SoylentNews is people
SoylentNews
Linus Torvalds On The Importance Of ECC RAM, Calls Out Intel's "Bad Policies" Over ECC
There's nothing quite like some fun holiday-weekend reading as a fiery mailing list post by Linus Torvalds. The Linux creator is out with one of his classical messages, which this time is arguing over the importance of ECC memory and his opinion on how Intel's "bad policies" and market segmentation have made ECC memory less widespread.
Linus argues that error-correcting code (ECC) memory "absolutely matters" but that "Intel has been instrumental in killing the whole ECC industry with it's horribly bad market segmentation... Intel has been detrimental to the whole industry and to users because of their bad and misguided policies wrt ECC. Seriously...The arguments against ECC were always complete and utter garbage... Now even the memory manufacturers are starting [to] do ECC internally because they finally owned up to the fact that they absolutely have to. And the memory manufacturers claim it's because of economics and lower power. And they are lying bastards - let me once again point to row-hammer about how those problems have existed for several generations already, but these f*ckers happily sold broken hardware to consumers and claimed it was an "attack", when it always was "we're cutting corners"."
Ian Cutress from AnandTech points out in a reply that AMD's Ryzen ECC support is not as solid as believed.
Related: Linus Torvalds: 'I'm Not a Programmer Anymore'
Linus Torvalds Rejects "Beyond Stupid" Intel Security Patch From Amazon Web Services
Linus Torvalds: Don't Hide Rust in Linux Kernel; Death to AVX-512
Linus Torvalds Doubts Linux will Get Ported to Apple M1 Hardware
(Score: 0) by Anonymous Coward on Wednesday January 06 2021, @11:34AM (14 children)
It all depends on what gets hit. A single flipped bit in the wrong place can mean a corrupted filesystem and lost data.
(Score: 2) by barbara hudson on Wednesday January 06 2021, @02:27PM (13 children)
Your desktop has a solid metal case? Problem solved for external gamma rays. Any bit flipping will be from the noisy Ed environment inside the case, not gamma radiation.
Now if you're a smoker, you're probably the source of decay particle radiation. Those broad tobacco leaves have been sitting out in the field collecting fallout as they grow, and your lungs are concentrating the particles. It's why, at the beginning of the 20th century lung cancer was so rare that if a case presented in a hospital, doctors on rounds would be told "look carefully, you'll probably never see another case in your career."
That all changed with atmospheric nuclear bomb testing. The fallout from fallout will be with us for another 100 years.
SoylentNews is social media. Says so right in the slogan. Soylentnews is people, not tech.
(Score: 0) by Anonymous Coward on Wednesday January 06 2021, @03:41PM (12 children)
The issue is cosmic rays and this was discovered before cases got to be so cheap.
(Score: 2) by RS3 on Wednesday January 06 2021, @03:54PM (11 children)
Yes, and AFAIK "cosmic rays" (high energy gamma) go through things pretty easily, including steel cases. You need very thick steel, lead, concrete, to significantly reduce them.
Although cosmic ray / random bit-flips are a problem, I think the "rowhammer" problem is what Linus was more unhappy about (and I with him).
(Score: 2) by barbara hudson on Wednesday January 06 2021, @06:54PM (1 child)
A steel case will do a better job than moving from your top floor to your basement with a plastic case ever will. Of course, in the basement you also have problems with radon and radiation from concrete and gravel construction materials. Even granite countertops.
Linus flipped a bit. It's not like the cpu, the various controllers, even the one in your keyboard, are immune. It's really a non issue for most people, so why should everyone pay extra for stuff they don't need that will reduce performance? Getting a bit spoiled in his dotage. If it's important to him, let him pay the extra cost and take the performance hit. He's sounding more like RMS every day - his way is the best way and we should all just do as he says.
SoylentNews is social media. Says so right in the slogan. Soylentnews is people, not tech.
(Score: 2) by RS3 on Wednesday January 06 2021, @07:12PM
All excellent points, thanks. Yes, I'm aware that for example, airline workers are at greater risk for radiation-induced health issues.
My concern, and what I think I'm gathering from Linus' rant, is the shift away from ECC. I think he's bothered by the lack of choice, and that choice being taken away for economic (cheapening) reasons, rather than solid technical / quality improvement reasons. As an engineer, the #1 thing I'm constantly told to put on my resume is how I cheapened something. The good news is that Intel is losing market share.
But I think another factor, at least has been in my professional life, is the people who make the purchasing decisions are generally not very technically-minded, and don't generally ask for technical advice from me or colleagues. They make "business" decisions about the hardware and software, OSes, development tools, etc., and foist it on us underlings. Specifically, when offered the option of ECC RAM, they simply say "oh, it costs more and gives indeterminate benefit? No thanks." So Intel starts cutting it out of their products. Hopefully it hurts them more and more.
But the point is, it's not just Intel, but (idiot) buyers who are driving ECC down.
All that said, as I posted elsewhere in this discussion, there are RAM chip manufacturers making chips with ECC built in. :) Not sure if ECC stats can be gleaned out though- maybe, could be...
https://www.intelligentmemory.com/ECC-DRAM/DDR3/ [intelligentmemory.com]
(Score: 3, Informative) by Immerman on Thursday January 07 2021, @05:36AM (8 children)
FYI cosmic rays are not high energy gamma radiation (electromagnetic radiation), instead they're protons and other atomic nuclei traveling at almost light speed. https://en.wikipedia.org/wiki/Cosmic_ray [wikipedia.org]
(Score: 2) by RS3 on Thursday January 07 2021, @05:53AM (7 children)
Thank you for the correction. I'm a bit confused as to how cosmic rays, being alpha particles? can penetrate things and mess with electronics? I thought only gamma could pass through objects... I guess it's their extreme energy? I see that cosmic rays were originally thought to be gamma, so maybe my brain is in 1920s, like my Model A Ford. :)
(Score: 3, Informative) by Immerman on Thursday January 07 2021, @04:46PM (6 children)
Yeah, seems like there's always another old misunderstanding to relearn, doesn't it? I suppose it keeps things from getting boring.
There's also neutron radiation that can pass through objects since they don't interact with the electron field - but neutrons are so unstable that they decay into hydrogen long before they cross the interstellar void - even at cosmic ray speeds.
Yeah, I think it's all about the energy, aka speed. Cosmic rays are traveling at particle-collider speeds, the really high energy ones absolutely dwarf anything we can produce in the LHC. They're just moving far too fast for the deflective force of electron clouds to have much effect before they're past. Kind of like comparing how far gravity deflects a softball pitch over the course of a single meter, versus how far it deflects a high-speed bullet.
As I understand it, high energy cosmic rays can actually have far more penetrating power than neutron radiation. I think that's probably down to speed again - QM is all probabilistic, so go fast enough and I think you can pass directly through the nucleus itself and come out the other side before the probability of interacting reaches 100%. Or maybe it's that the powerful repulsive effect of the nucleus on a solitary proton is enough to deflect a lot of would-be hits to near misses - a nucleus is an infinitesimally tiny target after all, about half a around one 26,000th the diameter of an atom, or almost a 700 millionth of the cross-sectional area. Possibly it's a combination of both.
(Score: 2) by RS3 on Thursday January 07 2021, @04:55PM (5 children)
Wow, thanks for all of that. Are you a physicist?
I've always been a science enthusiast, but I admit I don't delve in deeply, so I end up with "dangerous" knowledge. For instance, I know a little about nuclear reactors, but I didn't know that neutrons decay into hydrogen. Not sure how I missed that along the way. But it explains the problem with hydrogen in nuclear reactors. That aside, it kind of blows my mind. What a bizarre thing a neutron is! What the heck are we even made of anyway? :)
(Score: 2) by Immerman on Thursday January 07 2021, @05:50PM (4 children)
Nah, just a fellow science enthusiast - but I have a tendency to delve too deep for my own good sometimes, especially around physics. Probably still lots of "dangerous" knowledge mixed in, just a lot deeper into the details than usual.
I don't think neutron decay has anything to do with the hydrogen problem in reactors - I think that's more a matter of superheated water coolant chemically decomposing into H2 and O2. The amount of fuel in a reactor is tiny, and thus the maximum amount of hydrogen that could be produced through neutron decay is similarly tiny. Plus, virtually all emitted neutrons are absorbed into other nuclei, either triggering fission, or transmuting shielding, etc. into heavier isotopes. Free neutrons have a half-life of about 15 minutes - any neutron ejected from a reactor that didn't interact with anything for 15 minutes... would probably have already traveled a long way away from the power plant.
Basically though, beta radiation (an electron) is the result of a neutron within the nucleus decaying into a proton plus electron, with the electron carrying away the excess decay energy in the form of kinetic energy.
(Score: 2) by RS3 on Thursday January 07 2021, @06:17PM (3 children)
Good for you! I do delve into some things, but tend toward the more tactile- things I can take apart by myself with a few tools. :) Notice I didn't say anything about putting them back together again...
I was thinking the same thing- the neutron flux in a reactor wouldn't be enough to create significant hydrogen.
I never really studied neutrons. For whatever reason I had it in my head that neutrons were very stable things, like little lead balls.
I do know that neutrons must be moderated (whatever that really means) in a reactor- if they're too fast, reaction doesn't happen.
Here's a ponderment: do neutrons ever collide with their own electrons when trying to escape?
Another: do the neutrons and protons in a nucleus stay in a stable fixed relative position, or are they moving around and churning?
(Score: 3, Informative) by Immerman on Friday January 08 2021, @01:23AM (2 children)
>For whatever reason I had it in my head that neutrons were very stable things, like little lead balls.
I mean protons are, why not neutrons, right? I mean, you hear "bits of an atom, you figure that';s as far as it goes, right? Then you get into quarks and quantum chromodynamics... and everything you thought you knew goes weird.
I think (don't quote me) that moderation is basically a matter of putting a bunch of nuclei in the way (typically graphite or water molecules) that the neutrons will collide with without reacting. Each collision transfers some of the neutron's kinetic energy to the target, slowing the neutron to thermal speeds so that it can more easily react with more receptive nuclei.
>do neutrons ever collide with their own electrons when trying to escape?
I don't think that question makes sense. So long as a Neutron is a neutron, it doesn't have an electron - at best it'd be like asking if you ever collide with your own bones when trying to run. Except I don't think a neutron contains an electron either. It sounds like a neutron decays into a proton and W- boson (???), and the W- boson then transforms into an electron and electron antineutrino.
Particles physics is weird - properties like spin and charge (including color charge and other subatomic weirdness) are conserved, but particles themselves can convert to and from raw energy pop in and out of existence provided enough energy is available (or not - virtual particles spawn out of nothing, along with their antiparticle, by "borrowing" energy from the quantum vacuum, which is "repaid" when they collide and annihilate shortly thereafter.)
>do the neutrons and protons in a nucleus stay in a stable fixed relative position, or are they moving around and churning?
That's a good question. My gut feeling is that there'd be some churn, but I suspect it'll be a VERY long time before we can actually measure changes small enough to have any real clue as to the answer. Perhaps half-life is related to the amount of "churn" in the nucleus? The more things are moving around, the more likely it is that something will get collide just wrong and get ejected?
(Score: 2) by RS3 on Friday January 08 2021, @02:05AM (1 child)
Thank you for awesomeness!
>do neutrons ever collide with their own electrons when trying to escape?
What I meant was, when a neutron escapes from nucleus, how does it get through the electron clouds without ever colliding with one?
(Score: 2) by Immerman on Friday January 08 2021, @03:14PM
Gotcha. Hmm...I suppose it might - I don't really know QM well enough to say anything for sure. It's probably not very common though. To start with, a classic electron is far smaller than a nucleus, which is itself around 1/10,000th the size of the atom - even with a hundred classical "billiard ball" electrons orbiting the nucleus, the odds that any of them would be directly in the path of a neutron as it comes rocketing out of the nucleus are *extremely* low. And in reality electrons are more of a distributed wavefunction that interacts almost entirely via electric charge, which the neutron doesn't have. I believe neutrons do have spin, which might interact with an electron... but I think spin is mostly factor in when two identical wavefunctions are trying to occupy exactly the same space.