If I understand it right, all of the AMD Ryzen processors support ECC, provided the motherboard/chipset do as well. Which somewhat supports Linus' opinion.
Why don’t PCs use error correcting RAM? “Because Intel,” says Linus
11–20 of 30 posts
Re: Why don’t PCs use error correcting RAM? “Because Intel,” says Linus
#12While I don't think he should be flipping the bird, his technical arguments are absolutely spot on. An aside question for readers: How often does RAM error, and is it a significant problem in practice? I don't have any direct experience with this (that I could tell) using consumer computers for the past 15-odd years of programming.
That photo is from him saying "fuck you" to NVIDIA back in 2012: https://arstechnica.com/information-technology/2012/06/linus... FTA: "Bit flips can happen for many reasons, beginning with cosmic-ray impact or simple hardware failure. A large-scale study[0] of Google servers found that roughly 32 percent of all servers (and 8 percent of all DIMMs) in Google's fleet experience at least one memory error per year. But t…
Something that doesn't get said enough is that Google had a habit at the time of buying RAM chips that had failed manufacturer QA, stuck them on DIMMs themselves, and revalidated those DIMMs. They were really leaning into the whole "embrace failures if they're going to happen anyway and you can get cheaper servers out of it". So those numbers need to be taken with a grain of salt.
Re: Why don’t PCs use error correcting RAM? “Because Intel,” says Linus
#13I absolutely love that Linus speaks his mind. Candor and telling it honestly is getting rarer...
Unfortunately, "candor and telling it honestly" is frequently code for "I want to be hostile toward people I don't like, but I don't want to experience social consequences for it", so there's a lot of noise in the interpretation of this space.
Do you have any evidence for this claim?
Re: Why don’t PCs use error correcting RAM? “Because Intel,” says Linus
#14Given the current RAM and disk sizes, and how much we put trust on computers, every cell phone or computer should use ECC RAM and checksummed filesystems these days.
The density/process size of the RAM would seem to drive that as well. A cosmic ray, for example, has a better chance of flipping a bit on a denser collection of transistors. Assuming the die size is the same, due to the total amount of RAM increasing, as you mention.
Re: Why don’t PCs use error correcting RAM? “Because Intel,” says Linus
#15While I don't think he should be flipping the bird, his technical arguments are absolutely spot on. An aside question for readers: How often does RAM error, and is it a significant problem in practice? I don't have any direct experience with this (that I could tell) using consumer computers for the past 15-odd years of programming.
> While I don't think he should be flipping the bird, his technical arguments are absolutely spot on.
Linus is certainly known for some inappropriate rants against people which he has since toned down, apologized for, and worked to correct.
That said, this image is taken from an event where Linus was prompted about Nvidia in 2012 during some filming. For those who don't use Linux, these times were extremely chaotic in graphics, especially for Linux with respect to Nvidia, who were largely catering to Windows users. It's what kept Windows gaming in such a strategic position. This affects people who have lower end machines, people who can't afford Windows, and people who seek freedom from predatory data collection. Throwing up his middle finger at a company with those kind of practices is hardly revolutionary or inappropriate.
Source: https://www.wired.com/2012/06/torvalds-nvidia-linux/amp
Re: Why don’t PCs use error correcting RAM? “Because Intel,” says Linus
#16The market killed ECC for consumers. It used to be widely available to desktop class hardware (I used to have some), but it was slower and more expensive than non-ECC RAM. None of the marketing-types made a good case for why normal people should pay more money for less-performant hardware, so the market disappeared.
Even the security issues only meaningfully impact servers.
People like Linus can buy server-class hardware if ECC is so important to them.
Re: Why don’t PCs use error correcting RAM? “Because Intel,” says Linus
#17While I don't think he should be flipping the bird, his technical arguments are absolutely spot on. An aside question for readers: How often does RAM error, and is it a significant problem in practice? I don't have any direct experience with this (that I could tell) using consumer computers for the past 15-odd years of programming.
More often than you'd think. There's a paper from black hat 2011 [0] where bit flips and domains with a small hamming distance were used. Take google.com and woogle.com which in their ASCII representation only differ by one bit. The paper showed that these domains will receive traffic for their intended counterparts because of bit flips in the RAM in DNS servers.
[0]: https://media.blackhat.com/bh-us-11/Dinaburg/BH_US_11_Dinabu...
Re: Why don’t PCs use error correcting RAM? “Because Intel,” says Linus
#18The market killed ECC for consumers. It used to be widely available to desktop class hardware (I used to have some), but it was slower and more expensive than non-ECC RAM. None of the marketing-types made a good case for why normal people should pay more money for less-performant hardware, so the market disappeared. Even the security issues only meaningfully impact servers. People like Linus can buy server-class hard…
Old Sunblade desktops have a jumper switch where you can use either ECC or non-ECC ram.
Re: Why don’t PCs use error correcting RAM? “Because Intel,” says Linus
#19Earlier quoted context omitted.
That photo is from him saying "fuck you" to NVIDIA back in 2012: https://arstechnica.com/information-technology/2012/06/linus... FTA: "Bit flips can happen for many reasons, beginning with cosmic-ray impact or simple hardware failure. A large-scale study[0] of Google servers found that roughly 32 percent of all servers (and 8 percent of all DIMMs) in Google's fleet experience at least one memory error per year. But t…
Something that doesn't get said enough is that Google had a habit at the time of buying RAM chips that had failed manufacturer QA, stuck them on DIMMs themselves, and revalidated those DIMMs. They were really leaning into the whole "embrace failures if they're going to happen anyway and you can get cheaper servers out of it". So those numbers need to be taken with a grain of salt.
Even still, a one-third of a memory error per year is a rate that I'm totally comfortable with.
Re: Why don’t PCs use error correcting RAM? “Because Intel,” says Linus
#20If I understand it right, all of the AMD Ryzen processors support ECC, provided the motherboard/chipset do as well. Which somewhat supports Linus' opinion.
They don’t officially support it but they don’t disable it on consumer chips like intel does.
It’s supposed to work but if you have any problems with it you can’t go back to AMD and complain. IMHO that’s not really good enough, if ECC is going to protect you they need to officially support it.
It sounds like not officially supporting ECC is a CYA move for when it doesn’t work.