• 0 Posts
  • 185 Comments
Joined 1 year ago
cake
Cake day: July 29th, 2023

help-circle
  • I’d say about 99% is the same.

    Two notable things that were different were:

    • Podman config file is different which I needed to edit where containers are stored since I have a dedicated location I want to use
    • The preferred method for running Nvidia GPUs in containers is CDI, which imo is much more concise than Docker’s Nvidia GPU device setup.

    The second one is also documented on the CUDA Container Toolkit site, and very easy to edit a compose file to use CDI instead.

    There’s also some small differences here and there like podman asking for a preferred remote source instead of defaulting to dockerhub.


  • Yeah I should have mentioned the context is FBLA, and Google partially fixed the prompt.

    Original from a few weeks ago:

    BPA is another student org called Business Professionals of America

    The AI ignores the subject context and just compares whatever is the most common acronym.

    They lazy patched it by making the model do a subject check on the result, but not on the prompt so it still comes back with the chemical lol.




  • iirc due to some anti trust lawsuits, they cannot do that anymore.

    But it’s still easy to coerce OEMs to run Windows because they offer stuff like quick support and standardized IT support.

    If an OEM ships Linux, they don’t want to have to make an entire department to help troubleshoot the OS for users who will inevitably call for help. Ignoring them would only result in returns and loss of sales.

    I think some thinkpads actually do ship with some distro like redhat or opensuse as an option, but that’s because thinkpads are very popular in the business space which means lots of CS people use them, so it helps save some cost from a windows license that won’t get used.

    Like I said though, if windows really dives into the deep end, I think a potential market would open and some OEM will take a chance on it.


  • Not to be that guy but why not use Curve25519?

    I still remember all the conspiracies surrounding NIST and now 25519 is the default standard.

    In 2013, interest began to increase considerably when it was discovered that the NSA had potentially implemented a backdoor into the P-256 curve based Dual_EC_DRBG algorithm.[11] While not directly related,[12] suspicious aspects of the NIST’s P curve constants[13] led to concerns[14] that the NSA had chosen values that gave them an advantage in breaking the encryption.[15][16]


  • There’s plenty of videos on YouTube of people trying Linux for the first time, and it can be painful to watch how poorly they try to fix something or unintentionally break their system.

    That’s not to say windows is any better, because they’d do the same thing there.

    But people will only switch permanently if windows really falls off hard, which may or may not happen.

    You have to think of it like how people first learned to use a mouse and double click back in the 90s. It’s not immediately intuitive for everyone, they often have to start over.

    That being said, having a big OEM ship linux would do wonders, but Microsoft fights hard to make sure that almost never happens.


  • Yeah that means the driver is loaded fine, but it looks like it is selecting the iGPU by default. You have several options to fix this.

    1. You can disable integrated graphics in the bios if there is an option for it. This is the easiest, but if you’re on a laptop, leaving it enabled might save some battery in which case goto 2.

    2. You can tell either each program or the OS to prefer the Nvidia GPU. The way you do this also depends on how the gpu is set up (most laptops have it as secondary)

    You can test this by running __NV_PRIME_RENDER_OFFLOAD=1 __GLX_VENDOR_LIBRARY_NAME=nvidia glxgears in one terminal, and nividia-smi in a second terminal to verify a program (in this case glxgears) is running on the nvidia gpu.

    I’ll try to find a good guide, but depending on the setup, it could be a simple MUX switch you can flip to change between iGPU and Nvidia GPU, or with the use of some preference selector tool (I think it was called prime?).

    It’s confusing because lots of laptops essentially use the Nvidia GPU as offload which makes it a bit tricky to coaxe it into using the correct one.



  • It’s not 90s tech though, especially for China.

    Their latest x86 CPU is comparable to Kaby Lake in cycle speed which is only 8 years old, except it comes with more cores and supports DDR5 so it might as well be a first gen ryzen 7.

    They still haven’t revealed how they fabricated it or what process they used, probably because they want to keep the production chain and size a secret.

    Enriching uranium and making nukes, in comparison, is banging rocks together.

    No it isn’t, especially for weapons grade Uranium. Look at Iran, they’ve been perpetually “10% away from a bomb” for more than 20 years and still haven’t succeeded.

    The ridiculously high precision required to make the centrifuges, and then the scale required to make hundreds of thousands of them per plant just to reach 20% enrichment is insane.

    Reaching 90% is like taking all that and ramping it up several hundred times.

    The only reason Pakistan succeeded was because they got (stole) the critical design parameters needed for the centrifuges to work, and a rather brilliant metallurgist who took several years to figure out how to manufacture the centrifuges consistently at scale. Plus an entire set of physicists just to figure out the centrifuge physics in a way that would allow them to maximize refinement with dozens of design variables. It still took them a decade, but they eventually got it.

    It’s a pretty good comparison to lithography machines which requires similar dead precision with each decreasing size of transistor requiring an order of magnitude more precision in quality engineering.

    Also I don’t think the US is involved in this, at least not directly:

    I doubt it because they’ve been making it a pretty big deal for the past 4 years. Tons of Chinese tech OEMs are blacklisted, and the trade war keeps escalating with new bans/tariffs/exclusions every year. Plus they dumped billions of dollars into intel and TSMC in a desperate attempt to make a fab on the home front.

    It doesn’t matter that it’s DUV, they just want to ensure they make it harder for China to catch up, so even last gen tech is on the line because they believe it can be studied and reverse engineered.

    imo it’s a stupid shortsighted policy, but it’s nothing new for the US pulling these types of moves. I just wish for once they’d see that it’ll only delay the inevitable, and maybe they should put that effort into actually making quality products at home instead of throwing money at chip OEMs and expecting them to move out of Taiwan overnight.


  • People here (including the US govt apparently) acting like it’s actually going to take China a decade to figure out how to run a wafer machine bruh.

    Not only do they probably already have the procedures written down and kept safe, they’ve been already been experimenting with having to run the entire supply chain on their own for years now. Hell they’re even the ones basically carrying RISC-V development right now because they barely have OEM access to x86.

    And that’s all without the assumption that China hasn’t stolen some key trade secrets that would give them a head start. I highly doubt this equipment will actually go offline besides some practice runs and research application which they have likely already done without telling anyone.

    Pakistan’s entire nuclear arsenal only exists because one talented due working at URENCO (also coincidentally Dutch like ASML) took a few hundred documents and his years of work experience back to his home country. If broke ass Pakistan could figure out how to make fissile material and nukes in their backyard, China sure as hell gonna figure out how to fabricate chips without any external suppliers or contractors.






  • I should have added this in my original comment, but the top 3 feature requests on the crappy insider hub were dark mode, aero/glass effect, and getting rid of redundant or poorly made UWPs (Settings, video player, pictures, etc).

    In typical Microsoft fashion, they took 5 years to make functioning dark mode, told us aero/glass is impossible because of battery drain before adding it to only the edge browser, and promised “Settings” app would have all the stuff migrated from control panel “soon”.

    It’s been 9 years and they still haven’t migrated everything because turns out the control panel actually runs like a normal application that’s easy to use.

    They pulled the same stunts with Windows 11, which I still have left on a spare laptop on the insider release channel, and that’s when I noped out after they basically ignored almost all of the top requests with really stupid reasoning.





  • The OS is but the Google Play backend isn’t. Google has a monopoly over Android by keeping a monopoly on the appstore which dictates that you must allow google spyware to run on your OEM fork to be able to qualify as a “Secure Device”.

    Several Chinese OEMs have China only variants that don’t use GPlay and also ship with some some other cool apps, but they can’t sell it globally because Google says “screw you” since no one publishes apps outside Gplay, and because several major apps refuse to run on Googless android which GrapheneOS has threatened to sue

    This is still just the tip of the iceberg though. Google already got sued for GPlay monoply last year and reached a $700 settlement just for developers.

    On top of that, several of Android’s underlying features are considered archaic and dated. They always have huge kernel patching issues because no OEM (especially Qualcomm) releases the source code for proprietary binaries, meaning no one easily upgrade kernels (practically impossible for FOSS android, expensive for OEMs). The android runtime is imo a piece of crap compared to some low power optimized linux distros. ADB is still needed to delete system apps. Settings lies about permissions, which themselves are poorly sorted. Oh and Google hired the dev behind Android rooting (magisk) so they could kill magisk hide which circumvents system app abilities to tell if you are rooted and therefore not worthy of running proprietary apps.

    There’s so much more the deeper you go, it’s just really hard for any contender to step up because of the sheer might Google has over the market. They have so much power that they coerced Samsung into dropping RCS support which makes Google Messages the only app on android that supports RCS, even though RCS is an open OEM standard from 2008