Why isn't ECC memory used more?

[taltamir]

Capitalism is a concept that is not and cannot be implemented due to its simplistic approach of human behavior - that much is clear.

This is nonsense.
You can't make a false and completely baseless statement without a single shred of justification for it and say "that much is clear"... how is it clear? why is it clear? why is it even true? (it isn't true)

Capitalism is generally considered by scholars to be an economic system that includes private ownership of the means of production, creation of goods or services for profit or income, the accumulation of capital, competitive markets, voluntary exchange, and wage labor.

The issues are that "competitive markets" and "voluntary exchange" are suppressed in both the "State monopoly capitalism" (the problem) and the "socialism/communism" (the supposed solution).

Marx believed the solution to the encroachment of government on private liberty is to change the government from "greedy rich oligarchs" to "paragons of the people" and then give them even MORE power. This is wrong, stupid, and never works. The problem is NOT Capitalism, the problem is the LACK of capitalism.

However what happens currently is direct consequence of adopting the idea that every being searching for their own profit will result in a valid working system.

No it isn't, this would never have happened without the broken system of government mandated monopoly aka patents. Which are done with the intent of furthering society as a whole at the cost of individual pursuit of wealth. Yet actually holds society back while promoting the pursuit of wealth of a few individuals who exploit it to keep the rest down (which is most certainly not what capitalism is).

You know, at first I was worrying we are getting off topic... but the original question is "Why isn't ECC memory used more?" and the only answer is "politics & economics". The question is not a technical one and should probably have gone in the social forums.
The again, the author might have not realized that.

 

[anikhtos]

lol i think this thread is de railed
leave politics out
communism capitalism
who cares
just keep the new processors coming lol

 

[taltamir]

lol i think this thread is de railed

I disagree, as I have said before, this thread is an economical and the politics of economics question. It might easily be incorrectly classified as a technical question but it isn't.

You know, at first I was worrying we are getting off topic... but the original question is "Why isn't ECC memory used more?" and the only answer is "politics & economics". The question is not a technical one and should probably have gone in the social forums.
The again, the author might have not realized that.

Perhaps we should have an OP move this to social forum?

 

[anikhtos]

I disagree, as I have said before, this thread is an economical and the politics of economics question. It might easily be incorrectly classified as a technical question but it isn't.


Perhaps we should have an OP move this to social forum?

well first we had peopel arguing if ecc would made eny diferense to the sytem
now ti seems that we accepted that having ecc is better than non ecc
so now we debating why they do not implemate it?
the answer
because intel can go with that
it can simple disable ecc function from their chips so they can sell the more expensive platfomrs in a higher premium to earn more.
as for that
i3 is the only sandy bridge processor that supports ecc if compined with a server chipset c202 c204 c206 as an examble.

which states that all the sandy bridge line is the same
well given than e3 xeon line and the commercial line must be the same
simple disabling features or enalbing them

some can say a good way to max the production process good chips to a premium and some lacking something to a less value.

but nevertheless i3 supports ecc
so there is a cheap cpu for the consumer well it is among the cheapest

 

[Magic Carpet]

Great read, I had a blast

 

[Morg.]

This is nonsense.
You can't make a false and completely baseless statement without a single shred of justification for it and say "that much is clear"... how is it clear? why is it clear? why is it even true? (it isn't true)

Capitalism is a concept, and like all concepts the implementation will differ from the concept, that much is clear to everyone who understands both concepts :biggrin: (you know, concept and implementation).

Such a simplistic approach to commercial interactions doesn't have any chance to work.

For [you], the logical explanation is the following:

- Except for very rare cases (like newtonian vs quantum physics when you can safely ignore quantum effects for human-scale stuff), the only way to have a good theoretical model is to take into account all constituents.

- In effect it is impossible to determine if an engine is going to work before having documented every piece of it in a physics simulator (and even then ... you can't be 100% sure)

- The complexity of a human system is far superior to that of an engine or simple machinery like an airplane (and that .. is already incredibly complex)

So here we are, attempting to model a system for human commercial interaction.

It's based on humans, which we know are animals, but also self-aware beings, capable of great perversity and much much more.

That fact isn't even 1% covered in any theoretical model I've heard of, and that is why it is impossible (read 99.99999% unlikely) that any of these be implemented with 100% success.

Where in theory do you read about how (animals) reproduction instincts are going to drive the humans taking part in the system towards a direction or another ?

Where in theory do you read bout how (self-aware/perversion) frustrated individuals are going to twist appearances to take advantage of others / break the equilibrium ?

The conclusion is simple : capitalism or communism cannot be the answer, as to make anything work you require a global approach, which implies the subset covered by either theory will not be implemented in it's full simplicity / purity.

 

[taltamir]

There are several significant faults in your argument.

1. Prediction can be done: People make partial models of engines that work all the time without knowing and accounting for all the laws of reality. This is because you can use measured approximations rather then calculating every individual atom.

2. While you can't have a simulation 100% identical to reality, you can just BUILD something (or find it already existing) and observe it. This results in factual observation rather then a simulation.

3. We aren't simulating: All those economic models have been tried IRL. I am basing my discussion in reality and observed fact and evidence. It is now a question of our skill at interpreting this data, as well testing our objectivity (do we accept false data as true because it supports what we want to believe).

I discovered communism when I was 11, I fell in love. I told everyone about it and begun studying it. At 12 I had seen it has always failed and I have figured out exactly WHY and moved on. Although it was only recently due to some well thought arguments of others that I came to realize why it is an undesirable model (loss of freedom) rather then merely a desirable yet impractical one.

The conclusion is simple : capitalism or communism cannot be the answer, as to make anything work you require a global approach, which implies the subset covered by either theory will not be implemented in it's full simplicity / purity.

What tripe.

Also I would love to hear what part of

Capitalism is generally considered by scholars to be an economic system that includes private ownership of the means of production, creation of goods or services for profit or income, the accumulation of capital, competitive markets, voluntary exchange, and wage labor.

is the issue... is the voluntary exchange and wage labor? (no slavery or theft)...

 

[Morg.]

Obviously you won't ever get it but w/e -
I tell you the missing part is the human element, and you give me a quote that contains nothing about that - why ?

1. Partial models of engines are far more complete and no they aren't done without knowing and accounting for the laws of physics. Measured approximations are either related to extensive measurements or a theoretical model including proof.

It is science, not just talk or theory. There is absolutely no comparison between capitalism (an idea) and engine design (a reality).

2. You can just build and test, yes. Noone built capitalism and tested it so ... what's the idea there ?

3. None of those economic models have been tried IRL, there have been attempts at sticking more or less close to a concept or another but that's all there is to it.

Then I guess you must be 13, since you haven't yet realized that communism hasn't always failed.

The biggest loss for all humanity in the past century was the loss of the communist university of Moscow. - And by the way the USSR did not fail on its own.

The most powerful country today is a communist one, care to comment ?


And for the part you quoted ...

Do you realize accumulation of capital and creation of stuff for profit are two things that lead to optimization for profit ?

Do you realize that implies that to some degree, other variables will suffer, including everything since capitalism supports no other value ?

Do you realize competitive markets effectively push people towards hurting their competitors, thus winning the competition and wasting even more energy for futile purposes ?

Do you realize that the very idea of voluntary exchange means that all I have to do is to trick you into exchanges favorable to me in order to take advantage / make profit ?

Do you realize that by that point, those who have been tricked are left with no choices but to work for those who tricked them in exchange for a wage, not unlike in past history.

Do you eventually realize that all these evil plans, which are direct consequences of the simple basic capitalist idea, inevitably lead to a much greater loss of freedom in the end, where everyone w/ capital / profit / power is pushed (by the capitalistic rule of moar profitz) towards making sure the masses remain as oblivious as possible to their disadvantage in order to trick them into more profitable deals ?

As we're a species and our only goal is to reproduce as much as possible and become as tough as possible, any system limiting evolution or growth is a failure.

 

[taltamir]

Obviously you won't ever get it

I got it when I was 11, I got why it was wrong when I was 12.

It is science, not just talk or theory. There is absolutely no comparison between capitalism (an idea) and engine design (a reality).

You are the one who brought up engines and used them as an analogy, I was explaining why it is a faulty one. I am glad to see you admitting you were wrong and adopting my position on the issue.

The most powerful country today is a communist one, care to comment ?

I am curious, which country is that?

Do you realize <insert false statement>

Nice rhetoric, care you offer any evidence proof or justification?

 

[Morg.]

Congratulations on a successful troll. -- and there i was thinking you were just stupid ...

 

[Gundark]

Well, In my next build I'll deffinetly have ECC ram. 2 months ago I added 2 more ram sticks and problems start to rise.
My mb is old Asus M3A and with 2 ram sticks, BSOD or critical errors were in order of 2-3 times per year, nothing allarming. Since I heavily use emulation, and it's very error sensitive, for me it matters. But when I upgraded memory, first in some emulation scenarios that required 2+GB of memory, I experienced very frequent crashes or Out of memory errors ( like once in every two runnings or more ). So I ran Memtest and found a module with error ( it was Geil Black dragon, the other ones are Kingston ). However, if I test only Geil ram or only Kingston ram, Memtest could not detect any failure. So, I tried to test this ram on other machines ( Intels to be precise ). What I discovered is that in several 775 systems, 4 ram modules have errors if they are in twochannel mode, in single channel mode they are OK. Then I've exchanged my Geil with friend for Kingston, and I've passed the Memtest without error detection, but I still experienced ram related errors in emulation ( less frequently but still noticable ).
So, in my next upgrade I have two options.
1) Make sure that I have no more than 2 ram modules. If I need more in the future, I should replace that with larger sticks, and not add two more.
2) Make use of ECC ram. It seems to be feasible. Oh yes, and deffinetly by a bether mobo.

Now, my observation. Majority of users only have two ram modules. For them, errors are very rare, and can live happily ever after. Also, they aren't likey to use strictly error sensitive apps. Once they run 4 module ram, problems will skyrocket. However, I belive that this is mobo related, but this is no excuse. If manufacturers are selling mobo with 4 slots configuration, then they should guarranty that they are working errorfree.
So, after reading this thread, I'm a 100% supporter of ECC.

 

[anikhtos]

the same here
my next machine will be with ecc
and i think the memory controller has a problem with 4 memory channels
but then it is rediculous to sell a product that has 4 dimm
but if you use all 4 you will get stabi;ity issues more often.
and the older the machine gets the worse the ram behaves

ram ecc to regular is not that huge in money
the diferense will be only in mobo this days
i3 sandy supports ecc!!!!! with server chipsets
so thats 100$ more to get ecc in intel based systems
in amd hmmmmmmmmm its free
ecc is supported by amd

 

[Turbonium]

I just read through most of this thread, and I have to say, what a great read.

I've always been interested in the ECC vs. non-ECC memory debate (I've often wondered just how often real errors do happen with non-ECC RAM, and their potential impact), and this only adds to that interest.

Regarding Intel's server CPUs: are they just as good, if not better, than their equivalent consumer-level CPUs? For example, would a Xeon/Itanium be just as competent at gaming and content-creation as an i5/i7 would be?

Well, FWIW in this discussion, my NAS holding my photos seems to have developed RAM problems. When viewing pictures, some of them appear corrupted. Rebooting the NAS, and accessing those same pictures from a different, known-stable PC, shows different pictures in that series are corrupted. I have since shut down the NAS, to prevent further corruption.

But RAM corruption is real, and it's serious. Granted, this seems to be some kind of failure, but it would have been nice, if the NAS used ECC, and could detect corruption/failure, and automagically shut down, or at least gone into some sort of failsafe mode, where it would not operate as a NAS, but would blink a light on the front indicating hardware failure.

This post is the primary reason I resurrected this thread. It allows me to bring up something that happened to a system of mine some years ago.

The system in question was built on an Intel D875PBZ, had 2 sticks of Corsair RAM, and everything was running at stock speeds. Out of nowhere, both storage drives, which I had running in a RAID-1 (mirroring) array, developed data corruption issues simultaneously. Windows was citing all out CRC (cyclic redundancy check) errors.

Not taking any risks, I quickly backed up all my data, and did an RMA on the drives. In the end, I got lucky in that I only lost a single mp3 file.

However, I never quite found out what had caused the error. Honestly, I don't think either of the drives was actually defective, and I certainly don't think they could both have failed at the same time, as that is extremely unlikely to happen. I also don't think the RAID controller was defective, as following the incident, I used it to run a fresh RAID-1 array with absolutely no issues for years, even up until now.

Given that RAID-1 arrays transfer identical data to both drives (including any corrupted data), could it be that some sort of memory corruption led to the CRC errors (i.e. perceived hardware failure) of both drives? I would assume the RAID controller's function depends on data supplied by the system RAM.

If so, this leads me to seriously consider ECC memory for all future systems I assemble. I realize ECC memory isn't a magical solution to every form of data corruption, but I'd rather not go through an experience like that (or worse) again.

As for whether or not it's ethical to not have all forms of data protection possible in lower-tier (consumer) systems if possible (ignoring the economic side of things): I'm not sure where I stand on this. I do understand what tynopik is saying though, in that it can be argued that it is irresponsible to knowingly subject consumers to potential data loss and data corruption, especially without advertising its potential to happen. I also clearly understand the distinction he made, on more than one occasion, between data corruption/loss, and lost uptime, with the latter being an acceptable reality, the frequency of which does not necessarily need to be mitigated at all costs. All this being said, I do not wish to reopen the debate, or be the cause of any flaming.

 

[blastingcap]

I just read through most of this thread, and I have to say, what a great read.

I've always been interested in the ECC vs. non-ECC memory debate (I've often wondered just how often real errors do happen with non-ECC RAM, and their potential impact), and this only adds to that interest.

Regarding Intel's server CPUs: are they just as good, if not better, than their equivalent consumer-level CPUs? For example, would a Xeon/Itanium be just as competent at gaming and content-creation as an i5/i7 would be?


This post is the primary reason I resurrected this thread. It allows me to bring up something that happened to a system of mine some years ago.

The system in question was built on an Intel D875PBZ, had 2 sticks of Corsair RAM, and everything was running at stock speeds. Out of nowhere, both storage drives, which I had running in a RAID-1 (mirroring) array, developed data corruption issues simultaneously. Windows was citing all out CRC (cyclic redundancy check) errors.

Not taking any risks, I quickly backed up all my data, and did an RMA on the drives. In the end, I got lucky in that I only lost a single mp3 file.

However, I never quite found out what had caused the error. Honestly, I don't think either of the drives was actually defective, and I certainly don't think they could both have failed at the same time, as that is extremely unlikely to happen. I also don't think the RAID controller was defective, as following the incident, I used it to run a fresh RAID-1 array with absolutely no issues for years, even up until now.

Given that RAID-1 arrays transfer identical data to both drives (including any corrupted data), could it be that some sort of memory corruption led to the CRC errors (i.e. perceived hardware failure) of both drives? I would assume the RAID controller's function depends on data supplied by the system RAM.

If so, this leads me to seriously consider ECC memory for all future systems I assemble. I realize ECC memory isn't a magical solution to every form of data corruption, but I'd rather not go through an experience like that (or worse) again.

As for whether or not it's ethical to not have all forms of data protection possible in lower-tier (consumer) systems if possible (ignoring the economic side of things): I'm not sure where I stand on this. I do understand what tynopik is saying though, in that it can be argued that it is irresponsible to knowingly subject consumers to potential data loss and data corruption, especially without advertising its potential to happen. I also clearly understand the distinction he made, on more than one occasion, between data corruption/loss, and lost uptime, with the latter being an acceptable reality, the frequency of which does not necessarily need to be mitigated at all costs. All this being said, I do not wish to reopen the debate, or be the cause of any flaming.

I'm with the pro-ECC crowd on this one. An 8% annual soft error rate is bad enough but it's just going to get worse and worse as node shrinks keep happening. Shrinks mean less charge is necessary to flip bits. ECC should be standard from here on out, because that 8% will rapidly increase as we go down node sizes.

 

[Soulkeeper]

I want to get ecc memory for my next build.

I've had a few odd system lockups after 2+ weeks uptime since I upgraded to 16GB from 8GB that have me wondering.

 

[balnazzar]

I want to get ecc memory for my next build.

I've had a few odd system lockups after 2+ weeks uptime since I upgraded to 16GB from 8GB that have me wondering.

Do AMD FM2 systems support ecc unbuffered ram? (Like was for am3 systems).

 

[Charles Kozierok]

Do AMD FM2 systems support ecc unbuffered ram? (Like was for am3 systems).

Unfortunately, no.

It stinks but AMD is moving towards Intel on this, rather than the other way around.

The move away from error checking on memory has been a sore spot for me for many years. It was at least understandable when memory was expensive, but now it is inexcusable. It should at least be an option on at least all desktop platforms, IMO.

 

[taltamir]

Unfortunately, no.

It stinks but AMD is moving towards Intel on this, rather than the other way around.

The move away from error checking on memory has been a sore spot for me for many years. It was at least understandable when memory was expensive, but now it is inexcusable. It should at least be an option on at least all desktop platforms, IMO.

its a shame too because AMD can differentiate itself from intel in such a way. It can't compete with them on performance or features... but they could if they stopped playing the intel's game of artificial differentiation between server and home markets

 

[VirtualLarry]

its a shame too because AMD can differentiate itself from intel in such a way. It can't compete with them on performance or features... but they could if they stopped playing the intel's game of artificial differentiation between server and home markets

Given AMD's attitude towards market differentiation, vis-a-vis their support for VM opcodes and other ISA extensions - I am also surprised that AMD is dropping support for ECC.

 

[anikhtos]

Given AMD's attitude towards market differentiation, vis-a-vis their support for VM opcodes and other ISA extensions - I am also surprised that AMD is dropping support for ECC.

in a way it is funny if you consider it
intel fragment the market and you had to pay a premium for the ecc
amd respond by giving ecc support to all the cpus in the line

now amd droped ecc support :-(
and intel gives ecc support to few low end cpus!!!!!
i3 cpu when paired with a server chipset
new server atom with ecc support

even the e3 xeon line is the more or less the i3 i5 i7 line with all the feutures enabled and not that higher priced

now you can build a cheap ecc machine around intel but not amd?!?!?!?

lol amd shot itself on the foot
can not compete at raw power can not compete at power consumption and now even looses at the added feutures!??!?!?!?
if only the e350 supported ecc it would make suck a nice nas system around it :-(