ECC RAM

Technology exists to prevent silent data corruption.
It's been around for a long time. Do you want it?

>What? Nah... I'll pass...
>something something servers
>it's not """needed"""
t. Sup Forums RGB LED cr3w -=[MLG]=-

Other urls found in this thread:

cs.toronto.edu/~bianca/papers/sigmetrics09.pdf
myredditvideos.com/
twitter.com/NSFWRedditGif

what are you pushing through RAM so frequently that you have a significant risk of corruption?

I'm using it on my win98se pc

Nothing I do on my personal rig is critical enough for ECC, and according to Google's own testing, errors resulting from flipped bits happened 1% of the time on average, up to 5% from one unnamed supplier.

It's borderline meme tier if it didn't destroy your work with out it.

What would happen usually if i didn't have ECC, and for whatever reason, something got corrupted?

Would the computer crash, file not be written properly or something else?

ECC RAM is designed for systems that run critical software, and have months, or years of uptime.

Not your home PC that 90% of people restart, or shutdown every few days.

Also useful on some high end workstations doing large CAD projects and the like.

Back when I had a PIII 700 MHz CPU in a Dell, it came with DDR ECC RAM, so I used it. Ran it side-by-side with a system that didn't have ECC.

No difference that I ever noticed, for nearly 10 years.

In today's tech ecosystem (meaning: DDR4 and up), ECC is extremely exotic for very unusual applications. Nobody outside of a corporate server farm (or some equivalent application) has even the slightest need. There is no justifying the expense, unless you really, really have a need to max out your mommy's credit card on something that passed way outside the meme orbit around 2005.

all i do is shitpost on Sup Forums and play CSS so i reckon i dont need that crap

Why does it have to be "large CAD projects" and not ANYTHING of importance?

>Embedded firmware developer
>Compiles binary
>Bit flip while saving to disk
>Send to client
>Devices crash

Fun fact: Nobody has proven that ECC actually helps with anything meaningful. ECC corrects soft errors only, and to date nobody has demonstrated that these actually occur--it's all theoretical. ECC is largely just a way to rip off dumb middle managers.

>my ass
There are tons of studies.

Work published between 2007 and 2009 showed widely varying error rates with over 7 orders of magnitude difference, ranging from 10−10–10−17 error/bit·h, roughly one bit error, per hour, per gigabyte of memory to one bit error, per millennium, per gigabyte of memory.[5][6][7] A very large-scale study based on Google's very large number of servers was presented at the SIGMETRICS/Performance’09 conference.[6] The actual error rate found was several orders of magnitude higher than previous small-scale or laboratory studies, with 25,000 to 70,000 errors per billion device hours per megabit (about 2.5–7 × 10−11 error/bit·h) (i.e. about 5 single bit errors in 8 Gigabytes of RAM per hour using the top-end error rate), and more than 8% of DIMM memory modules affected by errors per year.

cs.toronto.edu/~bianca/papers/sigmetrics09.pdf

One of my servers has ECC RAM, my next file server will too.

>About a third of machines and over 8% of DIMMs in our fleet saw at least one correctable error per year

Don't you do anything of importance on your desktop?

Modern operating systems (since Win NT) and applications from the same era correct for corrupt data. There is a very slight performance degradation that more or less can't be detected, other than a 0.01% deviation in some benchmarks.

An actual crash _might_ occur in 1 out of a trillion thrown bits. Mostly, ECC helps out in server farms where massive data throughput, measured in petabytes per hour. There are slight performance increases and measurable energy use reduction. Google, for an example, probably saves several thousand dollars an hour in reduced energy use on all of their servers running all over the entire planet.

Which means nearly fuckall when you make the hundreds of billions Google does ... but there is also a reduction in general headaches when trouble-shooting any problems because you have eliminated RAM errors to the nth degree.

For that reason, also, places like NASA and CERN like ECC in their systems because who wants to track down a buggy RAM module when your space station or super collider blink or fart at a weird moment.

If you're worried about your gaymemes running an extra framerate faster every 100 hours, you should consider investing your money in psychiatric care.

yes but how many errors that actually affected meaningful data were corrected

I don't like this semantic nonsense, this doesn't necessarily address the heart of my point

Not really, anything I care about is done on remote servers. I just use my desktop for web browsing and SSH

>what's soft ECC correction

lmao id rather buy a better geforce gtx xxxtreme

>gaymemes
This website is 18+. Nobody plays video games here.

ECC RAM is best suited for mission critical virtualization.

>This website is 18+. Nobody plays video games here
Meanwhile this 18+ website has at least 4 boards dedicated to videogames

Yeah, and none of them talk about video games.

wrong

Sorry, ECCs premium isn't worth it to the consumer.

If you're even taking in to account some sort of esoteric security feature, if you can flip bits on my machine, you've already won.

I've run a server for years and never even worried about it. Didn't even know it existed when I started

>ECCs premium

It costs like 10% more. Let consumers decide if they want to pay for it.
Keep in mind that some very cheap Intel CPUs and chipset do support ECC. It's just the mid-high range non-server CPUs that don't support it. Pure market segmentation, nothing else.
IDK what AMD is doing nowadays. They used to have ECC enabled on more CPUs.

Fun fact: my uni has quadro/xeon with ecc memory workstations that we use for exams, and inventor crashed to me during a quite important exam even with such precision hardware.

Fun fact: 99% of crashes are because of software bugs.

Sup Forums cancer should be lynched

The motherboards are also more expensive. Some CPUs don't support it.

cucktel makes you purchase xeons for it, at least AMD does ECC on every cpu, even apus.

i dont understand people who make those home-nasses with intel gear, for less you have amd + ECC

>1 bit flips
>firefox crashes
>restart firefox
>continue on like normal

>If you're even taking in to account some sort of esoteric security feature, if you can flip bits on my machine, you've already won.
If someone was making modifications to the memory through software ECC wouldn't stop it anyway because the parity data would be updated as well. ECC like RAID solutions are only there to protect against hardware malfunctions. Also ECC only protects like single bit flips not whole rows of flipped bits

Random freak bit flips do not happen frequently enough for ECC to matter for most people. They very rarely happen in the first place but the chances of one happening in just the right spot to cause a system to become unstable is also incredibly unlikely. Systems that have to remain up constantly and will not be restarted very often should use them because while they are uncommon the chances of one happening increases with time and you can also accumulate errors which makes potential problems even worse. ECC is for high reliability in the same way that RAID solutions are.

>Random freak bit flips do not happen frequently enough for ECC to matter for most people.
[citation needed]

Google found 2.5–7 × 10−11 error/bit·h

What if you compile code for your customer and get a bit flip just when the binaries are written to disk? You would never notice until they report strange crashes weeks later.

But of course, the $50 are better invested in a faster graphics card and RGB LEDs.

>Random freak bit flips do not happen frequently enough for ECC to matter for most people.
I kinda agree, but that's like saying seat belts don't matter for most people, because car crashes don't happen frequently enough.

At the end of the day the only question is, does ECC actually cost so much more to implement? And the answer is no, it doesn't.
ECC could easily be the standard in computers and most people wouldn't notice, but they'd enjoy the potential benefits.

The only reason we're even having this discussion is because Jewtel is doing market segmentation to sell more Xeons. And motherboard manufacturers are happy to play along if they can save $2 per motherboard.

>At the end of the day the only question is, does ECC actually cost so much more to implement?
It requires additional memory chips and the circuitry is probably somewhat different/complicated. It would probably increase the costs whcih the end user would take. With a wider market ECC could be cheaper than it currently is though but there has to be a reason for the average person to care or for the manufacturers to stop producing regular DIMMs

Well I mean the average user isn't a software developer. Workstations would benefit from ECC and I'd kinda expect them to have it.

Kingston Value 16GB 288-Pin DDR4 SDRAM ECC Registered DDR4 2400

$118

G.SKILL Ripjaws V Series 16GB (2 x 8GB) 288-Pin DDR4 SDRAM DDR4 2400

$108

>wow, such expensive

If you really can't have downtime or any errors (Server) you get ECC. Otherwise there is no advantage and it's just a fucking meme.
/thread

>only servers are used for work

Ecc errors on servers are logged. I've only ever seen an error once. This is on hundreds of servers.

>error-critical hobby servers

>premium
it's not even 10%, like $10 for 16 GiB of RAM.

That was just an example user, of course there are other machines that can't have any problems.

>i dont understand people who make those home-nasses with intel gear
because you don't actually need ECC and even if you do, i3's and pentium/celeron-branded chips have ecc support

ZFS can suffer catastrophic loss without ECC.