Elpida Hyper Graveyard

[Leeghoofd]

Lol, I'd never have thought of putting a fan over my memory if you'd not have said so!

C'mon man what do you take us for?

You would be surprised how many run their mobo/ram combo's without a extra fan... and the IOH really dumps a lot of heat if not actively cooled.

Maybe ask which dimm in which slot gave in ? The one of my Kingston kit was the one nearest the IOH. Even though there was an Antec Spotcool directed at the IOH. Ram voltage never was above 1.63 (bios setting of the Gene II), about 1.65 reality... I was just testing 1900mhz and was still working my way up to 2000mhz... RMA here we go... the Corsairs and OCZ's keep on working fab on this board so it MUST be a bad dimm... if I kill another one on in the same slot then it's the mobo...

[Raja@ASUS]

You would be surprised how many run their mobo/ram combo's without a extra fan... and the IOH really dumps a lot of heat if not actively cooled.

Maybe ask which dimm in which slot gave in ? The one of my Kingston kit was the one nearest the IOH. Even though there was an Antec Spotcool directed at the IOH. Ram voltage never was above 1.63 (bios setting of the Gene II), about 1.65 reality... I was just testing 1900mhz and was still working my way up to 2000mhz... RMA here we go... the Corsairs and OCZ's keep on working fab on this board so it MUST be a bad dimm... if I kill another one on in the same slot then it's the mobo...

That would imply that 5-6 motherboards from different vendors are all faulty or users are not savvy enough to cool their IOH (including some non X58's chipsets too). Personally, my IOH is cooled with a Delta and the ram is cooled with another fan over it. I've got a second set going south as we speak - one module does not always map to the OS even though when I got the kit brand new it mapped every single time (this non-mapping thing that was the mark of demise for my last kit). The troublesome module on the new kit was not in the first slot either, plus Im running an open test bed. Cast your votes as to what it is, my guess is we'll never find out the truth on this one so can only speculate.

Also strange how if it's the mobo the non Elpida stuff carries on working even when substantially over-volted on the same boards. Whichever way you look at it, it's quite an Elpida sensitive affair. The IMC setup process on i7 is largely automatic too (no documented drive parameters have been found for those wondering about those next) - not much any of the board vendors can do in that regard. The rest; the ony real stipulation is that the VTT line is fully up before Vdimm is applied. In terms of power regulation they're virtully the same Vdimm circuits we've had on some boards for the past 18 months, no major changes from companies like ASUS. One wonders why these 'faulty board' issues did not show up in the past.

[KILLER_K]

Just got off the phone with a Kingston tech rep, and he said they are replacing the "elpida" ics with another brand. I asked what brand he said he wasn't sure but since the large failure rate that decided to change them out.

He gave me a choice to wait or get a full refund for my ram. As if i wait they are out of my kits and he wasn't sure what date the new ones would be in. So you can take it from there.

[eva2000]

That would imply that 5-6 motherboards from different vendors are all faulty or users are not savvy enough to cool their IOH (including some non X58's chipsets too). Personally, my IOH is cooled with a Delta and the ram is cooled with another fan over it. I've got a second set going south as we speak - one module does not always map to the OS even though when I got the kit brand new it mapped every single time (this non-mapping thing that was the mark of demise for my last kit). The troublesome module on the new kit was not in the first slot either, plus Im running an open test bed. Cast your votes as to what it is, my guess is we'll never find out the truth on this one so can only speculate.

Also strange how if it's the mobo the non Elpida stuff carries on working even when substantially over-volted on the same boards. Whichever way you look at it, it's quite an Elpida sensitive affair. The IMC setup process on i7 is largely automatic too (no documented drive parameters have been found for those wondering about those next) - not much any of the board vendors can do in that regard. The rest; the ony real stipulation is that the VTT line is fully up before Vdimm is applied. In terms of power regulation they're virtully the same Vdimm circuits we've had on some boards for the past 18 months, no major changes from companies like ASUS. One wonders why these 'faulty board' issues did not show up in the past.

Have you or anyone, with failed elpida hyper module(s), ran the remaining okay modules until they failed ? Or remaining modules still strong ? I guess most folks won't know as they have to send back for RMA the entire kit of 3 modules even if one module fails.

[Star_Hunter]

So Kingston would likely switch to 50nm Samsung? Are there any modules with 50nm Samsung already out?

[KILLER_K]

Seems the middle number on the ram sticks, indicate what ic's are on the pcb's from what i gathered talking to the tech. I will call him back tomorrow and chat again. As if they are going to be some better ic's and not randomly die, i'm game then but if not i will take a refund.

[eva2000]

Seems the middle number on the ram sticks, indicate what ic's are on the pcb's from what i gathered talking to the tech. I will call him back tomorrow and chat again. As if they are going to be some better ic's and not randomly die, i'm game then but if not i will take a refund.

you mean the serial or module id # ?

[KILLER_K]

The module id the last part, as ie ask me 3 times to make sure and said he would go check.

[tripgood]

Have you or anyone, with failed elpida hyper module(s), ran the remaining okay modules until they failed ? Or remaining modules still strong ? I guess most folks won't know as they have to send back for RMA the entire kit of 3 modules even if one module fails.

I ran my remaining 2 sticks of DOM GT's and they both failed in 1 week at reduced settings.

[saaya]

Lol, I'd never have thought of putting a fan over my memory if you'd not have said so!
C'mon man what do you take us for?

There might be 1 or 2 people without some form of cooling over the modules, but I'd hedge bets most are cooling the modules actively.

well hes got a point, a few people who reported dead sticks were people with few posts who might not be experienced... but while this might affect the situation, it doesnt explain the problem for sure... especially the dozen or more DOAs people reported...

1.66Volts in bios, it's 1.74Volts REAL...

raja, can you confirm this? might be a good idea for evga to rename the voltage table in bios to adjust to the offset, can you tell sham?

You would be surprised how many run their mobo/ram combo's without a extra fan... and the IOH really dumps a lot of heat if not actively cooled.

Maybe ask which dimm in which slot gave in ? The one of my Kingston kit was the one nearest the IOH. Even though there was an Antec Spotcool directed at the IOH. Ram voltage never was above 1.63 (bios setting of the Gene II), about 1.65 reality... I was just testing 1900mhz and was still working my way up to 2000mhz... RMA here we go... the Corsairs and OCZ's keep on working fab on this board so it MUST be a bad dimm... if I kill another one on in the same slot then it's the mobo...

mhhhh not necessarily... im not sure but isnt the first stick holding the lowest address space? so its getting stressed more than the other sticks, just like core0 always gets stressed more and the top vga in an sli or xfire setup gets stressed more... i doubt its a board fault... like raja pointed out, other mem works perfectly fine... you do have a good point there, heat might def play a part in that many elpida hypers going belly up, but i doubt thats the main problem.

Just got off the phone with a Kingston tech rep, and he said they are replacing the "elpida" ics with another brand. I asked what brand he said he wasn't sure but since the large failure rate that decided to change them out.

He gave me a choice to wait or get a full refund for my ram. As if i wait they are out of my kits and he wasn't sure what date the new ones would be in. So you can take it from there.

thats very interesting... please keep us up to date on how the replacement memory does!

So Kingston would likely switch to 50nm Samsung? Are there any modules with 50nm Samsung already out?

samsung? oh thats the new super mem rumors have been going around about? or is it just an upgrade of the current samsung stuff and it does about the same as elpida hyper?

Seems the middle number on the ram sticks, indicate what ic's are on the pcb's from what i gathered talking to the tech. I will call him back tomorrow and chat again. As if they are going to be some better ic's and not randomly die, i'm game then but if not i will take a refund.

might be good stuff... the previous samsung chips were the king of the hill until elpida hyper arrived, if this is really new samsung stuff it has a high chance of being at least as good as elpida hyper..

I ran my remaining 2 sticks of DOM GT's and they both failed in 1 week at reduced settings.

so its definately a problem of some chip batches! thanks, thats very good info there

[Raja@ASUS]

raja, can you confirm this? might be a good idea for evga to rename the voltage table in bios to adjust to the offset, can you tell sham?

On my E760 board 1.65Vdimm set in BIOS gives 1.652V idle in the OS and 1.648 under a single 32m thread load (8 thread load droop will be a bit more than this). So I'm not sure why some people have an issue with high idle voltages.

later
Raja

[Leeghoofd]

mhhhh not necessarily... im not sure but isnt the first stick holding the lowest address space? so its getting stressed more than the other sticks, just like core0 always gets stressed more and the top vga in an sli or xfire setup gets stressed more... i doubt its a board fault... like raja pointed out, other mem works perfectly fine... you do have a good point there, heat might def play a part in that many elpida hypers going belly up, but i doubt thats the main problem.

I'm still working with the 2 leftover Kingston Dimms and compared to my OCZ blades they do NOT get warm at all. Maybe there's bad contact with the IC's and that massive blue heatspreader ? I'm puzzled here as I had several Hypers here and this is the first kit that caps out on a sturdy mobo in under 2 days. Hope this is the first and last kit that goes down in my hands...

With my post I meant : if I see several guys that used Hypers and that have 3 or more diffferent ram kits that all die, there must be something else wrong ; either the mobo, settings (maybe due to bios) and or other factors... For me : one kit dies okay it happens, 2 is already a bit weird, 3 or more : time to digg deeper...

Here's the info for the Kingston kit :

[Raja@ASUS]

With my post I meant : if I see several guys that used Hypers and that have 3 or more diffferent ram kits that all die, there must be something else wrong ; either the mobo, settings (maybe due to bios) and or other factors... For me : one kit dies okay it happens, 2 is already a bit weird, 3 or more : time to digg deeper...

Well if the only stuff dying frequently is Hyper, it's kinda hard to start looking elsewhere.

[Leeghoofd]

I'm talking about several end users that had issues with several kits on their mobo(s), not of the issue in general. If it happens with ya setup time after time, kit after kit, then it could be something else then just the ram. (soz for my bad english if you missundastood).

I see it like this : It can be a big batch of bad Hyper IC's. Eg as some of the OC'ers at the GOOC had issues running several ram kits even at their default rated speeds. Some might be badly binned (very positive about that looking at the low availability of high end kits : 2000C7, the 2000C8,C9 are so widely available), some have bad contact with the heatspreader (manufacturer error)...

Though do not forget : some mobo's really do overvolt big time when under load, some mobo's biosses have Ram speed presets that pump 1.8-2 volts when selected,... just to name a few issues...

Really think you are looking at it a bit too narrow, it's not always the ram, it might be mobo, user settings, bad usage etc... and ofcourse we see lots of complaints from users that ofcourse never pushed their rams... it was the same with the CPU degredation thing : one had it,then another and all of the sudden almost everyone else had it too (snowball effect), but it's never the user's fault...

There's an issue with these rams for sure, hope there will be some sort of solution for it, besides RMA'ing all the time... but we need to keep our perspective objective

[drnip]

On my E760 board 1.65Vdimm set in BIOS gives 1.652V idle in the OS and 1.648 under a single 32m thread load (8 thread load droop will be a bit more than this). So I'm not sure why some people have an issue with high idle voltages.

later
Raja

My 759 1.65v DIMM set in BIOS gives me 1.71v OS and 1.70v under load per multimeter reading. So for me to run at at 1.65v I need a setting of 1.60v in BIOS. Mine is a lil more on the extreme difference side as most I have seen about a difference of .03v or so.

Heats not an issue for me as I have 2 San Aces blowing directly across my mobo. 1 of them directly across the RAM. Just wish I would have kept my Redlines now.

[Star_Hunter]

samsung? oh thats the new super mem rumors have been going around about? or is it just an upgrade of the current samsung stuff and it does about the same as elpida hyper?

Well after looking into it I guess the Samsung chips have not changed and are still quite a bit below the hypers, I guess the question would be does anyone know of any ics that will rival the hypers anytime soon.

[G-UNIT91]

my friends, I have buy a new CSX DIABLO3 Hyper....EPIC fail?

[bluehaze]

With my post I meant : if I see several guys that used Hypers and that have 3 or more diffferent ram kits that all die, there must be something else wrong ; either the mobo, settings (maybe due to bios) and or other factors... For me : one kit dies okay it happens, 2 is already a bit weird, 3 or more : time to digg deeper...

Not at all mate, on my 3rd set of Elpida Hypers now and my regular Corsair non Elpida Hyper has worked since I first got it 3 or 4 months ago. I keep it so that I have ram to install when I have to RMA the Hypers because it's become such a common thing for the Hypers to fail. It should be obvious to you, having had bad Hyper ram yourself, that there is a problem with them. Even the manufacturers are admitting it now

[saaya]

On my E760 board 1.65Vdimm set in BIOS gives 1.652V idle in the OS and 1.648 under a single 32m thread load (8 thread load droop will be a bit more than this). So I'm not sure why some people have an issue with high idle voltages.

later
Raja

different bios maybe... different pwm config? maybe hipro and others have droop disabled or changed the pwm clockspeed?
or maybe its board variations? that would be bad tho :S

mhhhh how do you meassure it?
im sure hipro meassures it with a dmm directly in the mem slot... which is the most accurate reading and the voltage value that actually matters

what ground do you use? maybe hipro uses another ground point? ive seen even bigger diferences by using an... not ideal ground point

[bingo13]

Have you or anyone, with failed elpida hyper module(s), ran the remaining okay modules until they failed ? Or remaining modules still strong ? I guess most folks won't know as they have to send back for RMA the entire kit of 3 modules even if one module fails.

I have run one kit all the way down. I only replaced the bad module with a good module, eventually, they all died, never did more than 1.72 VDimm and 1.375V VTT either. Two modules failed at 1600 C8 1.5V VDimm and stock VTT. I have had failures on everything from the Classified to the MSI X58M uATX board. All I can say, is good luck with the warranty process as eventually something will give on these ICs.

[Raja@ASUS]

different bios maybe... different pwm config? maybe hipro and others have droop disabled or changed the pwm clockspeed?
or maybe its board variations? that would be bad tho :S

mhhhh how do you meassure it?
im sure hipro meassures it with a dmm directly in the mem slot... which is the most accurate reading and the voltage value that actually matters

what ground do you use? maybe hipro uses another ground point? ive seen even bigger diferences by using an... not ideal ground point

I use the ground point nearest to the measurement pad. You can take the measurement from the last cap or the slot is best - either way the slot will never be higher than the pad (resistive losses). Likely a revision of the boards - the E760 I have here is the latest I believe but no way of telling, it's one for Shamino to answer.

[Raja@ASUS]

Not at all mate, on my 3rd set of Elpida Hypers now and my regular Corsair non Elpida Hyper has worked since I first got it 3 or 4 months ago. I keep it so that I have ram to install when I have to RMA the Hypers because it's become such a common thing for the Hypers to fail. It should be obvious to you, having had bad Hyper ram yourself, that there is a problem with them. Even the manufacturers are admitting it now

Exactly my point BlueHaze, all else keeps on working but this stuff on the same hardware setup, so it's very hard in cases like ours to start deflecting the blame at other components.

[hipro5]

I JUST LOST one of my Corsair 2000C7 1GB stick.....
BLUE SCREEN during loading windows.....

I was OK on my latest benches and yesterday I booted to install new Vistas.....NO GO....

I then checked the rig with ONE - BY - ONE stick.....
ONE stick didn't booted at 1.65Volts and at 1333MHz.....

I used to work them with 1.72Volts with a DELTA 8 x 8cm fun over them...

[Raja@ASUS]

I JUST LOST one of my Corsair 2000C7 1GB stick.....
BLUE SCREEN during loading windows.....

I was OK on my latest benches and yesterday I booted to install new Vistas.....NO GO....

I then checked the rig with ONE - BY - ONE stick.....
ONE stick didn't booted at 1.65Volts and at 1333MHz.....

I used to work them with 1.72Volts with a DELTA 8 x 8cm fun over them...

Welcome to the unwelcome club George. I'm finding they last a bit longer if you don't boot them any higher than 1.69V real (I use higher once in the OS only). Although that is probably not the case for everyone.

regards
Raja

[bingo13]

I'm talking about several end users that had issues with several kits on their mobo(s), not of the issue in general. If it happens with ya setup time after time, kit after kit, then it could be something else then just the ram. (soz for my bad english if you missundastood).

I see it like this : It can be a big batch of bad Hyper IC's. Eg as some of the OC'ers at the GOOC had issues running several ram kits even at their default rated speeds. Some might be badly binned (very positive about that looking at the low availability of high end kits : 2000C7, the 2000C8,C9 are so widely available), some have bad contact with the heatspreader (manufacturer error)...

Though do not forget : some mobo's really do overvolt big time when under load, some mobo's biosses have Ram speed presets that pump 1.8-2 volts when selected,... just to name a few issues...

Really think you are looking at it a bit too narrow, it's not always the ram, it might be mobo, user settings, bad usage etc... and ofcourse we see lots of complaints from users that ofcourse never pushed their rams... it was the same with the CPU degredation thing : one had it,then another and all of the sudden almost everyone else had it too (snowball effect), but it's never the user's fault...

There's an issue with these rams for sure, hope there will be some sort of solution for it, besides RMA'ing all the time... but we need to keep our perspective objective

It is the RAM, more specifically the IC, why is the question we are trying to answer. I am sorry, but this argument about it being the board is nonsense. We have had failures on five different boards (ASUS, MSI, EVGA, Gigabyte, ASRock) at various voltage settings ranging from 1.50V to 1.80V VDimm and with VTT at stock 1.15V settings up to 1.375V.

It made no difference as to the supplier of memory either, OCZ, Corsair, GSkill, Kingston, Patriot, and now Mushkin. The argument about IOH temperatures being a factor is nonsense also. The memory has failed with IOH temps ranging from 47C to 79C and in various slots.

So, it simply comes down to the fact that either there is a problem with the IC (maybe PCB combo used) or a problem with the memory manufacturers recommending voltages higher than the 1.50V limit suggested by Elpida. On a side note, as far as the possibility of a bad batch, our failures have occurred with first lot parts all the way through to the latest lot.