With AI seeming to consume all resources for hardware, I’m wondering what parts of those current systems we could see trickling down into componentry for desktop PC’s as they get outdated for AI tasks.

I know most of this hardware is pretty specific and integrated, but I do wonder if an eventual workaround to these hardware shortages are through recycling and repurposing of the very systems causing the shortage. We have seen things like dram, flash, and even motherboard chipsets be pulled from server equipment and find its way into suspiciously cheap hardware on eBay and AliExpress, so how much of the current crop of hardware will turn up there in the future?

How much of that hardware could even be useful to us? Will nvidia repo old systems and shoot them into the sun to keep it out of the hands of gamers? Perhaps only time will tell

  • bizarroland@lemmy.world
    link
    fedilink
    English
    arrow-up
    3
    ·
    13 hours ago

    You can already get SXM2 adapters and external boards that run to PCI-E X16 slots.

    I wouldn’t say it’s the smartest way to blow a thousand dollars, but you can add a few of the older, like, 2019 models to a computer for about a thousand dollars already.

    SXM4, I believe, has also been cracked, but it’s a lot more expensive, and I’m sure SXM5 will not be too far behind. The main difference between the PCI-E and SXM models as far as I’m aware, other than their interconnect and the built-in tethering between the multiple GPUs, is that they are able to run on 48 volt power, which means that the amperage running through the wires is much lower, and you’re less likely to cause everything to burst into flames randomly.

    • jj4211@lemmy.world
      link
      fedilink
      English
      arrow-up
      4
      ·
      7 hours ago

      Yes, they connect by PCIe and thus the physical mismatch may be overcome, but they also are now drawing 15kw. More wattage than any circuit in my residential breaker box can handle.

      Even if you did, there’s not even a whiff of driving circuitry for a video port, so your only application would be local models, and if the bubble bursts, well that would seem to indicate that use case would be not that popular.

      No I would expect that these systems get rented out of sold to supercomputer concerns for super cheap if a bubble pop should occur.