Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Non-ECC memory corrupted my hard drive image [video]

youtube.com

Post reply on HN
Print view View original

91–100 of 161 posts

Re: Non-ECC memory corrupted my hard drive image [video]

#91

Post by toast0 »

eternityforest wrote: post #71
simoncion wrote: post #63

Earlier quoted context omitted.

It is my understanding that the ECC that you're talking about only protects data-in-flight between the module and whatever is reading or writing the data. It does not protect against corruption of data-at-rest, which is what is protected with ECC in DDR4 and older. It's also my understanding that the DDR5 data-in-flight ECC is a mandatory feature because the link between the memory modules and everything else is so e…

Taking a quick glace at the articles, I think it's the opposite, DDR5 protects data at rest only, because they want to make the chips so unreliable it can't work without it, not the bus. But in practice, it will probably be more reliable than DDR4 without ECC, since now you need 2 cosmic ray flips, or 1 plus a manufacturing defect flip, and the defect flips will probably be uncommon-ish. It's too bad data in flight i…

The problem with the DDR5 approach is there's no reporting mechanism, so while it will reduce the error rate of a marginal module, it doesn't let you know so you can replace it. In my experience with ECC modules, a module with some errors is a lot more likely to get more errors than one that's operating with zero errors.

Re: Non-ECC memory corrupted my hard drive image [video]

#92

Post by helf »

moloch-hai wrote: post #9

We don't have ECC mainly because Intel has long been hostile to "consumer" access to ECC. Apparently this was conceived as a market segmentation scheme: people outfitting servers could get ECC when they pay a huge premium. They would thereby not be tempted to cheap out and buy consumer-grade equipment, otherwise wholly adequate to meet all their needs at a radically cheaper price. That we cannot get laptops or even d…

I see this a lot but I really don’t think it was/is that simplistic.

It’s added complexity and cost for something that rarely would benefit most consumers. Now, you can argue that the complexity and cost is a nonissue on modern setups and I would probably agree.

But Intel has long had desktop grade hardware with ECC support. The 440GX chipset supported ECC and I ran a Dell GX1 SFF with 768MB of ECC PC100 for yeeeears with a 450 MHz P3 and later upgraded to 1.4ghz tualatin-256 via a slotket adapter.

The 440HX /socket 7/ chipset supported ECC. And that’s a Pentium 1 chipset.

The 440BX/GX and 450NX supported ECC and that’s with desktop pentium 2 and 3 chips.

The 820/820E/840 supported ECC with desktop celeron and pentium2/3 chips

845/845e/850/850e/860 pentium4 chipsets support ECC

875/e7205/e7221/e7230 did with desktop pentium 4 and pentium d chips

925/925xe/955x/975x did with desktop pentium 4/pentium d/core 2

It’s more sparse now that they moved to the IMC, granted. But Intel has long had multiple chipsets per generation with ECC support for desktop grade hardware.

Re: Non-ECC memory corrupted my hard drive image [video]

#93

Post by Gordonjcp »

moloch-hai wrote: post #9

We don't have ECC mainly because Intel has long been hostile to "consumer" access to ECC. Apparently this was conceived as a market segmentation scheme: people outfitting servers could get ECC when they pay a huge premium. They would thereby not be tempted to cheap out and buy consumer-grade equipment, otherwise wholly adequate to meet all their needs at a radically cheaper price. That we cannot get laptops or even d…

> That we cannot get laptops or even desk machines with ECC, and so have them crash frequently, is seen as a trivial side effect of the strategy.

How frequently would you say you encounter a crash that you can pin down to a lack of ECC memory in your laptop or desktop?

Re: Non-ECC memory corrupted my hard drive image [video]

#94

Post by AaronFriel »

mrlonglong wrote: post #90

We've got a server that keeps rebooting due to a bad ECC DIMM chip. I thought the whole point of ECC was to keep the server going until we can replace the DIMM?

ECC can typically correct 1 bit errors and usually detect (and fault) on 2 bit errors.

Your server is faulting and preventing itself from corrupting data.

Re: Non-ECC memory corrupted my hard drive image [video]

#95

Post by Filligree »

Gordonjcp wrote: post #93
moloch-hai wrote: post #9

We don't have ECC mainly because Intel has long been hostile to "consumer" access to ECC. Apparently this was conceived as a market segmentation scheme: people outfitting servers could get ECC when they pay a huge premium. They would thereby not be tempted to cheap out and buy consumer-grade equipment, otherwise wholly adequate to meet all their needs at a radically cheaper price. That we cannot get laptops or even d…

> That we cannot get laptops or even desk machines with ECC, and so have them crash frequently, is seen as a trivial side effect of the strategy. How frequently would you say you encounter a crash that you can pin down to a lack of ECC memory in your laptop or desktop?

You can't, that's the thing, right?

I have a Ryzen desktop with ECC, and it registers about one bit-flip per week. I don't know how many of those would become crashes, but I'm more worried about the ones that wouldn't.

Re: Non-ECC memory corrupted my hard drive image [video]

#96

Post by Filligree »

kevingadd wrote: post #61
newZWhoDis wrote: post #30

Earlier quoted context omitted.

My favorite is when data silently corrupts, and then is happily propagated to your off-site recovery :/

Yeah, I don't know when the drive started failing, so in practice my backups are probably all screwed too. My only option would be to get a spare drive, restore a backup to it, then do a block level diff of the current and backup volumes and try to figure out whether any of the differences are file corruption.

Probably I'm saying nothing you haven't thought of, but ZFS is great for preventing this. If you also pair it with ECC, then you've eliminated most ways to cause corruption.

Re: Non-ECC memory corrupted my hard drive image [video]

#97

Post by aortega »

I have 128 GB of non-ecc memory in my notebook, never detected a single error, and has been on 24/7 for more than 4 years.

Unless you live over 4000 meters over the sea level, like to compile while flying or live close to an unshielded nuclear reactor, you don't need ECC.

And most memory problems you can fix by better cooling, and better shielding.

Re: Non-ECC memory corrupted my hard drive image [video]

#98

Post by aortega »

chunk_waffle wrote: post #51

For anyone curious (as I was once and looked into it) Dell and Lenovo both ship Laptops with Xeon's and ECC memory though they are very expensive.

I have a lenovo P52, it came with 16 GB ECC. The dimm is not super-expensive, but it's hard to find, and only comes in half the size of regular dimms.

I replaced it with non-ECC memory and never had an issue. In fact you can combine ECC and non-ECC memory without any problem.

Re: Non-ECC memory corrupted my hard drive image [video]

#100

Post by gumboza »

helf wrote: post #92
moloch-hai wrote: post #9

We don't have ECC mainly because Intel has long been hostile to "consumer" access to ECC. Apparently this was conceived as a market segmentation scheme: people outfitting servers could get ECC when they pay a huge premium. They would thereby not be tempted to cheap out and buy consumer-grade equipment, otherwise wholly adequate to meet all their needs at a radically cheaper price. That we cannot get laptops or even d…

I see this a lot but I really don’t think it was/is that simplistic. It’s added complexity and cost for something that rarely would benefit most consumers. Now, you can argue that the complexity and cost is a nonissue on modern setups and I would probably agree. But Intel has long had desktop grade hardware with ECC support. The 440GX chipset supported ECC and I ran a Dell GX1 SFF with 768MB of ECC PC100 for yeeeears…

I miss the BX chipset. The last PC that wasn't a complete piece of shit I had was a 440BX chipset P3.
Post reply on HN