Check if ECC is enabled (Windows 7)

[Vandroiy]

I got myself some ECC RAM now. So, I booted into memtest86+. First, it showed DDR-400 Non-ECC (What the ***?), on the second boot DDR-1333 ECC (yep, that's what I set it to).

So I'm like, WTF memtest86+, make up your mind? Oh, whatever, I'll just boot Windows and use CPU-Z. But CPU-Z shows nothing. The SPD->Correction field is just grayed out as if it was non-ECC.

Is ECC off now? On? How do I get my ECC correction count?

Help!

 

[Leaps-from-Shadows]

MemTest doesn't always get the RAM type or speed correct. You can ignore that.

What hardware do you have? Does your board even support ECC? Most consumer-level boards don't -- it's really just meant for servers.

 

[jaquith]

The CPU-z 'SPD' tab should show your error correction; keep in mind it just retrieves the JEDEC information. In order to 'use' ECC RAM both your MOBO and CPU must support ECC; e.g. Intel Xeon and AMD Opteron. Otherwise you can get oddball Memstest86+ readings on different boots, also I know the Memtest supports some chipsets and not others - this ECC bug has come up before.

The ECC is very limited with Memtest -> http://www.memtest86.com/tech.html

 

[Vandroiy]

MemTest doesn't always get the RAM type or speed correct. You can ignore that.

What hardware do you have? (...)

Good to hear that it's probably memtest failing sometimes, not ECC being switched off at random.

I use a Phenom II 965 on a Sabertooth 990FX. By official on-line specs, both support dual channel DDR3-1333 ECC as well as non-ECC RAM, as long as it's unbuffered. The manual doesn't say anything about ECC, but there's a setting in UEFI, too. If ECC is not supported, my board or CPU is broken, I care little whether by design or not.

But what I really need is a way to check whether ECC works from Windows, and get the ECC error count! The memory error detection feature goes to waste if I can't see the results!

There's also a temperature sensor on there, but I can't see the data. Blind diagnostics!

 

[Vandroiy]

The CPU-z 'SPD' tab should show your error correction; keep in mind it just retrieves the JEDEC information. In order to 'use' ECC RAM both your MOBO and CPU must support ECC; e.g. Intel Xeon and AMD Opteron. Otherwise you can get oddball Memstest86+ readings on different boots, also I know the Memtest supports some chipsets and not others - this ECC bug has come up before.

The CPU-Z 'SPD' tab is just grey on the error correction stuff. I don't believe that means "off", it rather looks like "lol no data" to me.

What do you mean with "this ECC bug has come up before"? ECC not showing even though it is enabled? As I said, often memtest86+ shows ECC just fine. I can enable and disable it for the testing, and it looks normal. But as the RAM doesn't produce any errors ( ), I can't really tell when/if ECC does anything or whether Windows would report that.

 

[Leaps-from-Shadows]

Is this what you're looking for? Windows Hardware Error Architecture

The errors are probably logged in Event Viewer somewhere.

 

[Vandroiy]

Is this what you're looking for? Windows Hardware Error Architecture

The errors are probably logged in Event Viewer somewhere.

If ECC is on, there should be an event for each error. I'm still browsing that site, but I don't see where I can check that ECC is on and has not produced an error.

Sure, I could just go on and assume my RAM is perfect AND soft-error proof now. But what if in reality ECC isn't even turned on and it's producing errors without me knowing?

 

[jaquith]

In order for Memtest to work properly the chipset must be supported otherwise ECC is ignored in its testing.

Memory Configuration
ECC Mode [Enabled]

My expertise is with Registered/Buffered ECC memory {RDIMM} on servers.

ECC:
Event Viewer http://msdn.microsoft.com/en-us/library/ff559390(v=vs.85).aspx

IMO - contact ASUS and get them to confirm that actual correction is taking place.

 

[Vandroiy]

running memtest86+ a few more times, it showed ECC capability every time.

I suppose when it showed DDR-400 non-ECC, that was just the program going haywire for some reason. I can't imagine my board underclocking my memory to DDR-400 anyways.

Also, within the .txt report of CPU-Z, I see the following entry:

DMI Physical Memory Array
location: Motherboard
usage: System Memory
correction: Multi-bit ECC
max capacity: 32768 MBytes
max# of devices: 4

It appears as if the entry is showing some combination of board and controller capabilities and the current mode the RAM is in... not very clear, but looking good.

Which leaves me with the point that I lack a neat GUI showing RAM state and corrected/detected ECC errors.

 

[jaquith]

Run Memtest for 24~48 hours then you'll know that ECC is working properly. You should have NO Errors. Based upon the info above it 'seems' ECC is recognized per CPU-z and as I stated above.

 

[wwest_32]

Run Memtest for 24~48 hours then you'll know that ECC is working properly. You should have NO Errors. Based upon the info above it 'seems' ECC is recognized per CPU-z and as I stated above.

That implies that with non-ECC I could expect an undected/ubndetectabe memory error within 48 hours?

How long before the disk data get corrupted?