315
Europe built sovereign clouds to escape US control. Then forgot about the processors (www.theregister.com)
submitted 3 months ago by to c/technology@lemmy.world
55 comments fedilink hide all child comments
you are viewing a single comment's thread
view the rest of the comments
[–] 28 points 3 months ago* (3 children)

It said RISC-V is decades away

"There is no immediate solution. RISC-V, the open source processor architecture European sovereignty advocates point to as a long-term alternative, remains years from competitive performance in datacenter workloads. "It will take decades,""

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 6 child comments
    •
  • [–] 19 points 3 months ago (1 child)

    RISC-V isn't in the same scenario. There's one company behind ARM with a few external companies with architecture licenses (who doesn't share their contributions), and ARM competes mostly just on the same commercial terms so for a long time it wasn't worth investing in single core performance because they could instead fill the efficiency niche.

    Also there's more knowledge on how to build high performance cores. Doesn't mean it's trivial, but it means the lead isn't several decades. With enough investment you can make it happen faster. And there's a national security motivation for investing.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] 15 points 3 months ago (1 child)

    RISC-V is more like 1-3 years away from CPUs existing that have competitive performance in datacenter workloads. Not decades.

    But they won't be manufactured in Europe. Getting fabs up and running is indeed something that takes a very long time.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] 6 points 3 months ago

    RISC-V is more like 1-3 years away from CPUs existing that have competitive performance in datacenter workloads. Not decades

    I’ve been hearing this for the past five years.

    People seem to forget that if one arch moves forward, so do every single competitor out there.

  • permalink
  • fedilink
  • source
  • parent
  • [–] 5 points 3 months ago (1 child)

    RISC-V is decades away

    Eh... what? I have a RISC-V SBC and it just works, running Debian on it in minutes of setup and it cost me peanuts.

    Sure it's not a state of the art CPU ... and if I wanted to run anything demanding on it, I'd have to be patient. Heck it's not even made in the EU but in China... but it works, today, it just depends on what your workload is. So yes it's not the fastest or has the best efficiency but still, it exists already.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] 4 points 3 months ago (1 child)

    What does that tell you about its performance under datacenter workloads?

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] 1 point 3 months ago* (1 child)

    Nothing because it depends on the workload? I mean if you run a static Website to few people it's more than enough. If you're trying to predict weather or render high definition 3D graphics in real-time it's not... but also nothing is so...

    Was it a rhetorical question and if so what were you implying?

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] 1 point 3 months ago (1 child)

    It was a rhetorical question. I wish all good for that architechture, but it doesn’t seem very competitive as for now.

    https://ben3d.ca/blog/risc-v-in-2024-is-slow

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] 1 point 3 months ago* (1 child)

    Depends entirely on the metrics you use for comparison. In terms of performances yes of course it's slower than others, nobody is contesting that. In terms of openness it fairs better than most. My point was solely that it's usable for some use cases and thus that it's not a theoretical architecture in 2026. It works. Yes it's slow but for use use cases it doesn't matter.

    If you don't care for openness then it's not competitive. Being competitive depends entirely on your constraints.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] 1 point 3 months ago (1 child)

    The case was described in:

    "There is no immediate solution. RISC-V, the open source processor architecture European sovereignty advocates point to as a long-term alternative, remains years from competitive performance in datacenter workloads. "It will take decades,""

    To which you replied with ”Eh.. What?” and went on to tell an anecdote about how it works well on a personal computer running linux. It doesn’t really relate to the problem in hand here although is neat.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] 1 point 3 months ago* (last edited 3 months ago) (1 child)

    Because those components are (theoretically) sold as equivalent. If you sell me cycles in a data center, one for 10e/h and another for 100e/h (because it's 10x slower and thus must have ~10x more instances) and you don't give me any details on why, I'll take the 10e and of course it won't be competitive. FWIW I do buy compute time in data centers and I'm also aware (but not involved with) https://www.top500.org/ and how none of them are RISC-V based, it's not my point. My point is that the metrics to compare will never make it competitive if we exclude its raison d'etre. RISC-V was never proposed to be the most efficient and powerful architecture (even though of course it'd be nice if it'd be).

    It's like apple versus orange then complaining that the apple doesn't taste orange-like enough. Sure, that's correct, but also pointless.

    Edit : it's not an "anecdote" it's a proof of existence, again RISC-V works today. It's not set of blueprints. It does compute, easy as that.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] 1 point 3 months ago (1 child)

    So even you know it’s not being used to any scale in data centers. In theory they could run on Raspberry PIs but that’s not useful to the problem in hand. Where did you get 10x slower by the way?

    An anecdote does not imply that something is not real.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] 1 point 3 months ago

    Starting to worry we're talking past each other.

    Yes, RISC-V isn't used at scale in data centers. Now though that NEW criteria are taking into account, namely sovereignty, they precisely might despite their limitations, including performances. If though it's just political signalling without any actual will and subsequent advantage and in reality only performance matters, they still won't be used.

  • permalink
  • fedilink
  • source
  • parent