Hardware Gurus - Lend me your brains (Machine Freezing)

seamus

Right, I'll try keep this short, but an input is much appreciated.

I built a machine back in March to be used as a small office server.

Abit NF-M2S Motherboard
2GB Corsair Value Select RAM
2 x 232GB SATA Seagate HDDs
Windows Server 2003

The drives are configured using onboard RAID in a mirror config.

About a month after building (though it may have been sooner), the server started falling off the network. Because this was more a nixer, it took forever to get back and forth to look at the machine. I assumed it was probably someone in the office doing something weird like kicking a plug or a network cable. Anytime I went out, the machine was working fine. Then it came to the end of August. It was still happening, but I hadn't really been called because of holidays and various other things. It had been steadily getting worse - 3 to 4 times a day, which would require hard reboots of the server to correct.

So I went out and the machine had crashed. All the fans on, but no OS interaction whatsoever. Reboot and I get back in. But it keeps crashing - complete freeze, the screen just stays on whatever it last showed, and there's no OS or HDD activity. I've been messing with the BIOS and it's getting worse, barely getting to log in and it freezes again.

The freezing isn't so regular if I'm not logged in at the console. If it's only dealing with network/disk traffic, it may stay alive for a couple of hours.

So last night I had come to the conclusion that is must be a screwed motherboard. I had tested and scanned both disks independently, and they're solid. If the processor was screwed, I probably wouldn't even get past the POST. If the RAM was boned, it would probably bluescreen. But I checked the abit forums on the off chance, took a quick look at a huge topic about audio drivers, and a number of people mentioned freezing. Turned out that all of their freezing problems were sorted by setting the RAM timings/voltage manually in BIOS. Naturally I had left the timings to auto.

Does this sound like a likely cause of my problems? It's kind of clicking with me - the freezing seems to get worse if I do something which requires more memory usage, i.e. logging in at the console. But why wouldn't it have appeared when I first built the machine?

Can anyone think of anything else? Cheers!

Khannie

I'd run memtest on it initially and turn off any on-board hardware that's not being used (sound card, etc.). After that, I'd try getting updated NIC drivers (gigabit NIC's have had some dodgy drivers thrown together). After that I'd try getting a PCI NIC and see if that sorts it.

Lots of hard reboots don't sit well with windows either. It's my experience that if you hard reboot windows enough times, it just slowly starts going to sh*t.

Stephen

I had a similar problem with a different Abit board (AB9 Pro) with 2 gigs of Corsair RAM - a BIOS update fixed it. It seemed to be reading the wrong timings from the RAM, and running it too aggressively, causing the machine to go bananas. Its been perfect ever since the BIOS ugprade.

seamus

Khannie wrote:

I'd run memtest on it initially and turn off any on-board hardware that's not being used (sound card, etc.). After that, I'd try getting updated NIC drivers (gigabit NIC's have had some dodgy drivers thrown together). After that I'd try getting a PCI NIC and see if that sorts it.

Yeah, initially when I'd use the onboard video, the screen would crash (go all scrambled), so I assumed that it was a video driver issue. So I plugged in a PCIE card and the screen stays nice and clear while it crashes
So I'm guessing now that video isn't a problem - the scrambling was just a symptom of something else.

Now that I think about that, I built a desktop machine at the same time with the same board, and 1Gig of the same type of RAM. I saw the exact same issue with the video on that one, which spontaneously went away. So perhaps this could be a RAM issue (fingers crossed!), or even a driver issue after all.

Lots of hard reboots don't sit well with windows either. It's my experience that if you hard reboot windows enough times, it just slowly starts going to sh*t.

Indeed
I had pleaded with him at the time to not reboot the server (thinking it was a network issue), but it's kind of unavoidable now.

I've downloaded any updated drivers for the board/BIOS, along with memtest and Hot Cpu.

Stephen, your story gives me hope

Cheers

Anti

I would reccomend you update all drivers, sound, video, network etc etc. And go with the latest bios available. Also set all memory timings yourself and voltage, just go one step above the standard. Dont worry, this wont harm the ram at all. Also, check the temps of the cpu. Install speedfan, and monitor the temps. Keep a eye out for anything over 60c. And yeah, run memtest too. 24 hours running should do it, so do it at a weekend to minimize downtime.

seamus

Heh, there's already downtime. It's sitting on my kitchen table. Having a crashing server is less productive than no server - I moved the data onto one of the desktops.

The main problem is that I would love to be able to run utilities, but it rarely stays alive long enough to install/uninstall anything. So bootable/command-line stuff like memtest are really all I have the patience to use.

I'm going to disconnect the disks and run Knoppix too. If Knoppix crashes, I know it's not a driver issue.

The BIOS has inbuilt temp. gauges (not sure how accurate they are?) which show the CPU idling at around 30c. I've also told the BIOS to reboot the machine at 70c, and that hasn't kicked in yet. So I'm doubtful about temp being an issue.

Anti

Have you tried another power supply? As unstable voltage on the rails can cause the machine to lock up too.

seamus

If it comes to that, I'll give that a go. Cheers

seamus

After all that, it would appear that in fact there's something wrong with one of the disks or the SATA controller :rolleyes:

I'm guessing the disk is screwed. I reinstalled Windows onto a single disk, then as soon as I plugged in the second disk and started reading data from it, crasharoo...

I'm running seagate's own diagnostic on it now to see if it's the disk or the controller.

Cheers for the help guys.

Khannie

Ah sure at least you have raid 1 love on it.

seamus

Yeah having an exact copy of the disk comes in handy

Left seatools on to scan both disks overnight. Nothing

So it only occurs when moving large amounts of data between the disks. I disabled write caching and kicked off software mirroring again before I left for work. Fingers crossed, but I'm doubtful.

seamus

Disabling write caching worked.

To say I'm surprised is an understatement. But I'm happy I've found the cause. But there's a performance hit, so I need to do a little more investigation so see if I can find a way to turn it back on. I was under the impression that windows used the disk cache for write caching, but seagate don't provide drivers for their disks. So maybe Windows uses part of RAM for write caching too...

Thanks for the responses guys.