Error Correction RAM

[Luba]

I’d guessing that 99% of computers out there don’t use ECC (error correction) RAM and they all seem to do fine, work fine. What happens when an error gets into RAM and is NOT corrected?

What are some situations were you would need/want a computer with ECC RAM?

 

[ActionableMango]

What happens when an error gets into RAM and is NOT corrected?

What are some situations were you would need/want a computer with ECC RAM?

A "0" is stored instead of a "1", or vice-versa. This is a more common occurrence than you would think, but it is not noticeable because most data is not critical to most people in most use cases.

Important situations would be anything where the integrity of your data is mission-critical. For example scientific research, financial transactions, and any kind of critical record-keeping.

A secondary benefit to error correction is notification. Because ECC RAM logs error corrections, when a RAM stick starts going bad you'll see the number of error corrections skyrocket and you'll know it's time to replace the RAM.

I did a poll here in the MP forum about ECC ram.

 

[msh]

A "0" is stored instead of a "1", or vice-versa. This is a more common occurrence than you would think, but it is not noticeable because most data is not critical to most people in most use cases.

Important situations would be anything where the integrity of your data is mission-critical. For example scientific research, financial transactions, and any kind of critical record-keeping.

A secondary benefit to error correction is notification. Because ECC RAM logs error corrections, when a RAM stick starts going bad you'll see the number of error corrections skyrocket and you'll know it's time to replace the RAM.

I did a poll here in the MP forum about ECC ram.

Also won’t errors that ecc ram cannot correct stop the system thereby preventing further corruption from being written to disk?

 

[AidenShaw]

Also won’t errors that ecc ram cannot correct stop the system thereby preventing further corruption from being written to disk?

Mostly. Historically, ECC has been able to correct single bit errors (one bit wrong in a 8-byte cell). It will detect all double bit errors (and throw an "uncorrectable memory error" kernel exception that typically causes a bugcheck or panic). Most three bit (or more) errors are detected, but some edge cases can be missed. Some newer ECC systems can correct some kinds of multi-bit errors

In a production environment, one of the best things about ECC memory is that you never need to worry that a crash or other problem is due to bad RAM. If your RAM is failing, you'll get a hard crash with "uncorrectable memory error". If a program dies or other panics happen - look at your drivers or other apps.

 

[msh]

Mostly. Historically, ECC has been able to correct single bit errors (one bit wrong in a 8-byte cell). It will detect all double bit errors (and throw an "uncorrectable memory error" kernel exception that typically causes a bugcheck or panic). Most three bit (or more) errors are detected, but some edge cases can be missed. Some newer ECC systems can correct some kinds of multi-bit errors

In a production environment, one of the best things about ECC memory is that you never need to worry that a crash or other problem is due to bad RAM. If your RAM is failing, you'll get a hard crash with "uncorrectable memory error". If a program dies or other panics happen - look at your drivers or other apps.

Thanks for the details on how it works. Hasn’t happened to me yet.

PS: Love your sig quote. I have added my own just for today.