ECC questions

[AfterBurn]

First of all, I know what ECC is and does, but I'd like some explanations on ECC settings. I have the following settings in my BIOS:

EC (Error Checking)
ECC (Error Checking & Correcting)
ECC SCRUB

The difference between the first and latter two is obvious, but the difference between the second and last are not so obvious to me. They both check, they both report, they both correct. So what does scrub do extra that regular ECC does not?

Further more, I have something called SERR, which seems to be related to ECC as its in the same submenu of my BIOS. It can be set to "None", "Single bit", "Multiple bits" and "Both". What's this for and what exactly does it do?

Manuals are not exactly much help on these issues and websearches don't reveal much more than explain what error correction is and which errors can occur.


Tia,
-AB-

 

[pm]

When a high energy radiation particle hits a section of memory, it normally nukes several bits in a circular region. Since memory is organized so that the same line in memory is spread across multiple sections of the memory chip and several other memory chips, you should only get a bit or two that is actually flipped. But the fact that one or two are flipped usually indicates that other bits are probably hosed too. And if you wait long enough, the chance that another particle will hit near those other bits and create more unrecoverable errors increases too.
So a memory scrub re-reads and writes all of the contents back to disk so that everything is refreshed (and thus correct) within memory.

The SERR setting is "Soft Error Rate Reporting". It's saying when do you want the error to be reported to the OS when one bit is bad, or when two bits are bad (or never report). I'd set it to both mostly because it's neat to see that ECC is doing something. It will pull an NMI and if the OS has the right hooks, then it will write this to a file. If the OS or a background program aren't monitoring for this event, it will get lost.

Nice BIOS options. Which board is it?

 

[AnoTech]

Just a few thoughts... from my understanding:

Power PCI chip, when ECC is enabled will detect correctable errors...on reads of memory and pass corrected data on...

Memory Scrubbing is a feature by which the Power PCI chip, without processor intervention, executes a write to also correct the error
in external memory....

I believe their should be both Active and Passive...

I am assuming though, that ECC Scrub... is the same as ECC, Memory Scrubbing

~ano

 

[AfterBurn]

Its the Tyan Thunder K7 and it has a Phoenix ServerBIOS. Gotta tell ya, love this board, its speed and its stability.

Thanks both of you for taking the time to explain what it actually does. I like knowing details of things, not just the superficial stuff.