Now really now guys....What's the deal with ECC/Registered Memory?

[lauxenburg]

Well, I've researched this topic time and time again and all I seem to ever see in a vague explanation here and there. I guess because just about 0% of consumers use ECC that it isn't worth throwing out there and people who know about it are professionals...

So, I am confused. A friend wants me to build a small server based off WHS for him to work as his small business/home server. When we venture into server motherboards, not only do I get even more confused and skeptical but they also become huge dual socket monsters with 20 slots I don't need nor do I want wasting space.

I really want to build a Mini-ITX system, slap in a SAS 6gb/s controller, four Seagate Cheetah 15.7ks, 2-4GB of RAM and a cheap CPU like an i3 or a Athlon.

But, with all of this talk on ECC and Registered memory, what's the big deal? Why do all enterprise servers come equipped with them? They say "data corruption". but I have been lacking ECC memory for the past 15 years and I haven't ran into any data corruption issues.

So, I am looking for an honest answer, and I am hoping you can help me.

Thanks.

 

[T_T]

Check this out... http://en.wikipedia.org/wiki/ECC_RAM#Errors_and_error_correction

The basic gist of ECC RAM is for correction from corruption. ECC support is generally more expensive, that is why a lot of manufacturers don't offer this in their non-server intended motherboards.

 

[lauxenburg]

Check this out... http://en.wikipedia.org/wiki/ECC_RAM#Errors_and_error_correction

The basic gist of ECC RAM is for correction from corruption. ECC support is generally more expensive, that is why a lot of manufacturers don't offer this in their non-server intended motherboards.

I understand they are "server oriented" but what is the real gain? What does it mean by data corruption? In the hard drive? I am pretty sure hard drives have ECC or else we would be screwed. ECC in the RAM makes the RAM last longer for extended operation? Maybe that?

Thanks.

 

[T_T]

I understand they are "server oriented" but what is the real gain?

Quoted from the link I posted...

"An ECC-capable memory controller as used in many modern PCs can typically detect and correct errors of a single bit per 64-bit "word" (the unit of bus transfer), and detect (but not correct) errors of two bits per 64-bit word. Some systems also 'scrub' the errors, by writing the corrected version back to memory. The BIOS in some computers, and operating systems such as Linux, allow counting of detected and corrected memory errors, in part to help identify failing memory modules before the problem becomes catastrophic."

The gain is the "correction" that happens with the ECC.

What does it mean by data corruption?

"Electrical or magnetic interference inside a computer system can cause a single bit of DRAM to spontaneously flip to the opposite state..."

In the hard drive? No. This is in reference to RAM. Corruption in the HDD is a different subject.

I am pretty sure hard drives have ECC or else we would be screwed. ECC in the RAM makes the RAM last longer for extended operation? Maybe that?

Thanks.

 

[lauxenburg]

Sorry for being stubborn. I DO appreciate all you have written. I assure this will be last post and then I select as Best Answer.

I know non-ECC is standard in desktops and even in the home server systems. So, would it be bad for a small server to use non-ECC?

Thanks.

 

[GhislainG]

It isn't necessarily bad; it all depends on how much RAM you'll install. ECC Registered memory is required when installing lots of memory on a server. I just installed a server that can use up to 192GB of RAM; that just can't be done with standard memory modules. You definitely want ECC when installing lots of RAM.

 

[EXT64]

My understanding also is that it is for most sensitive and constant running systems. Just by laws of nature there will be a random error in even a perfectly stable system. ECC would catch that error.

 

[GhislainG]

ECC would catch that error.

In addition, it would also correct it instead of possibly causing a crash or bad data.

 

[EXT64]

Yep, and the reason we all don't use ECC too is that error is very rare. It has some probability of occurring, but even if it does that does not necessarily mean a crash. You have to be accessing the area of the error to cause a crash, thus reducing the probability more. However, on a 24/7 high use server, the probabilities add up quickly.

 

[dkapke]

Just so I understand...you're wanting to build a WHS box...but with server hardware? You really don't need any of that. Your mini-ITX system with how ever many hard drives you need and an i3 is MORE than enough for WHS. A server board and ECC memory is complete overkill for a WHS box and won't do you a danged bit of good. My WHS box runs 24x7 and was a simple Sempron with 2Gb of memory and 3 1.5Tb hard drives. I will admit to upgrading it to a dual-core but that, it turns out, it pretty much worthless, too. That said, I use it only as a server and backup system - I don't stream music/movies from it. But that sounds like pretty much what you need anyways. Forget ECC/server stuff. WHS, being based on Server 2003, will take care of most issues you have in terms of memory.

 

[lauxenburg]

Just so I understand...you're wanting to build a WHS box...but with server hardware? You really don't need any of that. Your mini-ITX system with how ever many hard drives you need and an i3 is MORE than enough for WHS. A server board and ECC memory is complete overkill for a WHS box and won't do you a danged bit of good. My WHS box runs 24x7 and was a simple Sempron with 2Gb of memory and 3 1.5Tb hard drives. I will admit to upgrading it to a dual-core but that, it turns out, it pretty much worthless, too. That said, I use it only as a server and backup system - I don't stream music/movies from it. But that sounds like pretty much what you need anyways. Forget ECC/server stuff. WHS, being based on Server 2003, will take care of most issues you have in terms of memory.

Thanks for the clearing up to everyone and your nice reply.

I will attempt to build it using 2GB DDR3 a Pentium G6950 or i3 530 and 4 SAS drives.

 

[tcsenter]

ECC unbuffered is the same price as non-ECC, currently. Weirdly, Registered ECC DIMMs are actually less than unbuffered right now.

 

[pcy]

what's the big deal? Why do all enterprise servers come equipped with them? They say "data corruption". but I have been lacking ECC memory for the past 15 years and I haven't ran into any data corruption issues..

Cosmic or other radiation can induce the flip a single bit of memory. PROBABLY at a rate of around 1 bit per TB per year unless you like near a nuclear reactor or under a hole in the Ozone, then it would be a higher rate. So if you loaded a pattern of bits into a 1 TB of memory left it there for a year, 1 bit would flip. Which is not a big deal unless you are a Bank, or dealing with money. It really does depend on which bit is flipped.