• 0 Posts
  • 179 Comments
Joined 2 years ago
cake
Cake day: February 16th, 2024

help-circle
  • Yeah the utilities area is downright scary. Usually run on thin margins, really old specialists that are rapidly retiring, no IT budget to speak of, and most mission critical tech is 20+ years old, off the shelf from companies long dead, and loosely networked by people who’s objective was “make it work”, not security or resilience.

    The only way out- electrical companies basically HAVE to become public like all the other utilities, and funding, wages, job training, and modernization for all utilities needs to be prioritized massively.

    Of course, there’s lots of political reasons for idiots to fight against all this, so it won’t happen until the worst already has.


  • kata1yst@sh.itjust.worksto3DPrinting@lemmy.worldWhich 3d printer would you recommend?
    link
    fedilink
    English
    arrow-up
    16
    arrow-down
    1
    ·
    edit-2
    23 days ago

    Prusa undisputed #1 for their open and honest approach by reputable people who care, followed, in my opinion, by Snapmaker. Snapmaker is running Klipper with (mostly) open source stuff layered on top. They do have a few binary modules that run alongside, but overall an open ecosystem and a company that seems to thrive in it.

    Sovol in third for me. They’re very much just following what other companies and the voron project are doing, but they haven’t fucked up too massively yet, basically open source software, and everything in the printer is commodity, off the shelf hardware.








  • There was a lot more to it than just the fab failures.

    Full disclosure, never worked there, have shared many beers with fantastically interesting (and proudly deranged) hardware engineers and architects who did.

    Intel had transitioned from an engineering company to a company of managers, spreadsheets, and economic forecasting. They completely stagnated from a technology perspective, not just in the fab but also in the actual design and implementation. A Core 2 Duo was basically their last gasp of real innovation- after that even the business class cpus were basically stuck at the same core counts, same thread architecture, same memory architecture. They were just slowly shaving minuscule IPC, increasing minuscule clock, and adding small amounts of cache… but from a Wall Street perspective they were still posting good profits since the market was cornered and so engineering cuts kept coming, despite staff and principal engineers leaving in droves and yelling that the end was nigh.

    The basic fault was that they decided to build their engineering, economics, and strategies around Moore’s Law. The law that famously was a fun observation stupid people started taking seriously. ‘When will get 12nm, Johnson?’ ‘Forecasted for next Q2 boss!’ but, you know, without the fab architects actually having a clue how to get there without a moonshot project.

    Ryzen was a rude awakening for a company that had become an MBA zombie, slowly shambling towards the next quarter. Since that moment they’ve decapitated basically their entire management suite. And of course replaced them all with new managers. But they actually have managed (heh) to play decent catch up on the engineering side. But as it goes in the industry, 6 years later is 7 years too late. It’s AMD’s game to lose now.









  • I used ps5 controllers. Both developed stick drift within a year of light use. They’re beautiful and so nice to hold and use, but don’t waste your money.

    I recommend Gulikit TT Pro.

    Good ergonomics. Repairable (made by the company who got famous selling replacement Hall effect sticks for Nintendo switch controllers). Built like tanks with proper microswitches and Hall effect sticks. Inexpensive too, especially compared to the controllers that have similar features.

    Best of all, work out of the box on Linux.


  • I respect your perspective and personal experience, and I’m not trying to convince you (I’m not even downvoting you, as it’s not a disagreement button). I’m trying to convince whoever might come along and read this that the small extra price is worth it if their computer is going to hold data dear to them and be running 24x7.

    ECC is extremely good at covering cosmic ray bitflips, which happen with extreme regularity on software that runs and modifies data on the fly- server software. Yes, even home run stuff. That’s just playing Russian roulette, it probably won’t break anything, but why take the risk at least 10 times every day?

    It’s also great at catching failing RAM sticks and preventing them from doing horrible things to every bit of data running through them. This is the failure ECC caught for me at home twice.

    I have only 2.5 decades in the enterprise server and software space, I won’t claim to your 4. But I know I wouldn’t take that risk at work, and I value my home data more rather than less.

    I’m not a researcher or even a particularly well practiced rhetorician, so here’s probably a much more convincing argument.

    Or this perhaps.


  • I’ll strongly disagree. Anyone who cares about the data they store in their server should care about ECC. There’s a specific reason it’s used so widely by servers, not just financial databases or whatever.

    There’s also a ton of misinformation on the Internet about it, so don’t buy into the “ZFS write hole” or whatever. But ECC is very important in my experience. It’s saved my bacon in a material way twice now, and in ways that normal RAM would have just silently continued breaking things. Really that’s not much of a price premium if you’re willing to buy used, so it’s more a question of why not?

    There are many computers (especially business line computers including low power SFF) that will take ECC or even ship with it from ebay or whatever. Or you can build rigs with ECC, I’ve done this route twice and had good results.