3 ms·

You can't do that until someone has written and tested the code for the "easy defaults". And you can't do that until that someone has hardware in their hands.

You can't do that until someone has written and tested the code for the "easy defaults". And you can't do that until that someone has hardware in their hands. If you want only company insiders to have hardware until every last bit of software is perfected then you can do that ... with your company. And you'll be releasing things to the general public at the same time as the other companies are releasing the next 2x better generation of hardware. Or, you could just yourself personally not buy things until they've been in the market for two years. Heck, I don't update to the latest OS version on my Mac until the next one is on the verge of coming out. I'm running Sonoma and recently it's been nagging me to update to Sequoia. Nuh uh. It's a mature product, progress is slow, and I don't even remember the last time there was a new feature I cared about. It's the same in Windows. Microsoft can't get people to upgrade from Windows 7, let alone Windows 10. RISC-V on the other hand is changing incredibly quickly. There was not even one RISC-V chip or board on the market when Windows 10 came out. There is genuinely exciting new stuff coming out every 6 to 12 months. The P550 boards in the last month, The SpacemiT boards with 8 cores and RVV 1.0 with a decent amount of RAM in the middle of 2024. The 64 core 128 GB RAM Pioneer in January 2024. The TH1520 boards with OoO and high performance RVV 0.7 in mid 2023. The JH7110 boards, the first cheap multi-core, in January 2023. I, and many others, prefer to get new hardware the instant it comes out of the fab in mass production, even if there will always be some rough edges. But you don't have to do that. No one is forcing you. You can wait a little until others (whether in the manufacturer's staff or not) have smoothed off the edges.

phendrenad2 2y ago
> You can't do that until someone has written and tested the code for the "easy defaults" The VF2 has been out for over a year. How long does it take to write a bootloader that can load a kernel from USB? Earlier you said it already exists, so presumably it just needs tested. How long does it take to test? It's probably tested somewhat already. How much testing is "good enough"? The alternative is someone having to flash their own SD card to boot from USB. The bar to be better than that seems very low, no? > If you want only company insiders to have hardware until every last bit of software is perfected then you can do that If Microchip can ship me a microcontroller flashed with my custom firmware I wrote the day before, surely RISC-V board manufacturers can provide updates to their firmware after the fact, certainly a year later. > RISC-V on the other hand is changing incredibly quickly. There was not even one RISC-V chip or board on the market when Windows 10 came out. That was 10 years ago. Roughly the time period between Windows 3.1 and Windows XP. Are we really clinging to that tired excuse? Admit it or not, "RISC-V is changing rapidly" simply isn't true anymore. The ISA is done. There are some vector extension things that remain in beta, or have possibly just left beta, but you don't really need vector extensions to boot from USB, nor should you rely on them to do so. > I, and many others, prefer to get new hardware the instant it comes out of the fab in mass production Yeah and that's great, I'm just saying that maybe, after a year, it makes sense to release easier firmware. It can only help your chances of getting mass RISC-V adoption, right? I don't understand RISC-V fans who advocate for positions that only hurt the ecosystem. > But you don't have to do that. No one is forcing you. You can wait a little until others (whether in the manufacturer's staff or not) have smoothed off the edges. Who's smoothing out the edges? Maybe the wrong edges are being smoothed, and I'm just asking for a little re-allocation of smoothage.
- brucehoult 2y ago
  > The VF2 has been out for over a year. How long does it take to write a bootloader that can load a kernel from USB? Manpower is extremely limited -- unless of course you're offering to do that part. What would you like someone to not do so that they can do USB booting instead? > It can only help your chances of getting mass RISC-V adoption, right? There is not yet any RISC-V hardware in production that has performance suitable for mass adoption in laptops/desktops, just as there is none in the Arm world except from Apple. Maybe the Radxa Orion O6 and/or Qualcomm's Oryon/Snapdragon X Elite, I don't know as I haven't tried them yet. The Pi 5 / Rock 5 / Orange Pi 5 generation are kind of ok but not quite there as an x86 replacement. RISC-V is two or three years behind (non Apple) Arm. Pushing people into RISC-V before it is ready is what would be counter-productive. Maybe booting from USB is part of that, but general app performance is far more important. It's coming, but the Android phone/tablet market is both far larger and much easier and will happen first. No one expects their Android phone to boot from a USB stick. > The ISA is done. There are some vector extension things that remain in beta, or have possibly just left beta Showing how little you know about the current status. The Vector extension was ratified three years ago in November 2021. Hasn't been beta for ages. But there are lots of other very important things needed for feature parity with Arm and x86 which have just been ratified in 2024, culminating with RVA23 in December. That is not "finished" by any means, just as Arm and Intel/AMD continue to add things to their ISAs, but it's a good base for an ISA spec allowing general adoption. Chips implementing RVA23 will be out in two or three years AND should simultaneously hit a performance level that won't disappoint users AND all important software should be available at the same time. I expect including your USB booting. Waiting is...
  - phendrenad2 2y ago
    > Manpower is extremely limited And that's why this needs to be solved at a higher level than engineering. There's clearly ample investment into RISC-V - Tenstorrent just raised $693 million. Convincing investors to throw hundreds of millions at "AI magic" is easy. Convincing them to throw $50k at a contractor to make a USB bootloader... that's a harder sell - even if that small investment would have a huge impact on the public perception of RISC-V overall. I think this idea is wasted on "Hacker" News, I need to find "Tech Investor" News, but it doesn't seem to exist. :P

phendrenad2 2y ago

> You can't do that until someone has written and tested the code for the "easy defaults" The VF2 has been out for over a year. How long does it take to write a bootloader that can load a kernel from USB? Earlier you said it already exists, so presumably it just needs tested. How long does it take to test? It's probably tested somewhat already. How much testing is "good enough"? The alternative is someone having to flash their own SD card to boot from USB. The bar to be better than that seems very low, no? > If you want only company insiders to have hardware until every last bit of software is perfected then you can do that If Microchip can ship me a microcontroller flashed with my custom firmware I wrote the day before, surely RISC-V board manufacturers can provide updates to their firmware after the fact, certainly a year later. > RISC-V on the other hand is changing incredibly quickly. There was not even one RISC-V chip or board on the market when Windows 10 came out. That was 10 years ago. Roughly the time period between Windows 3.1 and Windows XP. Are we really clinging to that tired excuse? Admit it or not, "RISC-V is changing rapidly" simply isn't true anymore. The ISA is done. There are some vector extension things that remain in beta, or have possibly just left beta, but you don't really need vector extensions to boot from USB, nor should you rely on them to do so. > I, and many others, prefer to get new hardware the instant it comes out of the fab in mass production Yeah and that's great, I'm just saying that maybe, after a year, it makes sense to release easier firmware. It can only help your chances of getting mass RISC-V adoption, right? I don't understand RISC-V fans who advocate for positions that only hurt the ecosystem. > But you don't have to do that. No one is forcing you. You can wait a little until others (whether in the manufacturer's staff or not) have smoothed off the edges. Who's smoothing out the edges? Maybe the wrong edges are being smoothed, and I'm just asking for a little re-allocation of smoothage.

brucehoult 2y ago
> The VF2 has been out for over a year. How long does it take to write a bootloader that can load a kernel from USB? Manpower is extremely limited -- unless of course you're offering to do that part. What would you like someone to not do so that they can do USB booting instead? > It can only help your chances of getting mass RISC-V adoption, right? There is not yet any RISC-V hardware in production that has performance suitable for mass adoption in laptops/desktops, just as there is none in the Arm world except from Apple. Maybe the Radxa Orion O6 and/or Qualcomm's Oryon/Snapdragon X Elite, I don't know as I haven't tried them yet. The Pi 5 / Rock 5 / Orange Pi 5 generation are kind of ok but not quite there as an x86 replacement. RISC-V is two or three years behind (non Apple) Arm. Pushing people into RISC-V before it is ready is what would be counter-productive. Maybe booting from USB is part of that, but general app performance is far more important. It's coming, but the Android phone/tablet market is both far larger and much easier and will happen first. No one expects their Android phone to boot from a USB stick. > The ISA is done. There are some vector extension things that remain in beta, or have possibly just left beta Showing how little you know about the current status. The Vector extension was ratified three years ago in November 2021. Hasn't been beta for ages. But there are lots of other very important things needed for feature parity with Arm and x86 which have just been ratified in 2024, culminating with RVA23 in December. That is not "finished" by any means, just as Arm and Intel/AMD continue to add things to their ISAs, but it's a good base for an ISA spec allowing general adoption. Chips implementing RVA23 will be out in two or three years AND should simultaneously hit a performance level that won't disappoint users AND all important software should be available at the same time. I expect including your USB booting. Waiting is...
- phendrenad2 2y ago
  > Manpower is extremely limited And that's why this needs to be solved at a higher level than engineering. There's clearly ample investment into RISC-V - Tenstorrent just raised $693 million. Convincing investors to throw hundreds of millions at "AI magic" is easy. Convincing them to throw $50k at a contractor to make a USB bootloader... that's a harder sell - even if that small investment would have a huge impact on the public perception of RISC-V overall. I think this idea is wasted on "Hacker" News, I need to find "Tech Investor" News, but it doesn't seem to exist. :P