Crucial Ballistix Stability -- This doesn't make any sense

A bit of context first, sorry if it gets too long... Specs are as follows:

Mainboard: DFI Lanparty UT nF4 SLI-DR @ 623-3 BIOS, FSB200
CPU: Athlon64 X2 4400+ @ stock 2.2ghz
RAM: Crucial Ballistix PC4000 1gb (x2) @ 2-3-3-8
GPU: Asus GeForce 7800GTX (x2) @ stock
HDD: WD Raptor 74gb (x2 in RAID0)
PSU: Enermax 660W EG851AX-VH pricey, full of rails and stability
Case: Lian-Li PC-V1000+

I've been running my rig at stock FSB (200Mhz) ever since I got it, and I noticed that this RAM seemed good enough to run at 2-3-3-8 at this FSB, which was rather good for 1gb sticks.

However, about a month ago, I started getting random bluescreens, especially in memory intensive apps like Photoshop. It definitely seemed like a RAM problem to me, and since memtest86 didn't find anything, I ended up checking my voltages. Sure enough, while DRAM normally runs at 2.5V, these Ballistix are supposed to run at 2.8V. So I changed that, and had two weeks of uninterrupted uptime.

Then, a week ago, it craps out again. Except this time, it won't even boot into windows before BSODing, and memtest86 reports thousands of errors at random memory ranges. Loosening the timings to the specced 3-4-4-10 doesn't help, nor does swapping or reseating the RAM. I found it extremely weird that the RAM would mess up overnight like this and wondered whether the problem was actually elsewhere. Asking Crucial tech support got me a generic "Run Windows memory tester and ask for a RMA if it fails" response. I ran Windows memory tester, and sure enough, it got errors as well. So I took the sticks out while waiting for a RMA number.

Now this is where it gets strange. On a random fling, I decided to put the sticks back in while waiting and experiment some more. Timings are still 3-4-4-10. I start running windows memory tester and leave it on. It gets tons of errors for two or three passes, then... The errors stop. Leaving it overnight yields flawless results past the initial passes.

I figure what the heck, and boot into windows and play around a bit. Photoshop is smooth, Titan Quest as well, and get no crashes for 3-4 days. Today I decide to set the RAM back to the original 2-3-3-8 and see if that mysteriously fixed itself too. Memtest has been running for 2 hours now and it's absolutely clean.

I'm absolutely boggled at this, what might've happened? Divine intervention visited upon my aging, underexploited rig? Should I send in the RMA I asked for anyway, or chalk it up to a freak accident that won't happen again?

I put in for my second RMA just this morning on a 2GB kit. The first set died after a month back in '06. Things started going south again a couple of months ago.

Yours may be momentarily working fine, but I get them replaced. They'll eventually get bad again if they were showing those signs.

I am also having probs with my crucial ballistix ddr2 800 running 800 @ 2.2v 4-4-4-12 you say yours are supposed to run @ 2.8v why are mine only rated 2.2 @ 800? Anyway my problem is when i run memtest i get 2 passes then memtest freezes yet in vista i have no problems running any apps and orthos ran 2hrs stable on blend test which is stable enough for me im going to rma mine but i have a funny feeling they are going to return them saying no faults found anyone else had probs with memtest freezing and is it safe to up vdimm to 2.40 for stability testing?

Jon said:

I put in for my second RMA just this morning on a 2GB kit. The first set died after a month back in '06. Things started going south again a couple of months ago.

Yours may be momentarily working fine, but I get them replaced. They'll eventually get bad again if they were showing those signs.

Click to expand...

Yeah, I had a first RMA done a while ago. Looks like failures are somewhat commonplace with these...

Oc1Kenube said:

I am also having probs with my crucial ballistix ddr2 800 running 800 @ 2.2v 4-4-4-12 you say yours are supposed to run @ 2.8v why are mine only rated 2.2 @ 800?

Click to expand...

Because I have DDR PC4000, not DDR2 =)

oh i thought x2 processors were ddr2 only but thats am2 isnt it doh...:O)

It makes sense. Electronic failures often cause intermittent trouble. Get everything just so and it will work for a while, but that state is impossible to maintain and the part will usually continue to degrade if you keep using it. You're right to RMA the RAM.

If you have the same problem with your new sticks, your PSU might be going south.

Could it be a flaky voltage issue?

Yes, it could be. It might also be the memory controller. Intermittent errors like this are hard to diagnose because you can't tell whether it works because it just decided to work for a while or because of the change you made. But you've got to start somewhere, and if you get memory errors, the RAM is the most likely culprit. Now, if the errors only occurred when CNQ was enabled, I'd try swapping the CPU first.

i also have a busted 1gb kit... while its awsome memory it seems to die fast and easy

Going to bump this because it's summertime and, having no classes, I can afford to be without desktop for some weeks as hardware gets RMA'd. Anyway.

I'm having bluescreens galore again. Since I never actually did RMA the memory I referred to in my original post, my first suspicion was cast on it. So I ran memtest86 for 10 hours... 100% clean. Not a single error. The next day, run Windows mem diagnostic, also comes up entirely clean. By the looks of it, whatever issues were plaguing this memory before are altogether gone now.

Figured it might voltage. But measuring voltages directly on the power supply cables nets me 5.22V on the 5V rails, 12.24 on the 12V rails and 3.39 on the 3.3V rails. These numbers vary by barely 0.01V while running FEAR benchmarks in a loop. So if it's a voltage issue, it's from the motherboard, not the PSU.

For now the most likely culprit is overheating. After the aforementioned benchmark loop, I read 47 C on the CPU, 87 C on the GPUs and 70 C on the nForce4 chipset. This is pretty much due to the nF4 SLI-DR's bad positioning of one PCI-E 16x slot exactly over the chipset, so the heat from the 7800's dissipates directly into the chipset. While this definitely seems like the culprit, it hasn't at all caused a problem up until now.

So I'm faced with a conundrum:
Could it still be a RAM issue? Should I RMA just in case, even though it hasn't shown any errors in tests for a while now?
Could it be the chipset? Should I replace the bad stock fan on it? Or just say screw it and get another 939 board? If so, which one?
Could it be some other variable I haven't taken into account?

crucial & volts

hi'
I just changed my geil for crucial ballistix ddr2 pc 6400 c4/4.4.4.12 at 2.2v ...
on my first boot I went into bios to check if everything was ok, and set the vdimm on 2.0v ... boot, no problem!

then, I went on crucial site and saw this 2.2v ...

I didn't notice any major change with 2 or 2.2v on ram

so, might be that I'll set them back to 2.0v, as they are not OC
and will never be, even if I can reach 3.6ghz with a 9x multix400
it will be still 400, and so 4.4.4.12 @ 2.0v

my 2cents worth here

thinking also, that less volt on ram as for the cpu isn't bad at all !

i686

Oc1Kenube said:

I am also having probs with my crucial ballistix ddr2 800 running 800 @ 2.2v 4-4-4-12 you say yours are supposed to run @ 2.8v why are mine only rated 2.2 @ 800? Anyway my problem is when i run memtest i get 2 passes then memtest freezes yet in vista i have no problems running any apps and orthos ran 2hrs stable on blend test which is stable enough for me im going to rma mine but i have a funny feeling they are going to return them saying no faults found anyone else had probs with memtest freezing and is it safe to up vdimm to 2.40 for stability testing?

Click to expand...

Seems high. I would RMA immediately if they won't pass memtest @ stock settings. My Ballistix(DDR2 800 as well) wouldn't pass memtest after just sticking them in the board, which defaulted at 5-5-5-15. Had to crank the voltage to 2.25v and set them at 4-4-4-12. Been there ever since.

Yeah, DDR ram could take some serious voltage, for example Mushkin's Redline series which was rated @3.2v.

I think I found my culprit.

I investigated the southbridge overheating bit, and noticed the DFI stock fan on the nForce4 is *not spinning*. That'd explain the blessed silence coming from the PC lately, though apparently said silence comes at a steep price.

In spite of this, I remain skeptical about the likeliness of a nForce4 chipset overheating and causing bluescreens. The actual memory controller is on-die for this AMD64, so the chipset shouldn't matter that much -- and it seems to be crashing in memory intensive contexts rather than HD intensive.

I want to run some stress tests to get to the bottom of this, what should I use to stress nF4 chipset and RAM, respectively?

SiSoft Sandra does some burn-in stuff that I believe you can set to your desired length, so probably 8-12 hours should easily be sufficient.

Orthos is what I use generally. Very good program. It will stop you way before damage can occur.

I went through FIVE pairs of those sticks. One of the failures passed memtest all day, but I'd get bluescreens like you're talking about. The only way I was able to verify 100% is by putting in my old BH5 chips.

Basically, I think it's your RAM, but I'm biased.

I think it is both. My experience with that board is this. Immediately replace the fan with a evercool vc-re fan and put as5 on. The chipset fan is horrible cooler and most of the time the paste they use is not placed good on the chip. If your computer turns off and reboots then most likely the chipset is running too hot and is force to shutdown. It also may be damaged if you continue to use it but mostlikely as mine shut down over 70c it shouldnt be unless they completely did not put the paste on right. The fan dies out also quite fast and makes a horrible noise when spinning at max. My experience with ballistix is as follows. Great overclocking ram but longevity is short. First pair I bought were doa, second replacement dead after 2 weeks. Got them to refund me my money and bought g skills. I have been happy for nearly a year so far with g skillz. I had them running ddr500 and they are ddr400. When I installed my opteron I set them at 500 and they ran smooth with no change in timings. so I was at 250x10 and ddr 500. Was very pls.

I've got a dodgy 1GB set of TCCD

Passes memtest everytime, Orthos stable, but SuperPi32M crashes instantly, replace memory - it works.

If I try them in another motherboard, they fail memtest with thousands of errors. So I know the RAM is at fault. I'm getting them replaced under Corsairs lifetime warranty

add me to the list of RMA'ers for ballistix. went bad about 2 weeks ago. only problem is that they no longer offer what i bought so we'll see how they treat me.

2 rma's on pc2 6400 1gb set here

How is Crucial's RMA service?