Question about ECC memory

should be standard in all machines at this point

Garbage for serious gaming

Attached: ram-guide-v3.png (848x1492, 487K)

meow

Fuck you.

Fine for a workstation.

>Fun to play with
>Slightly lower clocks/latencies.
>More secure and safe, especially if you run big arrays of hard drives.

Is this legit? Please revoke all consumer rights to ever make decisions in technology.

Is 2400MHz the fastest ECC DDR4 available?

Yes, JEDEC standards are holding back mem speeds for servers. You can OC them pretty accurately though so it isn't an issue.

It's completely legit. If you can't see why then you must not understand how ram performance works on the hardware level.

No, Crucial sells 2666 MHz DDR4 ECC modules but Samsung's 2400 MHz ECC has Samsung B-dies which tend to overclock very well. There were several reports online about overclocked Samsung DDR4-2400 ECC so I tried it and voila, no problems at 2933
Haven't even tried more yet

Yeah but oc usually defeer the reason for ecc memory.
Anyway I always used it on servers, and used it and without it on workstations didn't get any difference what real reason for ecc ? I guess at this point programs and os can handle errors pretty well unless you need to do some kind of super precise calc? Also quadros gddr5 do any kind of ecc?

>oc usually defeer the reason for ecc memory.
How so?

I'd say overclocking is a good reason _for_ ECC RAM because it increases the likelihood of errors.

Gives me a safe feeling when doing 100 core years' worth of computation on university cluster.
Not sure what's the error rate without ECC, but I assume it would be at least one bit flip per run.

ecc is not needed in most cases. plus it doesnt even stop vulnerabilities, only bit flipping

ECC also protects against Rowhammer attacks which can be triggered by JavaScript, so it's useful in desktops, even if you don't care about rare random bit flips.

zdnet.com/article/dram-error-rates-nightmare-on-dimm-street/
Google reported 3.751 correctable errors per DIMM per year

Attached: picture-27.png (474x445, 10K)

>plus it doesnt even stop vulnerabilities
Not true.
googleprojectzero.blogspot.com/2015/03/exploiting-dram-rowhammer-bug-to-gain.html

>ddr 1 & 2

Attached: googleserver.jpg (1200x900, 249K)

You're right. DDR4 never has any errors.

my DDR4 strips passed a whole night of rowhammer tests on memtest x86, i thought ddr3 was the last standard that didn't have mitigations for it.

Attached: hd_spurdo_es_pilli.png (2000x1751, 108K)

It's still susceptible to random bit flips.
Of course that doesn't matter if all you do is playing video games.

Its overpriced stuff aimed for enterprise market.
As a home user you will never ever need it. Its a waste of money.

It kinda makes sense for stuff like online banking servers where errors are absolutely deadly.

>overpriced
Why aren't you competing if it's overpriced?
You could make millions.

Where to buy that isn't
>ebay

Samsung DIMM 16GB, DDR4-2400, CL17-17-17, ECC (M391A2K43BB1-CRC)

Get two of these

You buy ecc memory to have the absolute best reliability. Once you introduce oc it's running out of specs, so if you have any critical mission where reliability comes first before performance you stick with specs.

You're right, those once-in-6-months errors don't really bother me, i save my work and i think a power outage is much more likely.

ECC memory is for places where people lose their job if a server shits the bed, not for gaming or even rendering/coding/officing

Attached: 1486770632219.png (639x1499, 989K)

It's an overused meme really only suitable for high end workstations and servers.

Overclocked ECC is more reliable than overclocked non-ECC RAM though (= most "gaming" RAM).

It pretty much is by now, at least on late 2017 or newer hardware. All new Coffee Lake CPU's, i3 and up, support ecc as does Ryzen. With AMD it just depends on motherboard vendor, if they have enabled it at the bios level.

>Samsung DIMM 16GB, DDR4-2400, CL17-17-17, ECC (M391A2K43BB1-CRC)
€ 177,60
(OC to 2933 safely, after all you are noticed about any errors)

>cheapest gaymur RAM
> Kingston HyperX Fury schwarz DIMM 16GB, DDR4-2933, CL17 (HX429C17FB/16)
€ 169,84

>WOW, SO OVERPRICED
>NEVER UNDER ANY CIRCUMSTANCES USE IT IN ANYTHING BUT BANKING SERVERS

Unnecessary, in a lot of use cases, but when you want/need the reliability it buys you a safety net.

meme

DO NOT OC ECC RAM
REPEAT
DO NOT OC ECC RAM

Apples and oranges.
CL 17 @ 2933 is as good as CL 14 @ 2400

why not?

because oc dries up the magic ecc juice

Attached: 1474225715209.png (367x506, 212K)

But seriously don't overclock ECC ram.

Attached: 1525203350456.jpg (828x482, 289K)

no flashy 6gorillion colours LED's and big ass, useless heatsink on it. Checkm8

OCing ECC is fine if you're a retard who doesn't actually understand the benefits of ECC. It won't blow your computer up or anything.

Higher clock speeds than rated can lead to errors and make you more susceptible to rowhammer attacks. ECC provides defence against both of these, so overclocking ECC memory is just counterproductive.

Attached: 4kultra.jpg (3840x2160, 411K)

Holy shit, user on suicide watch, 100% dead and finished, how will he ever recover?

Attached: 1528279167375.jpg (1440x1080, 891K)

>Are inaccuracies caused by single bit flips completely unacceptable?
>Is downtime from crashes due to bit flips completely unacceptable?
Both are no on my gayman desktop so no, extra cost not justified.

This is perfectly justifiable. The ECC question is all about risk tolerance. You WILL get flipped bits, probably most of them will be harmless. If you're okay with this then running nornal memory overclocked to an absurd degree might be worth the tradeoffs.

I don't think higher clock speeds negate the protection provided by ECC.
You might find yourself running into more two-bit errors when overclocking, which will still cause problems, but it won't cause any more issues than it would on a non-ECC system. Same thing for rowhammer. It is no more risky than using non-ECC, so your argument is against overclocking in general, and the ECC part is irrelevant.

No. You pick safety or speed. If all you care about is speed then buy fast non-ecc memory and overclock it if you need to for the best results. If you care about safety then use ECC memory and run it as spec'd. Overclocking ECC memory is retarded because you paid more to get the same reliability as just buying fast memory.

>ECC magically stops working at higher clocks

Attached: 1411392105381.jpg (770x730, 129K)

The only RAM I've ever had go bad was ECC RAM. Go figure.

The functionality you paid for becomes degraded until the point that single bit corrections are common and 2-3 bit errors actually happen. At this point the ECC benefit is gone.

You're not going to get to the speed of actually fast normal memory before this happens.

Actually fast normal memory contains exactly the same dies if you choose wisely.

I can see what you're saying, but ECC would still be more reliable than non-ECC. Overclocked ECC will still catch single-bit errors.
I'd agree overclocking most servers would be kinda stupid, since usually you scale them out, instead of up, but overclocking ECC in a workstations or speed-specific servers would make a certain amount of sense.
Actually, given that ECC will catch the single bit errors introduced by memory overclocks, I'd bet ECC works very well for overclocking in TR processors.

That's like saying don't wear a condom when fucking a street hooker.

Condom is to prevent AIDS.
Street hooker probably has AIDS.
So it's "counterproductive" to combine the two.

this is a well thought out troll. 8/10 convinced me to read it.

Everything is binned these days.

It's no more reliable if you bring it out of the safety envelope ECC provides. You're far better off just using fast ram and checking it against several hours of memtest86. Servers very rarely need super fast memory, desktops very rarely need ECC reliability, and ECC is expensive. Overclocking ECC is a strange optimization.

No you absolute faggot. Let me put it into terms your shitty brainlet walnut can actually understand.
It's like saying you can go down to the seedy gaybar you love so much and get fucked by dozens of dicks as long as they use a condom. This is mostly true. You probably won't get aids this way. But then you say "oh boy, because of the condom I can get dicked even harder too!" so now you're getting fucked by lots of dicks AND you're getting fucked really hard. Well it turns out the condom breaks sometimes because of the rough fucking, but you think "oh that's okay, I'm still better off then when I was just fucking a few guys because the condom _mostly_ works".
Then you get AIDS one day when the condom breaks over a poz load and floods your battered ass. This could have happened back before you were using condoms too, but over the years with all that money spent on rubber was pointless because you still ended up with AIDS.

Safety devices have a specific envelope of usage. Step outside that envelope and the device becomes a false sense of security even if "it works most of the time".

Good for FreeNAS / ZFS

A must for accuracy and data integrity is critical.

gaymen motherboards dont even support ecc so fuck off

>intel motherboards for gaymen dont even support ecc so fuck off
ftfy