How reliable is RAM?

[Mark R]

Just wondering really.

I've just now finished repairing my rig after the 3rd DIMM failure in 3 years.

I originally had 2x Geil 1GB DDR2-3200 DIMMs. One died after 2 weeks. Ended up chucking them away, after I couldn't get an RMA. They tested fine at the vendor (but on my 3 separate PCs they the same DIMM failed memtest with thesame error), and each time I send them back, it was a $30 'testing fee'.

I replaced them with some Crucial Ballistix DDR2-3200, matched pair (2x 1GB). One stick died after 6 months. Couldn't be bothered to RMA after the first catharsis.

Added a Corsair XMS DDR-3200 matched pair (2x1GB).

And now, at 3 years - the other crucial Ballistix DIMM has died.

Always run at stock, in a reasonably well ventilated case. Although the DIMMs recommended 2.2 V, I always ran them at their 'design' rating of 1.8V as that worked fine and they ran cooler than at their recommended 'overclocking' voltage (except for final testing, where they were given their maximum recommended voltage).

As an aside, I pulled the heat spreader off the latest ballistix module, and it pulled all the chips off the PCB (all the solder balls fractured - apart from a couple, which just ripped up the PCB traces). Guess the soldering wasn't all that great.

 

[shortylickens]

On my systems, extremely. I've never burned up a SIMM or DIMM, ever. But I always buy good name brand RAM and motherboards. I also dont overclock anything.
Once, long ago, I found an article that explained how leaving your memory alone and screwing with the FSB can still burn up your RAM. Of course that was back in the Slot A and Socket A days.

 

[Yellowbeard]

I'm curious now. What exactly is your question? The title is bit vague.

 

[jvroig]

They are generally very reliable, especially when left at stock.

I've had 5 machines in the past 4-5 years (desktops and laptops), and I've never had a problem with any of their RAM. The only faulty RAM module I've encountered was a SODIMM I bought to upgrade my first laptop, but it was defective right from the shop (so I learned when I got home), so I returned it a day later and they replaced it.

So far, all 5 machines are still working (haven't thrown out any of them), not all of them are in perfect condition, but all their RAM modules are still running perfectly fine.

 

[Phynaz]

I've only seem one bad DIMM in my life.

 

[TemjinGold]

Just wondering really.

I've just now finished repairing my rig after the 3rd DIMM failure in 3 years.

I originally had 2x Geil 1GB DDR2-3200 DIMMs. One died after 2 weeks. Ended up chucking them away, after I couldn't get an RMA. They tested fine at the vendor (but on my 3 separate PCs they the same DIMM failed memtest with thesame error), and each time I send them back, it was a $30 'testing fee'.

I replaced them with some Crucial Ballistix DDR2-3200, matched pair (2x 1GB). One stick died after 6 months. Couldn't be bothered to RMA after the first catharsis.

Added a Corsair XMS DDR-3200 matched pair (2x1GB).

And now, at 3 years - the other crucial Ballistix DIMM has died.

Always run at stock, in a reasonably well ventilated case. Although the DIMMs recommended 2.2 V, I always ran them at their 'design' rating of 1.8V as that worked fine and they ran cooler than at their recommended 'overclocking' voltage (except for final testing, where they were given their maximum recommended voltage).

As an aside, I pulled the heat spreader off the latest ballistix module, and it pulled all the chips off the PCB (all the solder balls fractured - apart from a couple, which just ripped up the PCB traces). Guess the soldering wasn't all that great.

I don't know about the GEIL but that 2.2v Ballistix was not exactly known for its reliability. I think you just happened to keep on picking shoddy ram.

 

[veri745]

It's really difficult to correlate a single person's experience with product reliability with the overall market averages.

Historically, I've had terrible luck with harddrives. I've probably had 5-6 of them go bad on me in as many years, including an SSD.

I've never had such luck with RAM, but one of my roommates in college certainly did. It seems like he couldn't order a RAM kit without having to RMA it at least once.

It seems odd to me, but just about everyone I've talked to about this has similar experience with one kind of product or another.

 

[Cerb]

Get some plain old RAM, rated at JEDEC timings and lower voltage (FI, 1.5V max for a DDR3 pair), with at least 80% 5 eggs, if what you want is reliable RAM. IoW, cheap A-DATA, G.Skill, PQi, or Corsair. WinRAR might run 5% slower (when was the last time anybody needed WinRAR, anyway?), but not overvolting the crap out of it might be worth that.

Here's a good DDR3 RAM search:
http://www.newegg.com/Product/Produc...G&PageSize=100

 

[TemjinGold]

OP is on DDR2, not DDR3.

 

[Cerb]

OP is on DDR2, not DDR3.

Fixed (<=1.8V DDR2 800-1200, 1x2, 1x4 ,2x2):
http://www.newegg.com/Product/Produc...D=3332167&amp;SID=

 

[SilentAssassin]

I'd have to agree with Cerb atleast on DDR2. When I built my cousins pc I used some more expensive Mushkin ram that was good for overclocking and already had higher voltage than standard and it died within a year. The replacement from Mushkin died even faster so I ordered him some G.Skill rated at 1.8v and there hasn't been a single problem since. That doesn't mean that your ram will always die faster if you overvolt but if you want the best chances of long lasting DDR2 RAM go with something rated for 1.8v.

 

[Elixer]

If RAM was really reliable, then they wouldn't make ECC RAM
Then again, if you get a one bit failure after 10 billion operations, is that reliable ?
Just how do you define reliable ? Mission critical ?

 

[DrMrLordX]

There's a difference between periodic errors and complete hardware failure.

RAM rated for high voltages isn't necessarily less reliable, Ballistix DDR3 notwithstanding. You can buy RAM rated for high voltages and high speeds and try running it at lower voltages and (if necessary) lower speeds, and it will probably last as long as (if not longer than) the non-enthusiast stuff.

I have some OCZ Platinum 2s or something on my x2 rig that is rated for DDR2-800 with a maximum warrantied voltage of 2.4v. Obviously I don't run it at 2.4v all the time, but it has served well for over two years at around 1.8v-2.0v. It helps that I run the stuff below its rated clockspeed with extremely tight timings.

 

[VirtualLarry]

I had some Crucial Ballistix Tracers fail on me in less than a month or two. At least, I think it's the RAM that's failing. I haven't fully diagnosed yet.

 

[Jeff7]

Most of the memory I've dealt with came with a lifetime warranty. But of course, if you're buying from a low-end or no-name vendor, good luck getting them to honor it, assuming you can even find contact info. The brands you mentioned though are better-known vendors though, so I'd expect them to honor warranties.


Concerning problems with RAM as described in the OP, my experience has been that could be compatibility issues, voltage instability, bad timings, or cheap RAM.

Compatibility - I've seen some motherboards which were extremely picky about the RAM that they'd run with. One was an nForce2 IGP motherboard from Biostar - it had a lot of problems with some of the cheap brands of memory that would require higher voltages than normal to run properly, and it was fussy about the timings. On that one, I could even run Memtest86 on the problematic RAM without any problems, but booting to Windows gave a BSOD every time.

Voltage instability - cheap motherboard or cheap power supply.

Timings - this can go with the compatibility issue. The default timings should work fine in most cases, but some motherboards have trouble.

Cheap RAM - some cheap stuff seems to need a voltage boost to get it to run right.

 

[VirtualLarry]

It could be the PSU too. It's a 300W (labeled), in a micro-ATX case, custom size. But the fan built-in to the PSU is weak, and the entire PSU heats up really badly. Scarily, actually.

I don't even have that much of a load on it. BE-2400 running at 2.875Ghz, 1.25v, two 1GB DDR2-1000, 2.1v, and a WD Green 1TB HD and a DVD burner.

 

[mikeford]

Just wondering really.

I've just now finished repairing my rig after the 3rd DIMM failure in 3 years.

Maybe it has nothing to do with the ram, and its a mother board issue? Some mother boards are fussy about which ram works in them, sockets could be flaky.

My latest mother board included an automatic over clocking function, just tell it to go in the bios setup and it would "try" various over clocking schemes. Seemed great until I ran the full memtest86, and then I got half a dozen consistent failures of the same single bit on just one of the more obscure tests. Which is still fail, so I set the timing to stock.

 

[Emulex]

sounds like bad mobo or power system. i've never had bad ecc ram (Failure). i've had DOA but that's different. maybe mishandling.

with the whole alpha ray or whatever ray flipping bits around. 8-12gb in a pc without ECC imo is dangerous.

i buy german cars because they are safe. i could buy a tin can toyota; but the extra protection let's me sleep better at night. same dealio with ECC.

 

[Cerb]

i buy german cars because they are safe. i could buy a tin can toyota; but the extra protection let's me sleep better at night. same dealio with ECC.

If only it was more commonly supported w/o buying server hardware (some/most AM3 boards work fine w/ it on most CPUs, though, don't they?)...

Though, did you check out Google's study? They found it was mostly in actual transfer, not cosmic rays; even so, newer methods (like Chipkill) reduce uncorrectable errors.

 

[fffblackmage]

I have a set of those 2.2V 2x1GB DDR2-800 Ballistix ram, and I'm still using them in a spare computer without any problems. Then again, I've been running them at the normal 1.8V instead of 2.2V for 99&#37; of its life so far.

If you want reliable ram, the best bet are probably RAM with JEDEC specified timings and voltage, not the crazy tight timings and high clockspeeds that require higher voltages.

 

[Rubycon]

DDR-2 Ballistix were famous for going out, no doubts about it.

IIRC alpha triggered bit flips were caused by local radiation. (after all alpha does not travel very far!) The lead in the solder had traces of radium in it which is incredibly radioactive and a source of alpha (and gamma) radiation. Pb has been banned from solder for a while now...