Printable View
I have the same opinion about UTT ram.
They work fine (initially) ,but fairly large percentage of IC's die over a longer period. Good testing processes weeds out these "weak" IC's...
Clearly , this isn't happening and that's the real issue , IMO.
Also explains all the posts about failing UTT ram...
You don't hear about old Winbond BH-5/BH-6/CH-5 sticks failing , do you ?
There are some case reported , but less than 5 if I remember correctly ... Most of the dead sticks are all UTT sticks ... Some from OCZ , some from mushkin and others ... They die in various kind of voltage ... Lots of the case are reported in the DFI board with dram voltage > 3.0v , jumper in 3.3V or 5V ... Also there are case reported in other board + booster or vmod ... There are also case reported using the old NF2 board with high memory voltage ... If it's the voltage or the board kill these UTT stick , we should only see a common point of these case ... But the results now are way too much difference between them ... These stick are dead in various kind of platform , voltage and different chipset ... It's not possible that it's the voltage or platform or the board to kill these dram ... And what I found told me that this kind of situation is a typical problem of UTT chip maded module ... In the first , I think UTT is the recent thing ... By doing some research , I realize that these UTT chip are everywhere in the market ... But in the past , this type of module is so called downgrade module ... Just check my thread about the detail information ... http://www.xtremesystems.org/forums/...ad.php?t=65478Quote:
Originally Posted by Jaco
Was a cold boot indeed, took me about half an hour to sort it in the end.......so is this a board problem then, im assuming it is.
What ? OHG ... Let me explain once again ... It's not possible that the problem is the board !!! I think you must be able to use this dram to work under 2.6V to boot for daily usage when you first receive the dram ... But after some weeks or so , this dram start to degrade in it's electric characteristic and require more and more voltage to boot up ... This is not good !!! To be more precisely , we can call these chip as " Faulty ones " , because these chips are design to work under 2.6V ... If it's the board problem , then your dram should be able to boot in other platform ... If it's the board kill the dram , then we shall not see any other dead UTT case in any other platform ... You know what ? By searching around the internet , I found that almost every DDR I RAM is spec @ 3.6V absolutely maximum rating !!! If there is any dead case in DFI NF4 is based on 3.3V jumper location , it's not possible for the board to supply dram voltage any higher than 3.3V ... But now we got many reports here that the dram is dead under 3.2V or so when he only set the jumper @ 3.3V location , no matter what kind of situation happened , it's not possible for the dram voltage exceed the power source --> 3.3V ... Voltage does not kill these dram !!! The problem is UTT itself , ok ????Quote:
Originally Posted by Vincentvega18
honestly there is not enough information anywhere to make a real decision that is fact. However, ppl can post their situation and other ppl can come to their own conclusions, confusing yes, but at least your aware.
Simply check this ... Even samgsung TCCX chip is 3.6V Maximum rating ...
http://www.samsung.com/Products/Semi...8x16_rev13.pdf
http://www.nanya.com/attach_file/NTC-512M-B-DDR.pdf , Nanya ... 3.6V maximum rating !!!
http://www.elpida.com/pdfs/E0539E10.pdf ,elpida , 3.6V Maximum Rating !!!
http://www.hynix.com/datasheet/pdf/d...TP(Rev1.0).pdf , Hynix , 3.6V Maximum Rating
http://www.promos.com.tw/website/htm...C2512404SA.pdf , Promos , 3.6V maximum rating !!!
You are ignoring single glaringly important problem here...
4 STICKS DEAD EXACTLY AT SAME TIME.
Now tell me, all 4 bad sticks happened to die at exact same time?
Take my comment as whole not parts.
PS) whoever was asking about mushkin RMA, not as fast as OCZ but very courteous, informative and up to date RMA policy.
Quote:
Originally Posted by wtz54321
I don't know what happened to your dram or the board , but most of the reported case here only die with one of them ... Most of the dead case are 1 stick fail to boot and another stick is ok ... I can not say that no single DFI nf4 board will kill the memory , maybe your board is a bad one ... You are the only guy that I have seen here have 4 dead stick at the same time ...Quote:
Originally Posted by jinu117
I'm guessing bad board, I have had my vdimm at 3.4 for a while, no deaths yet
Just my .02 on this, take it for what you will:
I received a gig of Redline from Mushkin (for review) in late May. I didn't have the DFI board at that time, but I used the modules (at 2.7v) for well over a month. During that period, I tested them / ran them extensively on both the Gigabyte K8NXP-SLI and the ASUS A8N-SLI. They kept both systems absolutely stable under load and under heavy testing.
Last week my DFI NF4 board arrived. I fired up the FX-57 and ran another round of tests, again using Redline. After a week the modules were still completely stable, and still running perfectly at 2.7v.
I then flipped the 4v jumper, increased voltage to 3.3v (the maximum Mushkin recommends for the Redline 3500) and began testing the modules for latency and clockspeed. I hit 2-2-2-5 at 218, and 2.5 -3-3-6 at 250 MHz, still at 3.3v. Considering this was only PC3500 memory, I was quite happy with that.
I tested overclocking stability and headroom for about 3 days total. I then shut the system down for about two days.
Last night I turned the system back on to check a few things while working on the review, and discovered the RAM was no longer stable in Prime95. MemTest86+ confirms a damaged block on one DIMM at the 256.6 MB mark. The DIMM in question was in DIMM slot 3 (I was using DIMMs 1 & 3). I've run MemTest86+ on two separate systems--the memory in question is definitely damaged, and no BIOS flash on earth is going to cure it. The only way to stop the errors from appearing is to drop to around 133 MHz, at absurdedly high latencies.
Here's what I know:
Could it have been the memory? Yes. But the RAM was perfectly stable for over a month, even when stress-tested. Within four days of running on the DFI board, even though voltage never officially exceeded 3.3v, and the RAM was never swapped, it blew.
Given the multiplicity of issues reported with DFI's 4v setting, I can't say I see this as Mushkin's fault--at least not yet.
I can say for sure it is the motherboards fault, Nvidia chipset based ones are killing ram on the Intel side too.
My P5ND2 killed my OCZ PC4200 and killed my 600ADJ. It killed the ram with the system off. There was a surge of power when I went to swap the ram out and it killed both sticks.
One of my friends purchased the P5ND2 from Frys and it killed his Corsair's. I know now because I had to lend ram until payday. Im cringing now as if they kill once they can kill again.
I went to reboot without SLI enabled and it took out power supply.
Fugger,
What motherboard would you recommend for high voltages (3.0-3.5v)? I want to review this Redline, and I was happy with the memory itself--but I have no idea what motherboard to recommend.
It's not the voltage or the board kill the ram !!! It's the UTT problem ... With those good UTT chip , they can handle more than 3.6V for a long time ... But bad UTT chip , they are ok for a certain period , after that time , it will start to perform bad !!! ... The industrial standard is 1000hrs of operation ... And you can use high temp and high voltage to accelerate the time before these bad UTT chip start to degrade in it's spec !!! You never use high voltage in the previous two board for more than 2.8V , so these bad UTT chip still ok because the time is not coming yet !!! But you use high voltage to stress it and accelerate the time before it start to degrade it's capability !!! These bad UTT chip are those chip that have high failure rate in the first 1000 hrs of operation ... In standard chip spec , these chips will be sorting out and never been put in the module !!! But they are UTT now , and those bad ones are not sorting out and put into your module !!! You use high voltage , it die fast ... But even without high voltage , it will be dead after a longer period !!! And this is a typical UTT module sympton !!!Quote:
Originally Posted by Dputiger
WTF!, today my second stick also died... wats up with this crap? Well, atleast i can send them both out in the same RMA box... this sucks i thought the XP4000 can run up to 3.5v, is my DDR booster overfeeding them? DMM reads 3.48. I ran them at 3.7 for a few days under serious memtesting.. they don't even have a week on them how can they die so soon?
Trust Me !!! I am now 99.99% confident that the root cause of the dram death case is UTT problem itself !!! We did not see so many dram death case before !!! Even there is fault design K8 motherboard out there that will supply wrong DRAM voltage spike to kill the K8 CPU , we did not see any DRAM death case because of that board !!! But now with UTT DRAM , there are so many death case here ... If it's the fault design of the DFI NF4 board , we should only see dead DRAM in this board !!! But there are many other platform has this UTT death problem ...Quote:
Originally Posted by eBoy0
ive had twinmos speed premium die on a DFI UT nf3 250GB if it helps at all
it was at 3.6v
if you thought they could run up to 3.5v, why did you feed them 3.7v? Better hope Mushkin doesn't find out you feed 3.7v through them.Quote:
Originally Posted by eBoy0
On another note, when I first got DFI board I killed a couple of sticks of UTT ram. Since then, and after practicing some methodology that has been listed in some of the OCZ VX and Mushkin position statement threads, I haven't had any die.
Personally I think 50-75% of the problem is board issue. I Think about 10-20% is just nature of UTT ram, and no way to get around that unless manufacturers test these modules more, which will raise prices, and everyone would complain. Then I think about 25% or so is user error. I think people are spiking too much voltage, not reading information that is widely posted about how to setup and maintain the DFI when setup for 3.4v+ operation etc.... Or using too many volts. or whatever.
Fugger: I hadn't heard of the Nvidia boards killing off DDR2, good to know though so I don't get one :O :slapass:
Guys. as an engineer I need to point out that not only is the Vdimm important....and most of the DDR's have 3.6V ABSOLUTE MAXIMUM ratings (since the rating is based on the silicon wafer material and feature size)... but Vref/Vtt is important and the relationship between them.
Vref/Vtt needs to be 1/2 Vdimm(Vdd-Vddq) and if it diverges from this too far it can lead to failure. The chip is designed for Vref being 1/2 Vdimm and if Vref is "drooping" along with Vtt, then it is possible to break the rating on the MAXIMUM allowed voltage differential between Vref and Vdimm (Vdd/Vddq).
So it is possible the Vtt issue we see could have ramifications beyond just crashing Prime.
If this is true , how do you explain that there are many other dead case in other platform including NF2/NF3 ? IF it's the wrong voltage design of this DFI NF4 board , we should only see that dead dram case in this board no matter what kind of voltage design error to kill the dram ! But there are case reported in Old NF2 , NF3 with 3.3V Mod , NF4 with 3.3V Mod , 5V jumper Mode , And other board with booster ? Is that all these board all have wrong design in VTT circuit ???Quote:
Originally Posted by uwackme
the new DFI bios did not solve any cold boot ram issue or vdimm.
I got rid of my mushkin and went back to GSKILL thank god
The root cause of the cold boot problem is some chip in these UTT dram start to degrade in its electrical spec , they are original design to work under 2.6V SPD timing ... And I believe that all the user are able to use all the DIMM under 2.6V SPD timing in the beginning ... But after these bad chip start to degrade , they will need more and more voltage to work under 200mhz , spd timing ... Using the new beta bios is the workaround which set higher dram working voltage when bios detecting the dram ... But it does not mean that these bad chip is ok and the problem is solved !!! So ther are case reported that the beta bios solved his problem , period ... When these bad chips start to degrade once again , they will need even more bootup voltage ... If the bios did not set even higher bootup voltage , the cold boot problem came back once again !!! And tell me how could this be possible that it's the board supply the wrong voltage that kill the dram especially there are dead case is working with DDR booster or 3.3V mod ?Quote:
Originally Posted by Unrealcpu
WTZ, mushkin tests the modules you know? They have their very own binning procedure and facilities to do so.
It is not (soly) the chips themselves. I can understand some chips would slip through the net and stop working after a while, but this doesn't explain the mass deaths. I mean look it up on these forums. Or on muskin forum/webby, they have a own speed binnen procces which is in my opinion decent to say the least. They could do a lot less.
I suggest you don't look for the problem soly with the chips
If mushkin or OCZ can show us some pic or the detail about how these chip are tested , this will clarify the problem of UTT chip immediately !!! But we did not see any thing detail about this until now !!!Quote:
Originally Posted by Entity_Razer
Of course these module will pass the functional test !!! Do you know every UTT chip is functional from the FAB ? Do you know there is much more thing need to do to standard chip testing compare to UTT chip ? Please take a look at my thread about the industrial standard of how dram chip are tested and sorting out these bad chips ... It's almost impossible for module maker to do anything like that !!!
Take a look at this first ...
http://www.valueram.com/europe/newsl...s.asp#question
And read my thread about some industrial information about chip testing procedure !!! You will realize the main problem is UTT chip total quality problem !!!
http://www.xtremesystems.org/forums/...ad.php?t=65478
I only see a couple deaths here or there on platforms other than DFI.
That said, once I followed proper procedure I haven't had any problems with Redline, OCZ or G.Skill UTT.
Personally, I think this whole thing is getting blown way out of proportion. While it is quite irritating that these issues are present. These are certainly super hiperformance modules and do take some care in running. I would not recommend these for the novice, but with proper care and use, they are damm damm damm fast.
Since there is other dram dead case other than DFI board , it means that the total UTT dram quality is questionable ... No matter what kind of DRAM voltage mod is used in any other platform , it's hard for these board supply dram voltage higher than 3.6V , and this is below the DRAM spec 3.6V maximum rating ~~~ But the DRAM is still die in these platform !!! It can explain that voltage is not the main key reason that cause these dram to non-bootable !!!Quote:
Originally Posted by JNav89GT