What does the Western Digital Diagnostic Prove?

[BonzaiDuck]

With this thread, I'm looking for insight and consensus just to confirm what I'm thinking.

I have this old LGA-775 system with an E8400, GA-EP45-UD3R motherboard and G.SKILL DDR2-800 RAM. It had functioned well as a WHS v.1 server for about three years from 2009 to 2012. In 2012, I got a WHS-2011 install-disc set, and built a "new" server out of another LGA_775 mobo. The Gigabyte board was redeployed for a Visual Studio platform, and then in 2013 became a "business" system for condominium real-estate business. I added another kit of G.SKILL, but didn't think to test the new sticks. Turns out, the new kit was failing.

Early this year, it began to show evidence of disk corruption. I extracted the bad sticks and RMA'd them. I also did an "upgrade" install of the OS.

Recently, more troubles began to surface. I concluded either of two things: the motherboard was failing, or the hard disk (a WD Blue) was failing. I ran the WD Diagnostics for Windows. The "Quick" Test "quickly" turned up a read verify error.

Long and short of it, I determined to decommission the machine and ship the parts to a friend, making sure he understood what had happened.

So I took it off my LAN. I decided last night to do one more test -- this time with a bootable USB "DOS" version of the diagnostic.

The diagnostic gave it a full pass on "Extended" test. I am concluding from this two or three things:

1) There's nothing wrong with the board and its onboard controller.
2) There's nothing wrong with the disk.
3) The problem is more likely OS corruption which survived through an "upgrade" install, or possibly corruption of installed software.

I'm going to take the system apart and ship the board, RAM and CPU to my friend. Do my conclusions seem correct?

 

[Cerb]

All such diagnostics, from their tests, to simple SMART readings, can only positively prove a problem exists. A problem may still exist without being detected, if the testing comes back OK. They cannot prove the absence of a problem.

The rest is a judgment call, of the kind I hate making, when there isn't a clear cause.

 

[BonzaiDuck]

All such diagnostics, from their tests, to simple SMART readings, can only positively prove a problem exists. A problem may still exist without being detected, if the testing comes back OK. They cannot prove the absence of a problem.

The rest is a judgment call, of the kind I hate making, when there isn't a clear cause.

All I have is the history of troubles I experienced early this year with the bad RAM sticks, the troubles I endured with the "upgrade install" to fix OS or other corruption.

My friend on the East Coast seems to think that the "DOS-bootable" version of the WD diagnostic is more reliable than the Windows version. He's not software-savvy enough to explain "why." Of course, I would intuitively think the same thing -- another reason why I run HCI-Memtest-64 in the CD-bootable version as opposed to several instances running under Windows.

The hardware problem that was verified earlier this year was the defective 2x2GB G.SKILL RAM kit. At this point, I can't verify anything about the motherboard, but the WD Diagnostic would seem to fail under either of two conditions: bad hard disk, or failing controller (so I would think, anyway).

How would you otherwise test an onboard controller, anyway?!

 

[bryanl]

It's possible the drive cable simply wasn't seated properly.

It seems most diagnostics, including the drive's own long SMART test, allow several attempts to read each sector, perhaps 10, before they flag an error. One diagnostic, MHDD, available on The Ultimate Boot CD, will report retry counts lower than that.

I have no experience with the direct booting version of HCI Memtest but would prefer to use something better known, such as Memtest86+. However G.Skill uses the latter and proves that it isn't foolproof.

 

[BonzaiDuck]

It's possible the drive cable simply wasn't seated properly.

It seems most diagnostics, including the drive's own long SMART test, allow several attempts to read each sector, perhaps 10, before they flag an error. One diagnostic, MHDD, available on The Ultimate Boot CD, will report retry counts lower than that.

I have no experience with the direct booting version of HCI Memtest but would prefer to use something better known, such as Memtest86+. However G.Skill uses the latter and proves that it isn't foolproof.

Geez! You and another respondent make me wonder if it's ever really possible to know for sure that our hardware works properly!

I don't think it was the cable, because it would've turned up in the DOS-version test sequence. The only thing I'd known for certain: The second pair of G.SKILLs were proven defective, and I had indications before the proof of hard-disk corruption as it appeared in the Event Logs.

I suppose I just want to be sure that the WD Blue drive can be re-deployed. I've warned the future recipient of the board/CPU/RAM of my concerns and suspicions, but I still had to tell him I think the board is good. He gets it for free, anyway, as well as the processor and good RAM.