Nforce 4 Verification

[Gizmotoy]

Hello all,

 

I'm just getting started building an UnRAID box to serve media to a Boxee Box.  Part of the appeal was being able to reuse hardware I had lying around for the time being, and upgrading things as necessary.  The unfortunate part is, all of my spare motherboards are Nforce4/AMD X2 combos.  I've read about the three issues they have, and the only one that really affects me is the data corruption problem.

 

I decided to go ahead and give it a shot, since all it was going to cost me was some time.  It's up and running with two drives, one data, one parity.  I'd like to test for the data corruption.  Most of the posts made it seem like it'd be obvious, and immediately apparent.  Some said it would happen under congestion.

 

Before I added the parity disk, I copied a 25GB Blu Ray rip I made, and the MD5 matched the source.  I repeated the test after adding the parity drive, and it still matched.  Finally, I started copying three 9GB files onto the array simultaneously, figuring that's about as taxing as things will ever get.  The md5s take awhile to calculate, but so far 1 of the 3 match.  I have a parity check running to see if it finds any errors there.

 

Is there anything else I can test to see if I'll have a problem?  I'm not completely against swapping out the mobo/processor, which I'll likely do within 6 months or so anyway for power savings, but since it appears to be working so far I might just stick with it if the tests pass.

 

Thanks for your help!

 

The parts are:

ABIT AN8 939 NVIDIA nForce4 ATX AMD Motherboard

Foxconn NF4UK8AA-8EKRS 939 NVIDIA nForce4 Ultra ATX AMD Motherboard

AMD Athlon 64 X2 3800+ Manchester 2.0GHz Socket 939 Dual-Core Processor

[Rajahal]

I did the same type of testing with another nforce 4 board.  Read through this thread and you'll see all the testing that was recommended to me:

 

Testing the ZOTAC GF6100-E-E - (NOT recommended by the wiki)

[Gizmotoy]

Well, of the 5 tests I ran, the last one failed.  I got a md5 mismatch in one of the 3 simultaneous file copies.  So, I ran a parity check and got errors there as well (no big surprise).

 

A little skeptical, I checked the system log and found a bunch of read errors from one of my drives.  I checked the SMART status of the drives.  It turns out, one is failing.  SMART reports the drive still passes, but it has a bunch of read errors logged.  :'(  Unfortunately, those are my only two scrap SATA drives to try.  So, with the hardware at hand, I can't tell for sure if the motherboard is OK.  I grabbed an IDE drive to give it a shot, but as soon as I booted it up with it connected, it started smoking!!

 

I think it's trying to tell me something.

[Rajahal]

Haha, wow.  At very least it is telling you that you need some new drives!

[vsonerud]

One of my two unRAID servers is using an Epox 9NPA+Ultra (NForce4 Ultra) with an AMD Athlon 64 X2 3800+ Socket 939 CPU.

 

The server has 8 drives, using 4 drives on the nForce4 Ultra SATA controller, and also has 3 Sil3132 PCI-E x1 controller cards - for a total of max 10 drives.

 

The server has been running for a couple of years or so, and I have transferred HUGE amounts of data to/from the server, including checksum verification, and I have NEVER ever seen a single error.

 

 

[Gizmotoy]

Haha, wow.  At very least it is telling you that you need some new drives!

 

Yeah, they're all ancient.  I haven't used the PC in forever.  I did manage to find an IDE drive that will work, and perhaps unsurprisingly, it's a Maxtor, the kind that the Nforce4 boards are supposed to have trouble with.  UnRAID recognized it just fine, though, so I'm rerunning the tests.  I'm not sure if one IDE and one SATA drive will create the bus issues that are theorized to be causing these errors, but I'm interested to see the results either way.

 

The server has been running for a couple of years or so, and I have transferred HUGE amounts of data to/from the server, including checksum verification, and I have NEVER ever seen a single error.

 

That's good to hear.  I'll keep at it.  My gut feeling is that the board appears to be fine and I'm just having drive issues, but I'll have to see how it goes.  The new drives I ordered need to come faster.

 

Edit:

As a follow-up here, the new configuration passes every test I can throw at it.  I filled up the drives to 90% full, and all the md5s passed.  No parity errors reported.  I guess now all I can do is wait and try the tests again with the new drives.

[Gizmotoy]

Just a follow-up note that I've been running this thing through its paces and haven't had any trouble yet.

 

One note, as I opened the thing up to put new drives in, I realized I swapped around my motherboards and had forgotten about it.  I'm actually using a Foxconn NF4UK8AA-8EKRS.  It's quite a bit newer, and was probably among the last 939 NF4s produced.  I remember buying it when another NF4 chipset motherboard failed and I had to get the machine back up and running quickly in Windows, which hates it when you change chipsets.  It's from 11/2007, where my original Abit board was from 2005.

 

Given vsonerud's similar experience above, maybe the NF4 Ultras are OK.  Or at least a few of them are.

 

I copied over 750 GB of data, and ran md5deep on both ends.  I only got mismatches on two Mac OS DS_Store files that my desktop probably touched in the meantime.  I navigated to that directory in the Finder, so it's plausable.  All of the real data was fine.

 

I'm preclearing the parity drive now, and I should be inserting that into the array tonight.  Hopefully all goes well.