (V2.11 and NOT the 3.5V- 4.0V)

2. If it passes, on to Windows and SuperPi 32 mb test.

3. If it passes SuperPi, on to Orthos or Prime95.

If it passes all three, you're probably in good shape to begin your overclocks or your overclock is stable.

There are many, many examples of passing Memtest and failing SuperPi. There are also many, many examples of passing Memtest86 and SuperPi and failing Orthos (Prime95). You're not done until it passes all three.

Never make your initial or object memory test in Windows. If it's bad, you can corrupt your OS and will need to reinstall it. That's what Memtest86 from a boot disk is for as a start.

 

[Edi-no-son]

um.... noob question and sorry about the bumpage.
I've got memtest onto the disc, booted it and it works fine.... but what order should I test?

Like, I've got 4 sticks of ram.
Can I just test it as is, or should I take it out and test it one by one?
And also I read that I should test it until it passes 7x?? (Gonna take ages if I hafta do each stick, then do a total test with all 4 sticks in there!!)

Also, whichever windows it runs doesn't matter right??
Reason is winXP works fine, win7 freezes on me... hence testing my ram.
But if it's testing only the memory, it shouldn't matter what windows the HDD has right??

 

[johan851]

Windows doesn't matter, since Memtest86+ boots to itself, not to an operating system.

You don't need to test each stick individually - it'll test all the RAM in your system by default. If you find errors, you may want to try each stick individually after that to figure out which one is causing problems.

You can just leave everything default - it'll start scanning when it starts up - and let it run overnight. If it's clean the next day (no big red errors) then your memory is alright.

 

[Edi-no-son]

um... so lost...
I tested it and it only ran it twice and that was it :S

Kept waiting and it stopped scanning and the timer just kept counting!!
Did I need to change anything for it to continue scanning??

 

[johan851]

You shouldn't need to. When you say it stopped scanning, do you mean the progress bars stopped?

 

[Edi-no-son]

Well the number of pass showed '1', and the 2nd pass just finished, and said "no error" :S

 

[johan851]

Weird. It should just run happily until the end of time.

So the individual progress bars at the top right are completely stopped? But Memtest86+ is still running? The red "+" in the logo will keep flashing if it's still going.

 

[Edi-no-son]

I'll test it again and confirm a bit later

 

[Edi-no-son]

Alright just ran it for 4.5hrs and it did 4 passes.
Sorry the movement was just so slow compared to at the beginning that I thought it had stop!!

GOnna try it for 7-8hrs now.... but after 4hrs, it didn't have any errors so should be safe, but heading to work so might as well leave the comp on to test haha

 

[GizmoGuy]

Timings Different?

Running 8-8-8-24 memory DDR3-1600 PC3-12800 one 4G Module

Bios is manually set to 8-8-8-24 CMD 1

Memtest86+ v4.20 shows it to be 774MHz DDR3-1549 CAS 6-6-6-20

Why the CAS diff??

So far hase passed 2 full test cycles with 0 errors.

Great thread... thanks!

 

[johan851]

That's odd. It could be that Memtest is reporting the SPD values hardcoded in the RAM and not the values that are actually set. Are the values Memtest shows consistent with the RAM's spec?

 

[RollingThunder]

That's odd. It could be that Memtest is reporting the SPD values hardcoded in the RAM and not the values that are actually set. Are the values Memtest shows consistent with the RAM's spec?

johan,

I've had that happen with Memtest V4.x. Memtest V2.1x doesn't do that (for me). I have no idea why.

 

[GizmoGuy]

That's odd. It could be that Memtest is reporting the SPD values hardcoded in the RAM and not the values that are actually set. Are the values Memtest shows consistent with the RAM's spec?

No, the RAM spec is 8-8-8-24 1.5v

Doesn't mean, then, that Memtest has identified the RAM as operating at different timings or anything. Thought it might be indicating a problem with the RAM.

 

[GizmoGuy]

Test ended (PC hung.. little problem I'm having), pass 3, 0 errors.

Swaped my 2nd module in, and got the same wierd settings indicated for the 2nd mod (mod is identical to the first mod).