Elpida Hyper Graveyard

[saaya]

UPDATE:
Kingston, the worlds largest module maker/vendor and a very close partner of elpida has unofficially anounced that they will replace elpida hyper chips with another chip in the immediate future. faulty elpida hyper modules will be replaced with new modules based on another ddr3 chip (which one is unknown so far)

UPDATE2:
Corsair, one of the tier2 module maker/vendor has officially anounced that they are phasing out elpida hyper based modules and are recalling all elpida hyper based inventory from shops and distributors
http://www.xtremesystems.org/forums/...d.php?t=229050

UPDATE3: OCZ and Mushkin are putting elpida hyper based kits on hold as well, at least for now until more testing has been done

UPDATE4: the problems only affected a couple of batches of elpidas ddr3 chips, current batches and PSC chips (they licensed elpidas design) dont seem to have any reliability issues
UPDATE5: to make it more clear, PSC chips dont die at stock volts from what ive seen, they die very quickly with voltages above 1.9v though


thx to karmakazi for the pic

These kits tend to cost quite a lot, and even though they are rated to 1.65v and some are running them fine at above 2v, a lot of them seem to be dieing lately, even at as low as 1.5v vdimm...

If this happened to you with samsung based memory or any other memory chip or you are not sure what chips your sticks use, but they also died at low vdimm, please post your experience here as well, it might not be an elpida hyper problem after all...

please answer as many of the questions below, ideally all of them:
Just quote this post and then add your reply under each question

1.) After how much time did your memory fail? hours? weeks? months?
2.) Did you stress it a lot? (memtest/orthos/gaming)
3.) How did the memory die? Idle, load, reboot?
4.) What exactly happened? freeze, errors, crash, system turned off, system rebooted, blue screen, black screen
5.) What board, pcb rev and bios version did you use?
6.) What speed and timings did you run the memory at when it died?
7.) what speed and timings did you run the memory at most of the time?
6.) What vtt did you run when the memory died?
8.) what vtt did you run the memory at most of the time?

[Helfarch]

All stock settings on a Classified 760 and one stick died, DDR3 2000 C8s. No idea when it happened, it ran with the dead stick installed just showed 2GB ram, tested individualy and then with the dead stick out it went back to 4GB. Sent for RMA.

[saaya]

All stock settings on a Classified 760 and one stick died, DDR3 2000 C8s. No idea when it happened, it ran with the dead stick installed just showed 2GB ram, tested individualy and then with the dead stick out it went back to 4GB. Sent for RMA.

do all 3 memory slots work fine? its possible that a pin or two in the cpu socket is bent or has bad contact which leads to one of the memory channels to be dead/not operational.

what bios did you use and what vtt?
was the stick DOA or did it work fine at first?

[Kabauterman]

I think there is no Graveyard for the Hypers...

they dies like they would

[saaya]

I think there is no Graveyard for the Hypers...
they dies like they would

what do you mean?

[tekjunkie]

There seem to be at least half a dozen cases of dead elpida hyper based sticks that died nowhere near the known danger zone north of 2v but at a measly 1.65-1.75v... if you are one of the poor b4stards who had one of those sweet sticks roll over and die, please post as many details about it in this thread so we can find out whats causing it, and hopefully prevent it from happening in future

If this happened to you with samsung based memory or any other memory chip or you are not sure what chips your sticks use, but they also died at low vdimm, please post your experience here as well, it might not be an elpida hyper problem after all...

please answer as many of the questions below, ideally all of them:
Just quote this post and then add your reply under each question

1.) How did the memory die? Idle, load, reboot?
2.) What exactly happened? freeze, errors, crash, system turned off, system rebooted, blue screen, black screen
3.) What board, pcb rev and bios version did you use?
4.) What speed and timings did you run the memory at when it died?
5.) what speed and timings did you run the memory at most of the time?
6.) What vtt did you run when the memory died?
7.) what vtt did you run the memory at most of the time?

Kingston HyperX cl8 DDR3-2000(First Kit bought with Hypers)
Bought from Newegg(04/02/2009), was very happy to see them in stock as they were the only ones with Hypers in stock @ that time.The sticks came in with my classified(759), installed them and ran them everything defaults for about 2 weeks. Then as I found time to get into OCing, started to play around with my rig, encountered a ton of BSODs(pretty normal while OCing), decided to run Memtest @ 1.68 vDimm, 7-8-7 200 BCLK 2:10, went to bed after 3 successful passes, woke up in the morning to see a ton of memtest errors. The system then refused to post with all three sticks and found out that one of the sticks went bye bye(had to put in 1 stick @ a time in the same mem slot to find this out).

1.) How did the memory die?
During Overnight Memtest
2.) What exactly happened?
Please refer above
3.) What board, pcb rev and bios version did you use?
Classified 759
4.) What speed and timings did you run the memory at when it died? DDR3-2000, 7-8-7-20
5.) what speed and timings did you run the memory at most of the time?
Mostly @ default(DDR3-1333)
6.) What vtt did you run when the memory died?
Somewhere between 1.4 and 1.45(sorry can't remember)
7.) what vtt did you run the memory at most of the time?
Default(1.2)

I RMAed it to newegg and in the meantime saw the Dom GT 8-8-8 DDR3-2000 kits in stock and ordered them.

Corsair Dominator GT DDR3-2000 cl8(My second set of Hypers)
The Dom GT seemed to require lesser VTT than the HyperXs(1.375) and was happy about that, with being LinX stable @ 200x21 2:10 @ 1.65V 8-8-8.So I increased the volts to 1.68 again for 7-8-7, passes memtest and LinX, so did a Prime95 Blend run overnight, only to wake up and see a BSOD(I think it was MEMORY_MANAGEMENT BSOD).

1.) How did the memory die?
During Overnight Priming
2.) What exactly happened?
Please refer above
3.) What board, pcb rev and bios version did you use?
Classified 759
4.) What speed and timings did you run the memory at when it died? DDR3-2000, 7-8-7-20
5.) what speed and timings did you run the memory at most of the time?
DDR3-2000, 7-8-7-20
6.) What vtt did you run when the memory died?
1.375(+175mV)
7.) what vtt did you run the memory at most of the time?
1.375(+175mV)

So, since I was also waiting for My RMAed HyperX, I was running them fine @ DDR3-2000 7-8-7 in dual channel mode with two sticks for a while before my HyperX(third set of Hypers) came in and it came in as DOA.The PC was not even POSTing with the HyperXs on them @ DDR3-2000, although they did run fine @ DDR3-1333.

So sent the HyperXs and the GT back for a refund(Thanks Newegg) and also thought maybe the mobo had bad RAM slots and was killing the sticks and also RMAed my classified.In the meantime I got myself another board(Foxconn BR), another chip(W3520) and another set of RAM sticks(Patriot DDR3-2000 cl8) that had hypers on them.That was my fourth kit with Hyper on 'em.The patriots are the only ones until now that I never had any problem with(knock on wood).

I was also waiting for the DOM GT 7-8-7 kits to come in stock and was able to pull the trigger on them too to go with my classified.I also got a W3540 in the meantime.

Corsair Dominator GT DDR3-2000 cl7(My fifth kit with Hypers)
The RAM and my classified came in and I was running it fine with 200x21 with RAM 2:10 7-8-7 1.65V and alas after 2 days of mild activity(gaming and stuff), one of the stick degraded and it won't POST @ DDR3-2000 with that degarded stick and would work @ DDR3-1333 so I had to RMA 'em from Corsair.

1.) How did the memory die?
No death, just degraded.
2.) What exactly happened?
Please refer above
3.) What board, pcb rev and bios version did you use?
Classified 759
4.) What speed and timings did you run the memory at when it died? DDR3-2000, 7-8-7-20
5.) what speed and timings did you run the memory at most of the time?
DDR3-2000, 7-8-7-20
6.) What vtt did you run when the memory died?
1.375(+175mV)
7.) what vtt did you run the memory at most of the time?
1.375(+175mV)


Sorry about the long posting.

[tekjunkie]

I think there is no Graveyard for the Hypers...

they dies like they would

what do you mean?

I think what he means is that there is no specific pattern here(vDimm,VTT etc), they just sporadically die and with my experience I am with him on that.

[Helfarch]

do all 3 memory slots work fine? its possible that a pin or two in the cpu socket is bent or has bad contact which leads to one of the memory channels to be dead/not operational.

what bios did you use and what vtt?
was the stick DOA or did it work fine at first?

stock as in 0 settings changed with an i7 920, no attempt at OC yet as I had only just installed Win 7. I'll have to check the BIOS version, haven't updated that either.

GSkill ram is plugged into the same slots and working 100%.

[saaya]

very nice tek... but the last kit was on BR and not classified right?
or was it on the replacement classified board?

a friend just hinted towards me that some sticks even die with stock voltage and that elpida knows that there is a problem and working on it.
seems they have some problem with the chips... that might explain why they hesitated to launch the elpida hyper chips to begin with...

has anybody tried to revive his dead stick by putting it into the freezer?
this helps with most dead micron sticks, at least temporarily.... if that helps with the elpida chips as well then its most likely a silicon bonding/packaging problem...

[Kabauterman]

I think what he means is that there is no specific pattern here(vDimm,VTT etc), they just sporadically die and with my experience I am with him on that.

right...

my 2000C7 runs like hell, and after 1 week I don't used the System I want to boot and one stick was not longer detected... RMA...

also with the 2000C8 what I had before, they runs fine, but after a non specific point they don't want to run stable in Memory Based tests... RMA...

[tekjunkie]

very nice tek... but the last kit was on BR and not classified right?
or was it on the replacement classified board?

It was on the replacement Classified. The only sticks that I ever ran on the BR were the GTs (where one of 'em died on the Claasified)in dual channel mode and the Patriots.

a friend just hinted towards me that some sticks even die with stock voltage and that elpida knows that there is a problem and working on it.
seems they have some problem with the chips... that might explain why they hesitated to launch the elpida hyper chips to begin with...

Always thought that either they had problems with the yield(55nm) or the so called copper interconnects.Since both of them were new technologies, I had a hunch that, that might be the case. Thanks for posting this. I hope that the replacement GTs that I got from Corsair live longer and without degrading.

has anybody tried to revive his dead stick by putting it into the freezer?
this helps with most dead micron sticks, at least temporarily.... if that helps with the elpida chips as well then its most likely a silicon bonding/packaging problem...

Sorry, never thought about doing this.

[jagt]

Corsair Dominator GT 1866MHz 7-8-7-20

1.) How did the memory die?
During LinX testing with maximum memory size

2.) What exactly happened?
System froze. Rebooted and only 4GB was recognized. Tested each stick individually found one stick stone dead.

3.) What board, pcb rev and bios version did you use?
DFI LanParty T3eH8 bios 4/28

4.) What speed and timings did you run the memory at when it died?
2000MHz 7-7-7-20 1.66-1.67v

5.) what speed and timings did you run the memory at most of the time?
2000MHz 8-8-8-20 1.65v - 2000MHz 7-8-7-20 1.66v

6.) What vtt did you run when the memory died?
~1.32v

7.) what vtt did you run the memory at most of the time?
1.3-1.4v

Memory had only been in my system for around one week and had seen very little usage except a fair amount of stress testing (LinX, Prime blend, memtest86+) I had stability at 2000MHz 7-8-7 and was testing 7-7-7 when it happened.

This thread is a great idea!

[pinto]

Three kits dead here:

1 OCZ blade 2000C7, 1 Gkill PS 2000C7 and 1 Corsair GT 2000C7. Every time one stick no longer detect and no windows start or only 4gb are shown. All kits died on R2E, only the Gskill was test on Classified but it's not the board fault because this only one was dead on arrival. My use? between 875 C6 (Vdimm 1.58v VTT 1.3v) and 1100 C7 (never use voltage above 1.68v and VTT above 1.45v).
Really frustrating but every time RMA was really fast with Corsair (great corsair life time warranty); OCZ and Gkill came from a good french shop.

[saaya]

ok, thanks for all the feedback everybody!

we can already conclude that:
its not one board causing the problems
its not one brand/pcb design of mem that has the problems

just like kabauterman said, the kits seem to fail at random, though most of them seem to fail within the first few weeks, is that correct?
so if a kit has been running fine for a while already, it has a good chance of running fine for much longer if not forever...

tekjunkie, i suspect that its the chip bumps/packaging that is causing the problems...
the problems are very similar to nvidia chipset issues and micron d9gmh/gkx issues, only that temperatures and voltages dont seem to matter and chips seem to fail almost entirely at random...

remember, all it takes is a single bump/bond on the chip/package to fail and essentially the entire stick is dead. sure you can replace the chip and get the stick working again, but thats quite a hassle, not only for the end users but also the manufacturers. im very surprised about this cause elpida used to have the most reliable chips when i worked at cellshock... not the best performers, but they performed more than good enough and seemed to be almost indestructable... the rma rates of elpida based ddr2 were simply ridiculous...

[pinto]

My Gskill was dead on arrival but my Blade and GT kits died after 2 months 24/24 Saaya.

[emo]

I have the same problem... only one stick will die and another two will have no problem. System still boot and @BIOS only shows 4GB ram.. But inside CPUZ, you can still find 6GB ram and all 3 sticks SPD

[pinto]

+1 cpu-z still see 3 sticks/6gb and windows 4gb.

[Hazzan]

nice topic... Saaya..

I ll waiting this topic..and I will report later with my hyper chipset..

[eva2000]

Only tried 2 kits of Elpida Hyper - 1 out of 2 heh

kit 1: No problems at all with 6GB Corsair Dominator GT 2000 CAS7 which ran up to 1.96v vdimm with 1.65v max cpu vtt but usually run around 1.65-1.74v with 1.3-1.63v cpu vtt for several months now with no problems.

kit 2: Bought 2nd hand off these forums 6GB Kingston 2000 CAS8 - pretty much DOA/1 module failure out of the box - didn't know at the time as it was sometimes detecting 4GB instead of 6GB but recently tested each module on Asus Rampage Extreme and confirmed one module is a dud - full details at http://www.xtremesystems.org/forums/...d.php?t=226580

[bluehaze]

Dominator GT 1866 C7, 1st set had 1 stick DOA, Second set ran 2000mhz 7-7-7-20@1.68 volt for 2 weeks then computer crashed over night. Was just idling not priming. It wouldn't reboot had to take 1 stick out that was dead computer booted now, Then next time I tried to reboot same thing happened and another stick was dead.

This was with Classified 760 and 1.24v Vcore 1.4v VTT with 10x Mem, 20x Uncore, 200mhz bclk.

[Kabauterman]

I hope Elpida can fix this problems asap!

[fredyama]

I cannot remember the details of each incident but ONE stick is either half-dead (POST ok at bclock 133MHz only) or dead on Classified mobo @1.78 - 1.87Vdimm within a couple days.

2 kits of Corsair GT C7 3x2G (half dead): 1st set right after the first LN2 benching & 2nd set arrived half-dead

1 kit of OCZ Blade 3x2G (dead): Right after the first LN2 benching

2 kits of SanMax CL9 3x2G (half dead): during air testing for both kits

[George_o/c]

Quite interesting fredyama-san. Could you perhaps answer to more of the questions Sascha posted ?

1.) How did the memory die? Idle, load, reboot?
2.) What exactly happened? freeze, errors, crash, system turned off, system rebooted, blue screen, black screen
3.) What board, pcb rev and bios version did you use?
4.) What speed and timings did you run the memory at when it died?
5.) what speed and timings did you run the memory at most of the time?
6.) What vtt did you run when the memory died?
7.) what vtt did you run the memory at most of the time?

[TheKarmakazi]

I have a set of kingston 2000 c8's from the first batch newegg sold (when everyone thought it was a pricing error).

Set arrived "half dead" meaning 2 of 3 sticks worked perfectly (1050+ c7) but the third wouldnt even do 2000c7 (at normal voltages <1.7vdimm, <1.4vtt). After giving them 1.8vdimm/1.45vtt @ 2000c7 benching the "half dead" stick completely died.

Tested on REX48 (found the half dead stick), and classified (doing the benching that killed it).

Cpuz detects 6GB/3 sticks, but windows/etc only sees 4GB/2 sticks. Will be RMA'ing soon

[cdawall]

really good thread i have a kit on elpida non-hypers die on me no idea if those count oh and its AMD so my answers are off a smidge both of these were after about 3 days of benchmarking at these speeds

2X1GB elpida OEM 1066 CL7

1.) How did the memory die? Idle, load, reboot?
superpi32m

2.) What exactly happened? freeze, errors, crash, system turned off, system rebooted, blue screen, black screen
BSOD then no post over stock clocks

3.) What board, pcb rev and bios version did you use?
M4A78T-E BIOS 1401

4.) What speed and timings did you run the memory at when it died?
1800 7-7-7-21
5.) what speed and timings did you run the memory at most of the time?
1333 7-7-7-21
6.) What vtt did you run when the memory died?
on AMD....but the ram was @1.75v
7.) what vtt did you run the memory at most of the time?
1.5v

2X2GB elpida OEM 1066 CL7

1.) How did the memory die? Idle, load, reboot?
just died one day during a game

2.) What exactly happened? freeze, errors, crash, system turned off, system rebooted, blue screen, black screen
BSOD then no post on both sticks

3.) What board, pcb rev and bios version did you use?
M4A78T-E BIOS 1401

4.) What speed and timings did you run the memory at when it died?
1800 7-6-7-18
5.) what speed and timings did you run the memory at most of the time?
1800 7-6-7-18
6.) What vtt did you run when the memory died?
on AMD....but the ram was @1.8v
7.) what vtt did you run the memory at most of the time?
1.8v