How long should I run memtest?

[crazlunatic]

I have 4*2GB Patriot DDR2 RAM. Lately my computer has been blue screening and freezing up constantly. I ran a memtest last night on all 4 sticks and indeed there is an error.

Now I am testing each stick one by one to find the culprit. How long should I run memtest on each stick? Running memtest on 4 sticks is pretty time consuming so I want to know how many passes I should have before moving on to the next stick.

Thanks a lot!

 

[Idontcare]

I went thru this process myself yesterday...in my case running memtest+ v4.10 on one 2GB stick at a time for ~8-10min was enough to identify the failing stick. (test #4 in my case was catching the error)

If your particular error is not being caught until a later test then obviously it will take longer for you to have confidence any given stick is ok.

I replaced the one bad stick and retested my 4x2GB overnight (completed 4 runs in about 8hrs) and had zero errors. Best of luck, I hope it turns out as easy for you too.

 

[ch33zw1z]

Time saving: IMO...12 Hours each if the dimms are acting flaky, ie failing sometimes but not always. my rule of thumb: If purchased in pairs...test in pairs and RMA in pairs.

 

[crazlunatic]

I bought 2 pairs and I think I've mixed up the order by now =(
@Idon'tcare - that's a good idea. I'll try short tests for each stick first and then if nothing shows up I'll go for longer tests. Thanks.

--

But does anyone know a safe general amount of tests/passes to run? I'm on my 1st stick now. 1h30m passed and 3 passes.

 

[n7]

Bad RAM should usually error within a pass or two.

I run Memtest86+ overnight for thorough testing, but i'd just be doing a couple passes at most when sorting bad from good RAM.

If it's not failing when you test the DIMMs separately, it might be a mobo thing (more MCH voltage/vDIMM/etc. needed) or rarely, but possibly, a bad DIMM slot.

 

[Yellowbeard]

I have 4*2GB Patriot DDR2 RAM. Lately my computer has been blue screening and freezing up constantly. I ran a memtest last night on all 4 sticks and indeed there is an error.

Now I am testing each stick one by one to find the culprit. How long should I run memtest on each stick? Running memtest on 4 sticks is pretty time consuming so I want to know how many passes I should have before moving on to the next stick.

Thanks a lot!

What motherboard are you running?

 

[crazlunatic]

Gigabyte P35 DS3R

 

[Idontcare]

But does anyone know a safe general amount of tests/passes to run? I'm on my 1st stick now. 1h30m passed and 3 passes.

The answer to that question depends on another question - are you trying to detect hard fails (bits that simply don't work) or is your ram issue of the more insidious nature aka "a walking wounded" bit that works sometimes and other times chooses not to for a variety of reasons? (usually heat related, but can be vdimm fluctuation related)

The hard fail bits will be detected within a single pass. No need to run 3x the passes just to detect the same dead bits three times over.

The walking wounded are the problem. You can increase their fail rate, and thus improve the liklihood of detecting them (i.e. reduce the time for detection) by slightly lowering your Vdimm and/or putting them in a warmer ambient while testing.

The "warmer ambient" is basically what people are achieving by running memtest+ in a loop...letting the case internals warm up to a steady-state that then increases the odds of memtest+ catching a walking wounded dead bit.

In my case, the 8-10 min quick check, was catching a true hard fail. Then I tested for 8hrs (4 loops) to see if I could catch any walking wounded. You can never rule walking wounded bits, only can decrease the statistical liklihood of your ram having walked wounded bits by increasing the observation time (more passes in memtest).

 

[Hacp]

There is a way to reduce your memtesting. Do two sticks at a time. That way, you will need at most 3 tests to find the culprit.

 

[Knavish]

One warning: my 785G board (before a bios patch) was bluescreening & failing in memtest with 2 sticks, but ran perfectly with either stick by itself. After the bios patch, it was ok with both.

 

[RebateMonger]

One warning: my 785G board (before a bios patch) was bluescreening & failing in memtest with 2 sticks, but ran perfectly with either stick by itself. After the bios patch, it was ok with both.

Yeah, I've seen that, too. I recommend starting testing with the full compliment of RAM. If you get failures, you can test individual modules and, hopefully, determine if there's a single culprit. But in the end, you'll want to do another final test with all the memory installed.

 

[kalrith]

A week should be sufficient.

 

[crazlunatic]

Thanks for all the help. Here's an update. Ran 3 sticks for at least 2.5 hours and no errors. I'm on my 4th stick, just 1 hour in and no errors.

Any comments, suggestions or advice?

By the way, when I first bought this RAM, there were 2 sticks that were DOA (not exactly, they worked but had errors on arrival). I changed them and got working ones. Didn't bother to memtest the new ones. Not sure if telling you guys this does any good.
Thanks guys.

 

[Idontcare]

sorry, no ideas at this point being the usual "swap out, repeat test, swap out, repeat test" iterations. good luck escaping the "for...while" loop there

 

[crazlunatic]

I made a mistake. Actually my mobo is Asus P5KVM

 

[Elixer]

There is a way to reduce your memtesting. Do two sticks at a time. That way, you will need at most 3 tests to find the culprit.

I know on some of my older motherboards, this will NOT find RAM errors. When all the slots were populated, then and only then, did I see RAM errors.

Sometimes, that is the only way to test, and I leave it running 12-14 hours, up to 24hours in some cases.

 

[Idontcare]

I know on some of my older motherboards, this will NOT find RAM errors. When all the slots were populated, then and only then, did I see RAM errors.

Sometimes, that is the only way to test, and I leave it running 12-14 hours, up to 24hours in some cases.

That observation of a dimm slot population loading cause-and-effect certainly begs the question whether the observed ram instability itself is a symptom rather than the disease.

If my ram were throwing memtest+ errors but only when my dimms slots were fully populated I would have to assume the weakness lies somewhere on the mobo...either undervolted Vdimm or undervolted memory controller or simply a really crappy signal/noise mobo.

I'd be tempted to start boosting voltages and adding some fans to cool down components before just assuming my ram was bad in such a situation.

 

[crazlunatic]

anyone know how I can adjust ram voltage from my BIOS? My board isn't exactly an enthusiast board. It even has an integrated vid card. I only have this board so I could fit it in my lian li p351 case