Possible cause of DFI ram deaths.

@hovo73

I'm just being a bit sarcastic, you know? :D
The point I'm trying to make is that, If you have UTT sticks and are affraid to either: (a) use 5v feed -- or -- (b) use more than 3.2v all together, then don't. It's a risk and it seems that you are not willing to take, as you have UTT sticks. Which I perfectly uderstand, because something is wrong with those mems.
DFI has provided end-users with JP17 as an OPTION. From what we gather around the forums, it's seems that overwhelming majority of dead RAM cases has to do with UTT RAM and most of the hysteria comes from those users. There is nothing like "jumper issue"... It's just seems to be the problem with RAM as it somehow gets messed up after some time of using it.

If I were you I would get some 2x256MB of "old school" BH-5 (most of them will definetely do >250Mhz on DFI NF4) and go crazy with volts on them. Bench with it. But for everyday tasks get some TCCD/TCC5 as they are a alot more common these days that 2x512MB "old BH-5". It would make you feel a lot better.

My xp-90 covers one stick of ram nicely but this is how I cool the other and my mosfet, they are zip tied to the bar, the fans are 30 something cfm...

Also I've really cleaned up my wires since this picture :D

bachus_anonym, I understand what you are saying, but I have a better suggestion ...

How about DFI put it shiate back together, and actually come up with solution to this problem instead of bogus excuses? How about that?

I think I like that better :)

PS. Please understand I am not trying to annoy you or anything (i am saying this because i have several exchanges with you on this topic). I just want my damn board to work as advertised. Thanks.

Quote:

---

Originally Posted by hovo73

bachus_anonym, I understand what you are saying, but I have a better suggestion ...

How about DFI put it shiate back together, and actually come up with solution to this problem instead of bogus excuses? How about that?

I think I like that better :)

---

Did you read what OscarWu said recently? That's how I look at it...

DFI says everything is OK with the board and there's no "jumper issue" at all...
OCZ and Mushkin reps say RAM is OK and they have literally no RAM being killed in their lab testing...
Now, users have a TON of dead or messed up UTT RAM. Something like this phenomenon has never been happening before. But oddly enough, old BH-5 sticks are not dying and many still have those running flawlessly, including myself. Do you see the picture here? Someone out there is wrong and I think it's not DFI.

Are we back to the argument that only UTT sticks are dying? I thought we were over it?

I agree that the majority of sticks are UTT, because (a) that is what is available right now, (b) if anyone is using TCCD they are very unlikely to use the 5V line, (c) very little "old" BH-5 sticks around, especially 512MB modules, and most people would agree that 2x256 is not enough for today's needs.

But the fact is that non-UTT memory, such as TCCD and "old" BH-5 is also dying.

PS. So you take DFI word for it, but not OCZ and Mushkin? Do I understand it correctly, if not what is the problem than that so many people are getting these problems? It can't be simply because of accident or user error. Why are these sticks not dying on other boards (Abit), or are they?

I can tell you something honestly. If you look at my sig, you will notice that I have 4 sets of RAM, including TCCD. You'd be surprised If I tell you, that I've been running with JP17 on 5v feed, no matter what RAM I used.
Up to 3.1v on TCCD or 3.9v on BH-5. Literally, I'm on 5v feed since day one and never really run it otherwise. According to DFI, I should not use this option as it's unnecesary for TCCD. You know what? I don't care about that because I'm absolutely convinced that it's not the cause of dead RAM. Components on the mobo are not hot (as far as I can tell) untill I crank up vdimm to beyond 3.7v. When I ran 3.0v for TCCD on 5v feed, underside of the mobo was not hot. It only gets hot above 3.7v.

I strongly believe, that it's not that DFI has a problem...

Quote:

---

Originally Posted by hovo73

PS. So you take DFI word for it, but not OCZ and Mushkin? Do I understand it correctly, if not what is the problem than that so many people are getting these problems? It can't be simply because of accident or user error. Why are these sticks not dying on other boards (Abit), or are they?

---

That's the problme I mentioned before. We do not have a lot of folks with other NF4 boards and UTT RAM configs. So we cannot tell.
But, like I said above, I had all thogether 3 sets of OCZ VX PC3200 and two of them went back for RMA. Same symptoms as you guys are getting now on DFI NF4. But there's one thing I added... THAT WAS ON NEO2 w/ DDR BOOSTER ;) That's why I do not believe that DFI NF4 is responsible at all.

Quote:

---

Originally Posted by bachus_anonym

I can tell you something honestly. If you look at my sig, you will notice that I have 4 sets of RAM, including TCCD. You'd be surprised If I tell you, that I've been running with JP17 on 5v feed, no matter what RAM I used.
Up to 3.1v on TCCD or 3.9v on BH-5. Literally, I'm on 5v feed since day one and never really run it otherwise. According to DFI, I should not use this option as it's unnecesary for TCCD. You know what? I don't care about that because I'm absolutely convinced that it's not the cause of dead RAM. Components on the mobo are not hot (as far as I can tell) untill I crank up vdimm to beyond 3.7v. When I ran 3.0v for TCCD on 5v feed, underside of the mobo was not hot. It only gets hot above 3.7v.

I strongly believe, that it's not that DFI has a problem...


That's the problme I mentioned before. We do not have a lot of folks with other NF4 boards and UTT RAM configs. So we cannot tell.
But, like I said above, I had all thogether 3 sets of OCZ VX PC3200 and two of them went back for RMA. Same symptoms as you guys are getting now on DFI NF4. But there's one thing I added... THAT WAS ON NEO2 w/ DDR BOOSTER ;) That's why I do not believe that DFI NF4 is responsible at all.

---

So whats ur 24/7 rig then. Or do you just benchmark with 5v line??

Put a fan on that area and it wont be hot at all. Mine's not even warm after hours of memtest @3,5v using 5v rail.

I'm with bachus anonym on this one. Nobody is making you use the 5V line and the fact that you waited 2 months for the board is not a good reason to if you are afraid of your ram dying. I have some old school BH-5 running 3.7V daily and everything has been fine. I think the reason we have so many complaints is that everyone has a DFI as stated previously, since it is the board of choice. So of course we are going to get a lot more reports of this board and the pointing game will keep going back and forth.

But as always with overclocking, dead components are part of the risk and whether or not to take the risk is your own choice.

Quote:

---

Originally Posted by AMD-me

So whats ur 24/7 rig then. Or do you just benchmark with 5v line??

---

Currently I'm on:
- very trusty, over 1.5yr old Abit NF2 with every possible vmod done, running 2x256MB BH-5 @ 3.5v since almost day1. This PC has been crunching for DC projects for most of it life, day or night, 24/7. It's always turned ON, I ocassionally reset it for maintenence. Main day-to-day PC.
- 5mos old DFI NF4, on "5v feed" since day1. Before I mounted phase change cooler on it (until 2mos ago), it was run 24/7 with Gskill @ 3.0v 290 6-3-3-2.5-1T. Then it mostly run on 2x256MB Mushkins @ 3.8v 263 6-2-2-2.0-1T Mhz (3.6v is actually enough, but I didn't change it as I use 3.8v for benching :)), for about 12hrs/day. Currently, I'm awaiting new CPU, so rig should be back online early next week.
- I used to have MSI K8N Neo2 and DFI NF2 Infinity. Literally, every board I had run with vdimm north of 3.5v for 24/7 operation with BH-5. Neo2 worked great 24/7 (day-night 100% LOAD) @ 3.5v 256MHz 6-2-2-2.0-1T with OCZ VX3200 until mems died few weeks later. Got a new pair and soon same happened. Since then, I ditched UTT.

Please, try to look at it with your eyes very wide open. Don't get carried away, as many others did, by all the hysteria. I'm not surprised now, as DFI made HIGH VOLTAGE OVERCLOCKING a mainstream trend, not like it used to be before - reserved only for experienced users, eager to volt-mod their boards and take the risk.
Seems like many folks around still don't get it:
DFI gives you the tools that DO WORK properly, but cannot be responsible what you do with it. They ship the boards with JP17 in 3.3v position. Once you chose to switch it to 5v, you sign a waiver called "DFI is not responsible for RAM being killed due to use of extensive voltage", whatever RAM manufacturers claim is safe.
OCZ, Mushkin, etc... overclock Winbond chips that normally would be rated @ 200Mhz 6-2-2 2.7v. For them to achieve an e.g. 250Mhz 6-2-2-2.0 they NEED to raise voltage requirement, to e.g. 3.3v. That is already out of spec. Whatever they claim, you need to keep in mind that components run out of spec MIGHT die sooner. That's why they HAVE to give you a warranty for 3.3v as this is what is required to run them. But sticks MIGHT die @ 3.3v, manufacturers NEED to know that (I'm sure THEY DO) and have to be ready to give you replacement. They entered the area with high-voltage/high freq products where they need to be ready for stuff to die.
... and this is what is happening...

Another point is, that with UTT-CH5 based modules, there is one unknown factor - life expectancy when over-volted.
Not many people have run "real CH-5", months back, at anything north of 3.2v. CH-5 did not have much success and did not gain too much enthusiasm within OCers then. CH-5 surfaced again with introduction of Winbond UTT chips. So only now we are finding out how CH-5 behaves under new circumstances. Something messes up with those sticks. I don't know if ICs die or SPD chips get messed up. But certainly something seems to be wrong.

OK, I'm done... I wrote too much already :lol:

Somebody mentioned a good point earlier about this, and It seems feasible. Isn't CH-5 a newer, smaller process than BH-5? If UTT is a relative to that wouldn't that also be on a smaller process?
I'm certainly no expert, but even I know the smaller the process the less volts you can run thru it without burning it up. If these are more delicate chips and they run well right up to the burnout point, they probably wont give you too many obvious signs before they fry.
Either that or there's a wide inconsistancy in the process' voltage tolerance, and people are seeing what other guys do and find out the hard way that their particular chips cant handle that much voltage.

Quote:

---

Originally Posted by hovo73

bachus_anonym, I understand what you are saying, but I have a better suggestion ...

How about DFI put it shiate back together, and actually come up with solution to this problem instead of bogus excuses? How about that?

I think I like that better :)

PS. Please understand I am not trying to annoy you or anything (i am saying this because i have several exchanges with you on this topic). I just want my damn board to work as advertised. Thanks.

---

It's impossible for any MB company to gurantee and be responsible for the damage due to Over Voltage and Over Clocking ...

Remember ... In Fact , even 2.8V VDIMM is not guranteed for some dram ...

If you use 2.8V memory in MSI or ASUS board but the DIMM is dead , what will MSI or ASUS told you about this problem ? Will they say " Yes , our board kill the memory and will be reponsible for this ? "

If the problem belongs to the memory chip but not the board , how DFI can do to this problem ??? :confused:

Until now , we did not get a firm anwser that it's the board kill the memory chip ... If it's this case , DFI must do something to this ... But if it's not the case , I really don't think DFI can do anything ...

We are extreme overclocker , but we still need to know that overclocking and over voltage is not riskless and gurantee by the company until they say clearly about it ...

One important thing now ... And let's why I also think UTT is the trouble maker ...

We did not see 1st tier memory module company such as " KINGSTON " or " CORSAIR " have any UTT type DRAM product ... There must be something wrong for them to skip UTT DRAM product ...

I agree with bachus_anonym here, My old BH-5 is fine (although I'm only running 3.33v through them 24/7) and so is eva2000's and pretty much everyone else who is running 5v jumper with BH-5 ram. And the point that Corasir and Kingston havent made any products that run above 2.86v using UTT is also mor proof that UTT is probaly the cause of this RAM death.

But, and this is a big but. The DFI NF4 board when on the 5v jumper, supplys 2.7v to the ram, when the PC is switched off, after the AC power cord removed and then plugged in again. This is the casuse of the cold boot issues, and could contribute to the ram death's. Maybe?

Quote:

---

Originally Posted by loc.o

Put a fan on that area and it wont be hot at all. Mine's not even warm after hours of memtest @3,5v using 5v rail.

---

Yeah, I did that and with a fan aimed at it it's barely warm. I'm running it out of case so the only air on it was whatever could hit it from my heatsink fan.

this issue IS a problem, alone because of the critisism.......i just put my pc4000VX into the board, and got unsure of what to do...cooling the RAM, mosfets actively i should have done what i could to prevent the ram dying situation......but then somewhere in my old memory something popped to the surface........i read somewhere in her that it is possible only to move 1 jumper and then get 3.5v option...not the full 4v ...can anyone point me to that post....?

overvolting your ram is your decision, but we do need to figure out if it IS the Ram or if it IS the board... you could make accusaitons at both sides, like

its untested so it could be a cheaper process and lead to sooner death
its a smaller and more delicate chip than the bh-5 so less tolerance

and on the mobo side

we all know that mobo manufacturers often skimp on voltage cleanup and regulation to components... we all see that with cpu idle/load voltages and the same for ram, etc.... motherboard manufacturers could spend more and put in a power feed thad dosent droop... i mean heck, i can use a simple pencil and trace over 1 resistor and clean up the voltage and have a .01 droop on the vcore with my NF7-s.... so obviously that with the 5v jumper set that the voltage may be unclean and posibaly speratic when put under load at boot... i mean really, your running IC's that spec at 2.8v and alot of them are dieing at 3.4+ volts, but when the board applies the voltages it basically jams it through the stick from the initial 2.7v that the bios applies to "find the sitcks"

you could argue eitherway.... we do need to find a solution though

Quote:

---

Originally Posted by bachus_anonym

I can tell you something honestly. If you look at my sig, you will notice that I have 4 sets of RAM, including TCCD. You'd be surprised If I tell you, that I've been running with JP17 on 5v feed, no matter what RAM I used.
When I ran 3.0v for TCCD on 5v feed, underside of the mobo was not hot. It only gets hot above 3.7v.

I strongly believe, that it's not that DFI has a problem...

---

Micheal... I would very strongly suggest you don't do the <3V when on the 5V jumper anymore... at least using the standard jumper config ;) Put your finger on D8/D7 in the corner on top of the board....

Peace :toast:

Daniel,

I know, I know... I kind of :) forgot to say, that I use BH-5 on a daily basis, more than TCCD - at least recently. I'm not saying that everybody should stay on 5v no matter what. I was just trying to prove my point, that there's no danger in using JP17 to source vdimm from 5v rail :)

ygpm :)