Significance of a single memtest error??

[NewAMDUser]

Hi all --

I ran a memtest 2.01 session overnight on a new set of 2x2GB Corsair DDR2-800 C4 (4-4-4-12 2T). When I got up this morning, memtest was most of the way through pass 8, and showed that there had been a single error in pass 4. Should I worry about this or ignore it?

A bit more background: I ran the first test at DDR2-836, simply because the 2x1GB Corsair DDR2-800 C4 (same timings) I previously had in this machine was running at 836MHz (I have my X2 3800 oc'd to 2501 MHz, and I didn't bother to reset the bus to 2400 MHz to put the memory at 800MHz). Also, the 2x1GB ran at 1.9V; the 2x2GB wouldn't boot at 1.9V, so I bumped it to the 2.1V on the Corsair label before starting memtest.

After I found the one error this morning, I reset the machine to DDR2-800, and restarted memtest -- I'll see if any errors are present at when I get home this evening.

In the meantime -- what are your thoughts concerning the significance of the one reported error? My first reaction is that a single error means I can't run this memory at 836 MHz with these timings and voltage -- but that seems like a pretty low limit for decent memory.

Don't get me wrong -- if it passes at 800 at 4-4-4-12, I can't (and won't) complain -- it will have met Corsair's specs. I'm simply hoping for a bit better oc performance, as I'd like to have a bit of headroom available as I start experimenting with a new system (a Gigabyte GA-EX38-DQ6, an E8400 (if they ever reappear on the face of the planet!), Vista 64 and 8GB of Corsair DDR2-800 C4).

Last Q: Think it would be worth bumping voltage up to 2.15V? The memory is rated at 2.1V, and the last thing I want to do is cause damage by overvolting.

Thanks!
Mark

 

[leon2006]

Memory should be error-free.... Not a single bit error is acceptable.

One way or the other it going to create problems for you. It is a function of the location (Physical & logical Address) of the bits/bytes.

I have applications that use up all the available memory on my computer and it has no tolerance for bad bit.

Its your choice and decision...

 

[warezme]

If its Corsair, be afraid, very afraid....

I had two sets pf XMS2 800 4-4-4-12 modules fail on me (huge memtest errors and blue screen deaths in windows). I'm on a third set.

In your case it may just be the oc was a little to high and not a problem. I would NOT recommend over volting that memory. My memory was clocked to default 800 but manually set to 4-4-4-12 but the system auto volted it a little higher than 2.1, something like 2.2. I had to manually set it back to 2.1 in the voltage settings.

If they start to quirk out on you, get you some Patriot 800 4-4-4-12's, damn good memory and can handle voltage much better and a lot cheaper as well. I bought some to keep my system going between Corsair RMA's. They turned out to be faster, cheaper and more robust than any of the Corsair modules. I trust them completely. The fancy shmancy Corsair's no so much.

 

[endorphines]

i have 4gb of ddr2 800 overckocked to 880 (amd 6000+ overclocked to 3.3ghz from 3ghz) and let memtest run for 2 days while i was out of town. it did like 16 passes and found 2 different errors twice each in that time. it the system is rock solid. the only time it crashed it was a bug in a driver from nvidia that had a patch out for it already. when i say it's rock solid i mean it'll run prime 95 every time perfectly. if you've only got one error in multiple passes it's a weak spot in your ram that returns the wrong value 1 in a billion times. if you're unlucky enough to have windows using that paticulat bit of memory or an application it could cause a crash. in the last 6 months my machine has crashed only a handful of times and they seemed like normal windows stupidity, not a hardware problem.
bottom line is your memory's not perfect, but you're not likely to have major problems. relaxing your timings slightly and upping the voltave to it's rated maximum will likely aleviate any problems you could have experienced from that error.

 

[leon2006]

Memory must be error free. Not a single bit is acceptable....Thats what i paid for. I replace my memory module as soon as i confirmed that there is a bad location.

It is expected to perform to specification for the next 20 years. Thats how semiconductor are designed and manufactured. Reliability is guaranteed to 20 years when run to specification.

Not a single bit is acceptable. I expect mine to be perfect as my application use all the available memory in whatever computer i use.

You will get screwed by a single bad-bit. The infamous BSOD is one of those.

Memory companies don't issue errata or bug list as AMD/Intel do for CPU. You can check their websites won't find errata documents allowing for bad bit. DRAMs and SRAMS are expected to be error free when run to specifications.

Why settle for a bad or sub-standard product. DRAM modules have life time warranty. Unless you purchase a low quality brand.

Again its your choice and decision.

 

[jevon]

I've been fortunate enough to never have errors occur, but I think if I came across even a single error on a new stick of RAM I'd return it while I still could under warranty.

Since you're OCing the memory you would have to check with the manufacturer/vendor about returning it, or just say you got the error w/o mentioning the OC part I've never taken RAM past the recommended voltage and probably wouldn't try it either, but maybe someone else with experience can speak to that.

 

[Jim_L9]

It is acceptable to have a certain number of soft memeory errors in the PC industry. If you can't deal with "normal" soft errors you can use ECC memory. http://www.pcguide.com/ref/ram/err.htm

 

[leon2006]

Jim_l9

There is no categorical statement saying soft_errors are acceptable...

The document details memory errors and what cause it. The objective of the document is to present the risk of PC without ECC on its memory.

Even ECC modules are not 100% proof.

The bottom line is if you have a bad bit it will always be a bad bit.

The HW today are a lot better that soft errors are very minimal. It may happen on application but not on memory checker or diags.

Memory DIAGS/Checker failure is/are not soft error. The fact that it is repeatable it may potentially be a bad bit/bits/byte.

 

[RetiredChief]

If the memory only produces the error when overclocked and none at specifications - the memory is considered good by the company. it simply means you can not overclock it! And yes a single memory location can cause problems, the least is system BSOD's. If that location just happens to be in system and is a pointer to Bios call, or to HDD it can messy up bios (Rare) or corrupt a HDD.

Side note - with 4 1 Gig modules I had to bump up FSB voltage 0.1 V. Not sure if this would apply to 2 x 2 G. Also I have never had a bad stick of Corsair. Got a bad/incompadable 2 X 1 G patriot for lap top amont ago.

 

[bugmenot]

Question for you, if I run memtest and get an multiple errors within a minute. Then run memtest the next night with the same ram, and run error free for 12+ hours, whats up with that?

 

[antas]

I agree with RetiredChi ef: if it's running free of error on it's standard/stock specifications, then it will be considered as an error free.

I used XMS2 800 4-4-4-12 myself, and I found that it works perfectly fine at those settings (800MHz, 4-4-4-12). But if I increased the speed even just by a little bit, the system even won't POST. But if I loosen the timing (to 5-5-5-18), the memory will run fine.