Infinity SLI VDIMM problems

[SADtherobot]

I have an Infinity SLI running in my friend's computer, it's a Rev.A board. I'm using some old OCZ Value VX cause I wanted to see what I could do with it. I configured his system months ago and it was running for 2-3 months without any problems. He then mentioned that it started to blue screen on him, so I went to check it out. I noticed that the problems were RAM related, but I knew the RAM was good, since it had never given me problems, and I had thoroughly tested it (30 loops of memtest86+, 24 hours mprime, 24 hours Prime95, 12-16 hours 3DMark) on his board without a single error. I was running the RAM pretty fast (I can't remember exactly, it was in February when I first configured it) so I needed some pretty high voltage, so I had it set to 3.1V, but I knew it was not good to run too close to 3.3V, so I turned up the 3.3V rail a little (he has an OCZ powerstream 520W) to compensate. Since I knew the RAM was good and the PSU was giving decent voltage, I decided to look around the BIOS for any clues, that's when I noticed that the VDIMM voltage was jumping around from 2.65-3.2V in the BIOS's VDIMM voltage monitor. I was shocked to see something like that happening, so I knew that some component on the board had failed. I also noticed that after I exited the BIOS and rebooted that it wouldnt boot anymore cause the RAM wasn't getting enough voltage, so I cleared the CMOS and ran at much slower settings using a lower VDIMM setting. After rebooting this time, I noticed that the CPU fan had stopped spinning and I couldnt get it started again. I made sure the fan wasn't dead by plugging it into a different header, sure enough it worked fine. I tried the CPU fan header again and it still wouldnt work, so i tried lowering the CPU fan speed percentage in the BIOS from 100% to about 85%, this got the fan working again. Obviously the board had problems, so I put a temporary board in my friend's computer, then I went through a huge ordeal to get the board RMAd.

 

My friend lives far away, so I didn't get to put the Infinity back in for a while. I put the board in about 5 days ago and set it to some decent settings and again I used 3.1V just to see what would happen. I tested the board for days, more memtest, prime, and 3DMark, all passed without errors. I decided to do a single run of 3DMark yesterday, just to see a final score with the new drivers and other things I changed. At some point during the benchmark, the computer threw a blue screen, but no error message (like the usual "bad pool caller" or "IRQL not less or equal", etc.). I hard reseted the computer, but it wouldn't restart unless i held the reset button for a while. When it booted, it gave me some message about the system being in safe mode cause the CPU was OCed or something (though I hadnt OCed the CPU in any way, all clocks for everything were at stock at this point in time). I entered the BIOS and went straight to the VDIMM voltage read-out, sure enough, it was jumping around again. I don't know why I got a message about the CPU, cause the other voltages looked fine and the settings were all at the defaults, but I assume that the board doesn't have any messages about the RAM, so it just said something about the CPU cause it knew it couldn't boot for some reason (since the VDIMM voltage was too low when it tried to boot that time). Now I guess i'll have to RMA this board as well, but my friend leaves tomorrow, so I will have to give him a replacement again, and I have no idea when i'll get to try to swap this board back in (that is, if he wants to go through the trouble 1 more time). It's quite ridiculous that the board failed so quickly after it was repaired (I assumed it was repaired cause I got the same exact board back and it had a different smell, so they probably replaced a few components). I was wondering if other people have had this type of failure. I've read a few other threads that discuss both VDIMM and VTT problems, so maybe the Infinity boards are prone to failure. If this is true, does anyone know if the Rev.B have any of the same problems?

 

I'm currently having memtest failures even at 2.7V, so I'm guessing I should RMA the board, but should I even bother RMAing this board again?

 

If you must know the specs of this system (though they are almost irrelevant in this case cause the board IS broken, it's not some other component causing a problem), they are as follows:

 

Athlon64 3500+ (Newcastle core)

DFI Infinity SLI Rev. A (4/10/06 BIOS)

1GB OCZ Value VX

EVGA GeForce 6800GT

Maxtor DM10 250GB (yes, this drive sucks)

Toshiba DVD-ROM

Samsung CD-RW

OCZ Powerstream 520W

[red930]

It seems as if the CPU has gone goofy. The only test that puts any stress on the CPU is the 3DMark.

 

Since it's only bombing out on that test, I would look at the CPU.

[SADtherobot]

It seems as if the CPU has gone goofy.

 

The CPU works fine, I tested it in another computer, plus it worked fine in the other motherboard I put in my friend's computer.

 

The only test that puts any stress on the CPU is the 3DMark.

 

Incorrect, Prime95 puts plenty of stress on the CPU, it actually stresses it more than 3DMark does.

 

Since it's only bombing out on that test, I would look at the CPU.

 

Now it doesn't work on any test at the old settings (because the old settings require more VDIMM voltage), as I said, even Memtest86 is having errors now. It's because the RAM voltage keeps fluctuating. Which is the same problem I had with this board before, hence the reason I RMAd it for re-work. This time the board just failed a lot faster.

 

My entire post was about how the board has a problem with the VDIMM voltage jumping all over the place. It started failing after putting a huge load on the computer for days, it just happened to fail during the last run of 3DMark2001, but it could have failed at any time. The problem is 1 or more defective components on the motherboard that fail at regulating the VDIMM voltage. I was simply trying to find out if anyone knew if the Rev.B boards have the same problem. I have searched this forum and found a few posts about anomalous voltages and temperature readings on the NF4 Infinity boards, so it might be a common problem.

 

From what I've read, people have some very random results with the Infinity boards, so there seems to be a number of problems with them. I'm hoping that someone can get a Rev.B board to test if anything has been corrected.

[soundx98]

man I've never read about this problem before.

I'm sure you have swapped RAM and CPU to try and isolate the problem.

Earlier Infinity's seem to be famous for voltages jumping around.

Me, I'd go for an RMA. It may be something that was missed Or getting ready to fail in the previous RMA.

 

I know you are not a neophyte and even if you get it working again if it fails it all gets dumped on you again.

The problem with building a rig for a friend or family is the old "no good deed goes unpunished" adage.

Who knows if the user switched RAM while the board was on or a dozen other no-no's. DFI has the facilities and tech's to get it right or replaced.

I've talked to too many users that finally fessed up to "I even cleaned the memory contacts - the cat knocked it over - we had a big storm - it worked fine until the night of the Singapore Sling party - etc.

I'd start the process again, note the serial number and include a detailed explanation of what has occurred. (don't expect a reply but it couldn't hurt)

They'll probably put it on the "ghost in the machine" area and send a replacement.

 

I've found the RMA Customer Service to be helpful and prompt and never had to RMA a board more than once but I'm sure it has occurred.

 

EDIT: Sorry, I was typing while you were posting.

[red930]

Incorrect, Prime95 puts plenty of stress on the CPU, it actually stresses it more than 3DMark does.

LOL missed that one. But of course. Got Lost Vegas on the brain.