Elpida Hyper Graveyard

[[XC] gomeler]

So more or less these chips are guaranteed to die even at "stock" speeds from memory kit manufacturers. So I guess the lesson to pull from this would be to purchase memory from people with solid warranties.

Micron FTW. (anyone killed D9GTR yet?)

I killed 4 sticks, they all fizzled out after lots of 2.2-2.4vDIMM runs. Surely didn't die like D9GMH/GKX did

[Brian y.]

1.) How did the memory die? Idle, load, reboot?
9 of them were with in the first week or so and had never seen over 1.7v. I don't know exactly what caused it every time but they were all only used for 32M testing and I would notice in my task manager the missing memory that would lead me to it. 6 of them were ones that had seen upwards of 1.78v and would crap out during a session. 6 more were DOA.
2.) What exactly happened? freeze, errors, crash, system turned off, system rebooted, blue screen, black screenMost of the time it was missing available memory but if I was installing an OS it would just blue screen the few times I realized it was bad.
3.) What board, pcb rev and bios version did you use?
EVGA 759 with whatever original BIOS that came with at the time. Same with RE2.
4.) What speed and timings did you run the memory at when it died?
Most of my testing was done @ 7-7-7-20 1T/7-7-6-18 1T @ 2000MHz usually.Every one that died, I was not sure of the exact time of death except I did have one set that lost a stick after a 1050 7-7-6-18 1T session of 32M testing @1.78v where about two hours in I noticed the missing memory in the task manager....

5.) what speed and timings did you run the memory at most of the time?
answered above
6.) What vtt did you run when the memory died?
anywhere from 1.4 to 1.6v
7.) what vtt did you run the memory at most of the time?same as above

All these sticks were Kingston Hyperx sticks and while I have lost quite a few as you can see I do have one 3x2GB kit that I have been benching up to 1.85v for many many hours over the past month. I have not seen any degradation of any kind. I have become quite superstitious with these IC's and am convinced that as long as I do not remove the sticks from the slots I will be ok. It may sound crazy but this crazy theory has worked for me so far

One last observation....it has always been every time for me on both boards used that it was the 1st channel that always went bad. EVERY TIME.

[Gautam]

Micron FTW. (anyone killed D9GTR yet?)

10+ gigs never more than 2.16v vdimm

I'll add my name to this elpida graveyard as soon I even think about running 32M. (yes, the thought is enough)

[Mescalamba]

Micron is tricky......but the most important factor is micron d9 gtr can not do these speeds in 2gig modules.

If your fancy is running tight like me 1720 6-6-5-15-11, your not gonna do so at reasonable volts.....not even after binning 3 sets of some of the best stuff out there....and they will degrade.

My set used to pull 6-5-5 but bsod'd while running 32m PI and never ran 6-5-5 again.

Lets not mention that I can get anyway with alot more voltage than most of the intel guys can for extended periods of time.....

Best bet is GTR/GTS 1g modules for 2d and elpida is really the only game in town for 2gig modules and 3d benching, if attemtping to run tight latencies at rather tight timings....

i7 is ok up to 1.8V .. maybe 1.9V .. which is enough for D9GTR/GTS. Just significant increase in VTT is needed and good cooling. If you have problem with some set.. try freezer, works quite good.

[Mescalamba]

So more or less these chips are guaranteed to die even at "stock" speeds from memory kit manufacturers. So I guess the lesson to pull from this would be to purchase memory from people with solid warranties.



I killed 4 sticks, they all fizzled out after lots of 2.2-2.4vDIMM runs. Surely didn't die like D9GMH/GKX did

Its DDR3 not DDR2.. impressive that they can live at 2V

[SAE]

Hey Sascha,

I just wanna contribute to this useful thread (not that I wanted those sticks to die in order to participate here ):

6GB OCZ Blade 2000 7-8-7-20 1T

1.) How did the memory die?

Been memtesting all the evening to determine minimum voltages for stability to avoid corrupting my win install. The next day turning the rig on it seemed there were no problems with the sticks at all... they went into windows fine. But then each bench I was starting failed immediately...

2.) What exactly happened?

Things got worse quickly... SPI threw errors, Prime totally quit itself, LinX BSODed. Then memtesting again there was only 4GB recognized. I removed all the sticks testing them individually then. There appeared to be a dead stick #3...

3.) What board, pcb rev and bios version did you use?

EVGA X58 Classified 759, Rev1.0, S21S.

4.) What speed and timings did you run the memory at when it died?

1025MHz 7-8-7-20-88-1T @ 1.67Vdimm

5.) what speed and timings did you run the memory at most of the time?

1002MHz 7-8-7-20-88-1T, 8xxMHz 6-6-5-18-59-1T

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

1.37Vtt iirc.

7.) what vtt did you run the memory at most of the time?

1.32-1.40Vtt.

Greetz,
Stefan.

[leotonch]

"Kingston T1 6G kit DDR3-2000 C8", I have RMAed 4 pack, 1 stick from the kit suddenly die repeated 4 times irregularly & 1 new pack with a new date lot yet test

1.) How did the memory die? Anytime, never known !
2.) What exactly happened? 2G ram disappear from system & tested it stick by stick in x48 to ceritify which one is suck !
3.) What board, pcb rev and bios version did you use? Asus P6T pcb 1.01 bios 0503
4.) What speed and timings did you run the memory at when it died? Some are running 1600 CL7 also die & some running at DDR-2000 CL8
5.) what speed and timings did you run the memory at most of the time? 1866 CL7
6.) What vtt did you run when the memory died? 1.375 ~ 1.6v never >1.6v
7.) what vtt did you run the memory at most of the time? 1.55v

[Brian y.]

I'll add my name to this elpida graveyard as soon I even think about running 32M. (yes, the thought is enough)

LOL

sigged!

[saaya]

I tried, put ADATA and Kingston into DICE and it didn't help.

damn...

[George_o/c]

OMG look at these dead sticks around here ... WTF is going on with these Elpida Hypers ? Aaaah, I've missed the old BH5 days ... Running 3.65 - 3.8V and everything worked perfectly, time after time after time.

[hiwa]

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

2.) What exactly happened? freeze, errors, crash, system turned off, system rebooted, blue screen, black screen
mobo shows just 4gb

3.) What board, pcb rev and bios version did you use?
classified E759 ,05/04/2009 BETA,

4.) What speed and timings did you run the memory at when it died?
2210 7-7-7-21

5.) what speed and timings did you run the memory at most of the time?
2240 7-7-7-21

6.) What vtt did you run when the memory died?
1.425

7.) what vtt did you run the memory at most of the time?
1.55

[Star_Hunter]

http://www.newegg.com/Product/Produc...82E16820226059

I was going to get some of the Mushkin Cl8 kit but now I am thinking of holding off because of this. Some of these chips are dieing so I guess the question is what percentage of these are dieing and should I hold off until they fix the issue?

[[XC] gomeler]

LOL! My memory just died between pulling it from one board and putting it into another board!

1.) How did the memory die? Idle, load, reboot? Removing sticks from 1 board to a 2nd board for mobo testing
2.) What exactly happened? Lockup posting memory error
3.) What board, pcb rev and bios version did you use? Asus M4A79T-DLX and Gigabyte 790FXT-UD5P
4.) What speed and timings did you run the memory at when it died? 1600-6-6-6-18-24-1T @ 1.6v
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? 1.6v
7.) what vtt did you run the memory at most of the time? 1.6v


Time to call up Corsair and see if I can get 2 replacement sticks just for when this surely happens again.

[drnip]

OCZ Blade Series 6GB DDR3 2000 OCZ3B2000LV6GK

1. Dunno why or how it happened.
2. ONE DAY of ownership and 1 of the 3 sticks shat itself. Was working fine the night I received them and didn't do anything too heavy on them. Went to work the next day, got home and BSOD's yet agian!!!
3. EVGA Classified
4. 2000 CL7 8-7-20 1T
5. 2000 CL7 8-7-20 1T to 2100 CL7 8-7-20 1T
6. 1.41v
7. 1.325v to 1.41v.

This RAM was brand spankin new and hadn't had time to play with it really. Maybe 1 hour of run time with them. Very disappointed in this $400 investment. The other two sticks are great though.

[Tony]

If the boards are killing the ram shouldn't the board manufacturers being paying for the ram RMA and not the memory manufacturers?

Hyper is not cheap to buy, if you RMA 1 kit thru death, get another and that kit dies also does that not point at the board as the issue?

Im not saying THIS is what should happen, I am asking a serious question here as I see memory dying that has not been massively overvolted and is within the specs laid down by the dram IC manufacturer.

[K404]

As a general comment on a related problem...are Asus willing to ackowledge that they overvolt by towards 0.1v on some boards compared to the BIOS label?

I dont think board makers will take responsbility for someone elses component dying.

[Brian y.]

All these sticks were Kingston Hyperx sticks and while I have lost quite a few as you can see I do have one 3x2GB kit that I have been benching up to 1.85v for many many hours over the past month. I have not seen any degradation of any kind.I have become quite superstitious with these IC's and am convinced that as long as I do not remove the sticks from the slots I will be ok. It may sound crazy but this crazy theory has worked for me so far

LOL! My memory just died between pulling it from one board and putting it into another board!

maybe I am not so crazy after all!

[eva2000]

Folks mentioning not pulling modules from boards - how are you guys removing modules ? Shutting off psu and power down board and wait a few seconds before removing modules, or just shutdown system with psu power still active to remove modules ?

OCZ Blade Series 6GB DDR3 2000 OCZ3B2000LV6GK

1. Dunno why or how it happened.
2. ONE DAY of ownership and 1 of the 3 sticks shat itself. Was working fine the night I received them and didn't do anything too heavy on them. Went to work the next day, got home and BSOD's yet agian!!!
3. EVGA Classified
4. 2000 CL7 8-7-20 1T
5. 2000 CL7 8-7-20 1T to 2100 CL7 8-7-20 1T
6. 1.41v
7. 1.325v to 1.41v.

This RAM was brand spankin new and hadn't had time to play with it really. Maybe 1 hour of run time with them. Very disappointed in this $400 investment. The other two sticks are great though.

ouch sorry to hear mate

If the boards are killing the ram shouldn't the board manufacturers being paying for the ram RMA and not the memory manufacturers?

Hyper is not cheap to buy, if you RMA 1 kit thru death, get another and that kit dies also does that not point at the board as the issue?

Im not saying THIS is what should happen, I am asking a serious question here as I see memory dying that has not been massively overvolted and is within the specs laid down by the dram IC manufacturer.

Maybe the problem is at elpida's end ? Very old elpida hyper ics 847/848 - week 47, 48, 2008 ones are clearly different to latest batches and batches in between have varied in performance/oc'ing.

I just received my RMA replacements for Kingston 2000C8 and these are way better than original modules (arrived DOA 1 module and other not hitting rated speeds) i received from just testing the first 2x 2GB modules, testing 3rd 2GB module soon

[drnip]

I always turn my PSU off at the back, wait a bit then pull whatever hdd I need to. So far I have my Kingston's in for RMA and submitted a ticket on my OCZ's today. My Mushkin Redlines I had for a week worked fine but they were only Elpida's not Elipda Hyper's. Of course they were only a week old too as my HyperX's lasted a month.

[nut311]

Nice Post Saaya

G.Skill F3-16000CL9T 6GPS

1.) How did the memory die?

After benching with stock voltage 1.65v and Stock speed without Fan the RAM has died...i touch heat sink is very Hot, i think temperature is make died

2.) What exactly happened?

Only detect 4GB, install XP SP3 no problem it all, and run Vista 64Bit....BSOD or Crash program

3.) What board, pcb rev and bios version did you use?
i used DFI Lanparty UT X58 T3eH8 and BIOS 0408

4.) What speed and timings did you run the memory at when it died?

2040 8-8-8-24 1.65v

5.) what speed and timings did you run the memory at most of the time?

2000 8-8-7-21 1.65v

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

1.47v

7.) what vtt did you run the memory at most of the time?

1.37v


Second run , i few minute later died one Corsair Dominator GT 2000C8

1.) How did the memory die?

Died with memtest 1.74v @2000 7-7-7-20, after passed 51x% OS Bluescreen

2.) What exactly happened?

2 Died Stick

3.) What board, pcb rev and bios version did you use?

Rampage 2

4.) What speed and timings did you run the memory at when it died?

2000 7-7-7-21

5.) what speed and timings did you run the memory at most of the time?

1900 7-8-7-21 1.65v

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

1.51V

7.) what vtt did you run the memory at most of the time?

1.35V

I think the temperature also influence what causes death, because it seems Vdimm using the default 1.2v is too hot and i 1.65-1.75v is too high in case Elpida hyper, look like scaling may be the same as the D9GTR / S at 2.3V ( remember Saaya D9GTR topic in first out D9G)

[Raja@ASUS]

I've had reports that at least one memory vendor has admitted the Hyper can be 'fragile'. Also of note, I know people that have older Hyper and it seems to stand up a little better than the newer stuff (they've RMA'd the newer sticks while the older ones are still working OK).

later
Raja

[drnip]

I wonder if VTT has to do with dying sticks and if so what are the safe volts to be ran at various speeds?

[Hazzan]

I kill two elpida hyper ...

1. G Skill 2000C7
2. OZC Blade 2000C7

I never give Vdimm higher 1.65 volt...

1.) How did the memory die?

daily use...

2.) What exactly happened?

I cant boot on Vista 64bit..but its ok on XP ..always BSOD on vista...and on bios just 4G with CPU Z normal 6G

3.) What board, pcb rev and bios version did you use?

DFI UT X58 and ASUS R2E

4.) What speed and timings did you run the memory at when it died?

by SPD 7.8.7.20 1.65 volt running 2000

5.) what speed and timings did you run the memory at most of the time?

1025 7.8.7.20

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

DFI 1.4 volt with R2E 1.45 volt

7.) what vtt did you run the memory at most of the time?

same...

[illuminatiASUS]

This seriously a concerting issues. It really is something that makes me unsure about investing into a higher speed memory kit. Really the saving grace has been at least these memory modules are supplied through companies we know have strong support and RMA policies. I am going to see if I can get any test reports on my end and see if can provide you guys with any additional information. Additionally has anyone had the chance to try the memory linked below or know what IC's are based on? I think it is also something we at ASUS have talked about in trying to work with memory vendors and testing these modules for a longer period of time to see if issues can be found before they are experienced. This does make wonder though if something is wrong or off with the binning process currently.

http://www.ocztechnology.com/product...triple_channel

[bluehaze]

If the boards are killing the ram shouldn't the board manufacturers being paying for the ram RMA and not the memory manufacturers?

Hyper is not cheap to buy, if you RMA 1 kit thru death, get another and that kit dies also does that not point at the board as the issue?

Im not saying THIS is what should happen, I am asking a serious question here as I see memory dying that has not been massively overvolted and is within the specs laid down by the dram IC manufacturer.

It would seem so, Evga Classified overvolts .4-.5 volts on vdimm. I wouldn't think this would be enough to kill ram with good cooling but if it does, good luck getting them to replace your ram for you, it's hard enough to get them to rma thier own mobos lately