Why is my new build locking up?

Korrd

Limp Gawd
Joined
Apr 19, 2005
Messages
237
I just put together a machine for someone and I've run into some trouble I'm hoping you fine folks can help me out of.

Here's the system in question:
Q6600 (stock cooler, no overclocking)
MSI P6N SLI Platinum
2x2GB G.Skill PC6400
512MB Radeon X1650 Pro PCIe
(2) 320GB Western Digital (reused from old build)
500w Sunmax PSU (reused from old build)
Generic PCI USB, Firewire800, and Serial Cards
Belkin Memory Card Reader

I got the system assembled, installed Windows XP, loaded drivers, patched, etc., etc. It wasn't until after all that, when I tried to copy some backed up data from the secondary drive to the primary/OS drive, that the system crashed with a BSOD.

STOP: 0x00000003 (0x000000F4, [varies], [varies], [varies])

I tried again with the same results each time. After a while, Event Viewer reported an error reading data from the secondary drive. I switched the drive to another port and thought for a while the problem was solved... until it crashed again after copying ~45GB.

The only other error in Event Viewer during all this is about the parallel port service failing to start, which I believe is somehow related to the "Unusable Parallel Port" installed as part of the 2-port serial PCI card.

A reinstall of the onboard SATA drivers again led me to believe the problem was solved, as I was able to copy ~120GB without issue. I decided to fire up Orthos while I copied some more data, just to put the system through its paces (the system is not and will not be overclocked). A few seconds in, the system locked up solid.

Through much trial and error I've learned the following: two instances of Orthos (with affinity for two cores each) and moving data results in the computer locking up inside of a minute. Running one instance of Orthos while copying, or two instances without, also locks up the system, albeit after a slightly longer period (up to 20 minutes). One Orthos instance or copying by itself works for longer still (a transfer of ~150GB was able to complete), but I can't very well give this machine back to the poor guy with a stipulation that he never multitask.

EDIT: Both Orthos and Prime95 v25+ yield similar results.

Heat does not seem to be an issue, since both the CPU and northbridge temps are well within reason. I threw in my PSU (Antec 480w NEO... Sunmax doesn't inspire much confidence), but it changed nothing. Memtest checks out, though I only did a single loop. Frustratingly, I don't have the parts to swap out (CPU, MB, RAM) to properly diagnose the problem. My assumption would be a bad motherboard, but I thought I might get a few opinions before I decide it's conclusive enough to start an RMA.

Thoughts?

EDIT: Also, as an aside, I'm confused by PAE. I thought PAE was only used for over 4GB RAM, yet it's active on this computer. System Properties also reports only 3GB... is that because of PAE or should I be looking at that as a symptom of the above issues?
 
XP 32 will not recognize all your 4GB systerm memory, you will need a 64 bit Operating System for that.
 
Sigh.. here we go again.. :D

XP WILL recognize 4Gb of memory. However, with a 32-bit OS, it will NOT all utilize all of it. It will subtract your video card memory, and that will be the functional memory.

Hence, 4Gb, with a 512 video card, it will show up as roughly 3.5GB memory. More than enough for anything outside of massive adobe / video editing at this time.

If and when you choose to transition over to a 64bit OS, at that point it will utilize all 4Gb.
Hope this clears things up.

PS: I'm not sure if by enabling "Memory Hole" (can't remember the exact wording) that will help. I'm tired, and it's time for bed.

On a different note, that PSU is defiantly hitting the lower end of the scale for the hardware you are running. I personally have no experience with Antec Neo, but your hardware may require something a bit more robust and cleaner.

Another site of frequent issue. It can happen that when you install your memory, the BIOS does not accuratly detect the voltage requirements. It can be undervolted. As a result, the more the memory gets stressed, the less voltage, and the more errors ensue. If that occurs with a fresh OS install, you will have instabilities throughout. Double check the voltage requirements, and make sure they match up in your BIOS. Hell, I would just bump it up a notch no matter what. A small voltage increase will NOT harm your sticks.

The culprit seems to be heavy CPU usage IE ortho's / Prime. I would check both of the above stated.

Hope this helps!
 
Get a copy of Memtest and run /each/ stick separately through a couple hours of tests. I think it might be your RAM. Also, what voltage is your RAM rated for? Try bumping it up a couple of notches. It may just be undervolted and unhappy.
 
Thanks, guys. I appreciate the responses.

The system's resumed crashing with BSODs, it seems (no more mere locking up :p ). I've upped my voltages to see if that helps. Currently my RAM is set to 1.95v (manufacturer specs state 1.8-1.9, SpeedFan reports it as 1.90) and the CPU is at +0.075v (BIOS reported default as 1.288, SpeedFan of 1.34). It's still crashing.

I've got it running Memtest on one of the modules right now, but I'm not expecting anything conclusive. Call me pessimistic, but that's how this thing's played out so far. :(

I wrote in my original post that temps were fine. However, I just loaded up SpeedFan, Core Temp, and Intel's TAT. SpeedFan reports what I expected (max ~55C under load), but Core Temp and TAT both show temps hovering around 70C. I understand some people run with temps even into the 80s, but... should I be concerned about stability issues resulting from these temps?

Also, I had the STOP codes reversed. For example, just now it crashed with:

STOP 0x000000F4 (0x00000003, 0x891EEC08, 0x891EED7C, 0x805D11F8)

Hence, 4Gb, with a 512 video card, it will show up as roughly 3.5GB memory. More than enough for anything outside of massive adobe / video editing at this time.
Okay, but I've got a 512MB card in there and it only shows 3GB. Why would that be? And... why does it subtract the amount of video RAM? I don't want to get OT since all I care about now is making this thing stable, but I'm curious. Also... what happens if, with 4GB, you add the /nopae switch to boot.ini?
 
I ran one module through almost 7 loops of Memtest, which it completed without any problems. The other module, however, generated some 40,000 errors partway through a second loop. I ran Memtest again when I went out for an hour, and came back to a completely gray screen. I'm wondering if maybe it was due to heat, as now that it's cooled off a bit this evening it's completed two loops without any errors...

What I forgot to mention previously is that I've run this machine with each RAM module individually and the crashing/locking up behavior was identical. As it stands, it looks like I'll have to have someone with parts for a proper diagnosis check it out, since as far as I'm able to tell none of this is terribly conclusive. :mad:

EDIT:
The module that had generated errors has now also run 8 loops without a(nother) problem. At this point I'd say it has to be the motherboard or power supply.

eXtreme PSU Calculator estimates the requirements of the system at 538w. I have a 550w Antec Trio that I picked up, but I can't return that after opening it. If that doesn't work I'll have solved nothing, since it could still be a power issue (12w isn't much in the way of breathing room), and I'm out money for a part I don't need.

I'm not fond of builds that don't go smoothly.


*UPDATE*:
I'm updating this just in case it ever comes up in a search. It turned out to be a compound problem consisting of an nForce NCQ issue and bad RAM.

It just so happened that the head technician at the shop I took the machine to (where I worked way back when), had just seen a similar problem recently. It was a day or two in that instance before he even thought to look at an NCQ/nForce conflict, but after turning NCQ off the problem was resolved.

It worked with this system, to a point. The system stopped locking up, but continued encountering BSODs under load with either or both of of its RAM modules installed. Replacing the RAM allowed the system to run under load with complete stability for over 20 hours.

It's important to note that the RAM was never tested in another board, so I can't say for certain whether it's simply a defective kit or if there's some incompatibility with the board. Either way, though I have no grudge with G.Skill, I'll be ordering a different brand as a replacement.
 
Back
Top