• 0 Posts
  • 104 Comments
Joined 3 years ago
cake
Cake day: June 17th, 2023

help-circle


  • Intel Arc is actually decent from a power efficiency standpoint, and they were coming out with cards that had good amounts of memory for home or workstation AI before prices went nuts. My A770 actually does pretty decently at mid models etc and I’d love to get my hands on a B70.

    AMD was never really competitive at the top-end, but their APU’s are very slick and they did have a decent power-per-priced on GPU’s. Even in the ATI days they often made good hardware but the driver stack was abysmal. On non-Windows platforms they’ve been awesome for a long time though, especially purpose-built APU platforms in gaming systems (PS4/5, Xbox after 360, Steam Deck, etc).

    Nvidia always kinda tried to capture the “elite” market and went for top dollar, but part of that is marketing given what people were actually doing in terms of gaming.



  • Because typically the resources for a particular AI job need to be centralized in order to obtain usable speeds. Even a machine with a decent 16GB GPU is going to lose speed swapping in and out of system RAM - though it does allow for larger models - so machines in a LAN can’t currently split the workloads without significant efficiency losses. It’s also why AI boards are typically using soldered LPDDR rather than swappable DIMM’s (can’t get an AI CPU with a non-soldered board).

    Peers over the Internet would be abysmally slow or untenable for many workloads, at least with the way a lot of stuff is built currently.

    HOWEVER, that’s assuming monolithic non-specialized/optimized models. In the future as we see models built and trained for more specific jobs, it should be possible to have individual students run more as specialists and swap data. There are models that are specialized currently but not really an ecosystem for them:

    • General LLM/Chat or
    • Image recognition
    • Image generation
    • 3D Spatial navigation/manipulation
    • Code generation/processing
    • Network analysis
    • Audio processing/generation
    • Etc

    If we think of something like creating an MMO (and assuming we didn’t include the moral issues of AI generation), then a pipeline could have different assets building a soundtrack, 3d models, scenes/animation per a script, character voicing, various dialog, 3d world space mechanics, etc all running from different clustered machines







  • For people that promote one over the other, my answer is always “why not both”?

    Areas that are good for nuclear are likely not as good for solar or wind. The latter two are generally more sustainable, but also more temperamental towards environmental conditions for output.

    So my thought is, built all three in the locations best suited to whatever, and also build to storage capacity where possible to manage peak/off-peak loads.







  • It’s a bit of both. Yes, Republicans have a bad track record in general, bit also has they’ve gotten worse and more blatant about it.

    I don’t know if this guy is a “Republican who voted in 2024 and now regrets the choices coming home to roost” or “somebody who has been a long time Republican but at least had enough brains not to vote for the racist goon squad more recently”, but at least there’s some growing awareness there.

    The Republican party is gone. Now you’ve got the Appeasement party and the MAGA party and both could really benefit from jettisoning a bunch the members that have pushed them towards such and away from the people.