ECC RAM underreporting vs an identical computer, identify/remove damaged RAM or leave it?

Nicole Rachelle

Reputable
Mar 20, 2014
10
0
4,510
Hi,

I need to prep two Dell Precision WorkStation 490s as servers at work. They're old, but they have each have two awesome Xeon dual-core processors. One computer works great, but the other is a little hinkey. We kinda need to end up with two servers so I need to make the hinkey one work.

They have ECC DDR2 RAM, which I know very little about. One computer reports 8x512MB as 3GB of available RAM (losing the rest to error-correction bits, I presume).

The other (identical) computer reports 8x512MB for 2.75GB of available RAM. I assume this means there are RAM errors and some of the memory addresses have been marked bad? Is this a correct assumption? (This computer had the first computer's video card in it since it came to us without a video card. I *think* all the other hardware was exactly the same.)

Is it better for me to run that server as-is and count on the ECC to resolve any RAM errors reliably, or is it better for me to try to isolate which stick(s) be all underreportin' up in there and remove the affected stick(s) from the computer?

I'm not sure if they have to be in pairs, but if I removed two sticks and ran 6x512MB I think that would be enough RAM for our applications. Can the server be run with 6 sticks of RAM?

We're no-budget, a small theater. So purchasing expensive RAM to get this computer working is not an option.

I'm such a noob at ECC, so if there are any links you can refer me to, I'd appreciate that as well as your opinions. I tried teh Googlez, but didn't find enough information to know what to do in this situation.

Thanks!
~Nicole
 
Solution
Nicole,

It sounds like you're expecting to see 4GB of RAM on these machines. How are these systems reporting the ram? Is the report through an operating system utility after boot? Or through the machine BIOS?

If the report is via an operating system, there are a number of configuration options in the both the OS and BIOS that could explain this without indicating a RAM failure. Most 32 bit operating system won't recognize a full 4GB of RAM.

Another possible explanation could be in various BIOS config options. Without botching the explanation, if you have a 32bit CPU or the BIOS is configured for a 32bit OS then there may BIOS options mapping some of your ram for a video I/O or other onboard host adapters using MMIO. This...
Nicole,

It sounds like you're expecting to see 4GB of RAM on these machines. How are these systems reporting the ram? Is the report through an operating system utility after boot? Or through the machine BIOS?

If the report is via an operating system, there are a number of configuration options in the both the OS and BIOS that could explain this without indicating a RAM failure. Most 32 bit operating system won't recognize a full 4GB of RAM.

Another possible explanation could be in various BIOS config options. Without botching the explanation, if you have a 32bit CPU or the BIOS is configured for a 32bit OS then there may BIOS options mapping some of your ram for a video I/O or other onboard host adapters using MMIO. This can take up a significant amount of ram. Based on your reports of 2.75GB and 3GB, I'd say this is likely. If you can, check out the specs for those machines. I think you'll find that this is normal. Probably nothing wrong at all with your hardware.
 
Solution
It bothers me that with the same type & number of RAM chips, exact same physical video card and the same physical Windows XP hard drive installed, the two machines report different amounts of RAM available.

I'll go through them and list what the BIOS reports versus what Windows XP reports. I have to run downstairs to bus tables (intermission) then I'll be back up here on the second floor of our theater working on teh computerz and can get more accurate information for you. Thanks for responding!
 


Yep, based on what you've reported so far, I'm willing to bet that if you browse the bios settings, you'll probably find that one of the machines has a different amount of ram dedicated to Video MMIO or something to that effect.
 
ok . . . . . the BIOS reports the exact same amount of RAM in each of the two computers, 8x512MB = 3GB. I presume the other 1GB is eaten by ECC error-correction. I could be wrong about that, since I don't know how ECC works.

One computer reports 3GB in Windows. The other reports 2.75GB in Windows. That's what concerns me. Why does one report less RAM? Has it mapped around "bad sectors" of RAM like a hard drive would map around bad sectors?

I physically move the video card, hard drive, & CDRom drive between the machines; I can only have one up at a time--one wasn't complete. The incomplete one is the one that reports less RAM in Windows. Maybe they scuttled it for parts because it was acting up?

I guess I could buy Memtest86 Version 5 for $40 with ECC support . . . . . but I'm trying to keep the budget down for the theater.

The RAM chip I read says Hynix 512MB 1Rx8 PC2-5300F-555-11 HYMP564F72CP5N3-Y5-AC-C-0731

Is this special ECC RAM or can we get any PC2-5300 or DDR2 and stick that in? I don't know if the ECC is done on the motherboard or on the RAM chips.

I just talked to one of the two techies I report to at the theater, and we definitely need two good servers. I have to fix the wonky one or we'll have to find a replacement somehow.

I appreciate any help you can provide, even mundane details or links to documents that explain ECC would be helpful. As I said I tried Google but I haven't found the information I need.
 
Apparently this is Fully Buffered or FB DIMM RAM . . . .

http://en.wikipedia.org/wiki/Fully_Buffered_DIMM

http://www.geeks.com/details.asp?invtid=PRECISION-490-NOOS-6R

ooh I found RAM cheap!!!!! So if it turns out this is actually bad RAM I can do something about it. http://www.ebay.com/bhp/dell-precision-490-ram

Maybe next time I'm here at work I should switch all the RAM from one computer to the other. If it's a BIOS setting the "underreported" RAM value will stay with the computer. If it's bad RAM the underreported RAM value with move with the RAM . . . . .
 
I'll check in later, gotta go home! I might be gone for a couple days, and definitely won't be back in the theater until next week. This upgrade is going more slowly than I had hoped but oh well. Thanks agin for your help CEH3 :)
 


You're doing fine. You're a natural trouble shooter. Before you spend your hard earned money try the following search term: "memory map win32 4GB system," you'll see what I mean when I say the results are to be expected on those systems. .. And, I think you can get a free version of Memtest that will verify all 4GB of ram is OK on both machines. Look for the "Ultimate Boot Disk" for Memtest and a bunch of other handy utilities.

What do you want to do with this hardware?
 
hi CEH3!

The servers will run Windows 7 or Windows Small Business Server 2008 depending on what Mark & Chad decide. We may start using domains. Currently one server supports our accounting & financial software (quickbooks??) and the other might be some phone software. I'm a little fuzzy on that part of this 😉 I'm just supposed to test out the hardware & make sure it's ready.

I am aware of the 4GB limit on RAM in Windows XP, but that doesn't seem to come into play in this case. They report 3GB in the BIOS, but one reports 2.75GB in Windows whereas the other reports 3GB in windows. the hardwarwe is identical. So I was just concerned to know if the RAM that's reporting 2.75 might be damaged?

Memtest86 free version doesn't support ECC, so if there are RAM errors being corrected it will just show no errors.

If the machine would be super-reliable even with damaged RAM since it's fully buffered, that would be fine with me. I just don't know enough to be able to guarantee to Chad & Mark that it's 100%.
 
This is solved.

I switched the RAM from one computer to the other and the problem stayed with the motherboard, not with the sticks of RAM.

I discovered the computer reporting less RAM has a reeeaaallllyyyy old BIOS.

I'm flashing it. Actually I'm flashing them both because even the newer one was out of date. I shoulda checked that first!