11 ms·

This is super neat, but it serves to remind me of the difficult (as in decision-making) position I'm in. I recently refreshed so have a few years to go until i

This is super neat, but it serves to remind me of the difficult (as in decision-making) position I'm in. I recently refreshed so have a few years to go until it's time again, but I would love to go RISC-V, but I want high performance that is going to last me several years. For comparison sake, right now my desktop has an AMD 5950x 16-core (32v) in it, and my Framework 16 laptop has a Ryzen 7 7840HS (8-core (16v)). The reason this matters is that I expect to spend around $2k for a machine, but if I'm going to spend that kind of money, it's going to be my only workstation. I suspect it will take some time for RISC-V chips to get anywhere near competitive. But if I assume everybody is like me, that presents a real chicken-and-egg problem for high performance RISC-V laptops.

throw8076987234 2y ago
Since you own a Framework in particular, I think a more likely scenario is that 2-4 years from now you have the option to upgrade your mainboard to the latest Intel/AMD x86 silicon, or to the latest Qualcomm Snapdragon/Oryon ARM platform. I can imagine the next upgrade 2-4 years after that maybe having a RISC-V option.
- freedomben 2y ago
  Indeed, that's a really great point, and the timeline seems pretty reasonable as well.
brucehoult 2y ago
I've got a similar spec laptop with 24 core (8P + 16E) i9-13900HX that is probably quite comparable or maybe a bit better especially on burst workloads e.g. building software as the max turbo is 500 MHz higher and both similar all-cores boost (3.8-3.9 or so) but I have more cores. It's going to be a long time before we see RISC-V at this level, at least on a per-core basis. I think we'll see around 2020 x86/Apple M1 performance in 2026/27. Late this year we'll see a 16 core RISC-V with around Nehalem or Sandy Bridge performance. But 16 core, so something like a Xeon of that generation -- I had one in a Mac Pro back in 2009. That's a pretty respectable computer. Milk-V says they'll have a board with one for $120 -- I suspect that's with zero or very low RAM, no storage. SiPeed says $300 and that's probably with a good chunk of RAM and eMMC. That will be RISC-V's RK3588 (Rock 5, Orange Pi 5) or Raspberry Pi 5 moment, though it actually leapfrogs those by quite a bit, in both per core speed and number of cores.
- freedomben 2y ago
  Thanks, I think your timeline makes a lot of sense. Considering that, my gut tells me it may be 15 to 20 years before RISC-V catches up (since Intel and AMD and ARM won't be laying down). What do you think the timeline looks like? Do you think RISC-V will catch up anytime within the next 15 to 20 years? Is it more optimistic than that? If a company like AMD spun up a RISC-V division (oh god that would be so amazing), that would obviously greatly accelerate the timetable. I lack a whole lot of strategic info for decision making, but my instinct is to think that could be very good for AMD. There would be some cannabilization, but I don't think nearly as much as you would think, but the big payoff is that if RISC-V is going to be big (which I think it is), they become the de-facto leaders and can own it, much the way Nvidia did with GPUs.

throw8076987234 2y ago

Since you own a Framework in particular, I think a more likely scenario is that 2-4 years from now you have the option to upgrade your mainboard to the latest Intel/AMD x86 silicon, or to the latest Qualcomm Snapdragon/Oryon ARM platform. I can imagine the next upgrade 2-4 years after that maybe having a RISC-V option.

freedomben 2y ago
Indeed, that's a really great point, and the timeline seems pretty reasonable as well.

brucehoult 2y ago

I've got a similar spec laptop with 24 core (8P + 16E) i9-13900HX that is probably quite comparable or maybe a bit better especially on burst workloads e.g. building software as the max turbo is 500 MHz higher and both similar all-cores boost (3.8-3.9 or so) but I have more cores. It's going to be a long time before we see RISC-V at this level, at least on a per-core basis. I think we'll see around 2020 x86/Apple M1 performance in 2026/27. Late this year we'll see a 16 core RISC-V with around Nehalem or Sandy Bridge performance. But 16 core, so something like a Xeon of that generation -- I had one in a Mac Pro back in 2009. That's a pretty respectable computer. Milk-V says they'll have a board with one for $120 -- I suspect that's with zero or very low RAM, no storage. SiPeed says $300 and that's probably with a good chunk of RAM and eMMC. That will be RISC-V's RK3588 (Rock 5, Orange Pi 5) or Raspberry Pi 5 moment, though it actually leapfrogs those by quite a bit, in both per core speed and number of cores.

freedomben 2y ago
Thanks, I think your timeline makes a lot of sense. Considering that, my gut tells me it may be 15 to 20 years before RISC-V catches up (since Intel and AMD and ARM won't be laying down). What do you think the timeline looks like? Do you think RISC-V will catch up anytime within the next 15 to 20 years? Is it more optimistic than that? If a company like AMD spun up a RISC-V division (oh god that would be so amazing), that would obviously greatly accelerate the timetable. I lack a whole lot of strategic info for decision making, but my instinct is to think that could be very good for AMD. There would be some cannabilization, but I don't think nearly as much as you would think, but the big payoff is that if RISC-V is going to be big (which I think it is), they become the de-facto leaders and can own it, much the way Nvidia did with GPUs.