Homebuilt PC locks up randomly

[muffinman]

AMD Athlon64 x2 3800+
1GB OCZ Memory (ECC/Reg)
XFX 7800GT factory overclocked @ 450mhz
ECS nForce-A939
WD 36.4GB 10k RPM SATA
Maxtor 200GB EIDE
Lite-on DVD burner
Antec P180 case (three 120mm fans)
Antec True Power 430w

Computer freezes up seemingly at random. Might run for 2 hours, might run for 10 minutes. The 36GB hard drive does get pretty hot, although the ambient temps inside the case seem extremely cool. Running the latest nVidia drivers.

Not even sure where I should start to try and diagnose this problem?

 

[linux_0]

http://memtest86.com/memtest86-3.2.iso.zip

download the 3MB ISO burn onto CDR, boot up and let is test your RAM for about 4 passes.

You can also try downloading the 700MB Knoppix Live Linux CD which has memtest86 already on it and 1200 software packages.

http://www.kernel.org/pub/dist/knoppix/KNOPPIX_V4.0.2CD-2005-09-23-EN.iso

With Knoppix you can torture test your system to see to find out what the problem is.

Also check to make sure your fans are spinning and your CPU and motherboard are not overheating (check the temps and fan speeds in the BIOS).

Semper Fi carry on

 

[muffinman]

I'll definitely give memtest a shot. I just rolled back my nVidia driver to the latest one on the XFX website (which was slightly older than the latest nvidia). I'm really hoping that is/was the problem.

I appreciate your response.

 

[linux_0]

Let us know what memtest86 reports.

You may also want to check your temperatures and voltages in your BIOS.

What kind of ECS motherboard do you have?

ECS boards are known to have issues. The ATX PWR connector on mine melted.

Thanks!

 

[hella-d]

10,000RPM Do Run A Bit "Warm" So Thats Not The Culprit (Not Likely) But Did You Say Your RAM Was Registered/ECC RAM?

 

[linux_0]

Ummm..... virtually all socket 939 boards cannot use registered ECC RAM.

You have to use Unbuffered 184pin DDR SDRAM.

Only socket 940 boards use registered ECC.

 

[hella-d]

LOL, I Already Use Firefox

 

[linux_0]

 

[hella-d]

Im Running Solaris In VMware

 

[linux_0]

I'm running GNU / Linux on AMD64

 

[muffinman]

What kind of ECS motherboard do you have?

Its an nForce4-A939

10,000RPM Do Run A Bit "Warm" So Thats Not The Culprit (Not Likely) But Did You Say Your RAM Was Registered/ECC RAM?

Correct - Registered/ECC RAM (Ocz 1gb).

Ummm..... virtually all socket 939 boards cannot use registered ECC RAM.

You have to use Unbuffered 184pin DDR SDRAM.

Only socket 940 boards use registered ECC.

I thought only socket 940 boards *required* it - but it would still work with 939 boards/cpus. Is this incorrect?

I'm running GNU / Linux on AMD64

fc4 x86_64 on my a64 averatec laptop, but just windows xp on this one.

...Well at least for another week until windows drives me insane and I install something else

EDIT: pc has been humming along for about 2.5 hours so ar. This may be a new record. Which is great, and somehow sad at the same time

 

[linux_0]

I checked the ECS website and the manual for this board and as far as I can determine it only officially supports unbuffered 184pin DDR.

http://www.ecs.com.tw/ECSWeb/Products/ProductsDetail.aspx?MenuID=0&LanID=9&DetailID=493&DetailName=Specification

I'm not sure how it even works with REG ECC most motherboards won't take it.



Please see here:

http://www.ocztechnology.com/products/memory/ocz_pc_3200_ecc_registered_dual_channel_ddr

* This product has been specifically designed for use o­n the AMD Athlon64 FX Socket 940 platform and should not be used o­n any other desktop platform. AMD Athlon XP and Intel Pentium 4 users should not purchase this memory for use o­n their systems. This product does not work with Socket 939-based mainboards, as they are not compatible with Registered memory.

 

[muffinman]

Oh man if it was just the ram, that would make my day. I'd gladly replace it. A gig of unbuffered memory costs what, $75 or so?

If I continue to have issues, I'll go ahead and pickup some unbuffered ram and give that a shot.

Here's a link to the RAM:
http://www.monarchcomputer.com/Merchant2/merchant.mv?Screen=PROD&Store_Code=M&Product_Code=140526

Definitely REG/ECC. Detects perfectly fine during POST too.

EDIT: Anyone want to recommend some RAM? Preferrably with a part number

 

[linux_0]

http://www.ocztechnology.com/products/memory/tyan_qualified-br_ocz_ddrpc_3200_server_series

If this really is what you have then yes it IS REG ECC and it is not supposed to work with your ECS 939 board.

This is intended for High end Dual Opteron Tyan motherboards.

That's why the DIMM has 9 chips instead of 8 and 2 small ICs in the middle.

The 9th Chip holds the parity info.

NON registered NON-ECC RAM typically has 8 chips (NOT 9) and no small ICs in the middle.



Have you had a chance to try memtest86? My guess is it will fail if you run it.

 

[muffinman]

Oddly enough, PC is still running. Ran all last night and still humming along. Almost 24 hours without a hiccup. If I have any more problems I'll just pickup some unbuffered RAM.

How do I determine the correct memory timings for my board/cpu?

 

[linux_0]

Have you had a chance to try memtest?

http://memtest86.com/memtest86-3.2.iso.zip

or Knoppix?

http://www.kernel.org/pub/dist/knoppix/KNOPPIX_V4.0.2CD-2005-09-23-EN.iso

memtest and Knoppix can probably help you diagnose the system better.