Become a fan of Slashdot on Facebook

 



Forgot your password?
typodupeerror
×
AI Intel

"We Asked Intel To Define 'AI PC.' Its reply: 'Anything With Our Latest CPUs'" (theregister.com) 35

An anonymous reader shares a report: If you're confused about what makes a PC an "AI PC," you're not alone. But finally have something of an answer: if it packs a GPU, a processor that boasts a neural processing unit and can handle VNNI and Dp4a instructions, it qualifies -- at least according to Robert Hallock, Intel's senior director of technical marketing. As luck would have it, that combo is present in Intel's current-generation desktop processors -- 14th-gen Core, aka Core Ultra, aka "Meteor Lake." All models feature a GPU, NPU, and can handle Vector Neural Network Instructions (VNNI) that speed some -- surprise! -- neural networking tasks, and the DP4a instructions that help GPUs to process video.

Because AI PCs are therefore just PCs with current processors, Intel doesn't consider "AI PC" to be a brand that denotes conformity with a spec or a particular capability not present in other PCs. Intel used the "Centrino" brand to distinguish Wi-Fi-enabled PCs, and did likewise by giving home entertainment PCs the "Viiv" moniker. Chipzilla still uses the tactic with "vPro" -- a brand that denotes processors that include manageability and security for business users. But AI PCs are neither a brand nor a spec. "The reason we have not created a category for it like Centrino is we believe this is simply what a PC will be like in four or five years time," Hallock told The Register, adding that Intel's recipe for an AI PC doesn't include specific requirements for memory, storage, or I/O speeds. "There are cases where a very large LLM might require 32GB of RAM," he noted. "Everything else will fit comfortably in a 16GB system."

This discussion has been archived. No new comments can be posted.

"We Asked Intel To Define 'AI PC.' Its reply: 'Anything With Our Latest CPUs'"

Comments Filter:
  • The hard part with AI is training. Once you've trained the neural network, it's relatively easy to run it, but training can take massive clusters that can run into the millions or even cross into billions in extreme cases.

    If course, everyone is hyping this now, but there isn't that much out there that people actually want to run on their local computer just yet. Too much is web based where your don't need any new resources, and the companies gathering the data would like to keep it that way.

    • Depending on the model, running it can be tricky. If the model takes hundreds of GB of memory and the access pattern is essentially random, most systems won't have an easy time running it. Also, most models are going to need to change as they're either not perfect or there's a desire for them to be able to incorporate new information or the preferences of their users, which necessitates hardware that can handle doing that.
      • Depending on the model, running it can be tricky. If the model takes hundreds of GB of memory and the access pattern is essentially random, most systems won't have an easy time running it. Also, most models are going to need to change as they're either not perfect or there's a desire for them to be able to incorporate new information or the preferences of their users, which necessitates hardware that can handle doing that.

        Translation: An otherwise still-worthless feature will be sold as a ‘critical’ need for future computing, and abused as a fearmongering tactic to financially rape the ignorant PC buyer who doesn’t know any better.

        Sounds about right.

    • You will have to assume that the context of AI run instructions will be completely different from a normal user's context. And that telemetry will be on another level by many factors.
    • AI PC about running models, not creating them.

      Intel is basically claiming they have / are coming out with chips that are in the same league as Apple Silicon with respect to AI. Basically their CPUs will also contain neural engines and other hardware to assist in the running of ML models. This is something real. For example, the relatively modest neural engine in an Apple Watch allows it to run small models doing various kinds of speech analysis.

      Neural Engines in CPUs are a thing. They will become stan
    • by gweihir ( 88907 )

      Indeed. Essentially Intel is trying to sell people expensive but useless hardware to make a big pile of cash. I expect too many will fall for it.

      The whole idea is bogus: People will want to use the Artificial Idiots on their phones and that means online services.

  • by Kelxin ( 3417093 ) on Tuesday March 12, 2024 @04:54PM (#64310501)
    Quote: "according to Robert Hallock, Intel's senior director of technical marketing." A marketing person will literally say anything to sell their products.
  • by joh ( 27088 )

    Really. When did you hear the last time of anything future-oriented from Intel? It's quite a bit strange actually. Intel is a bit like Boeing...

  • O snap! So this AI stuff is just software? Quick! Sell off NVDA

  • They haven’t figured out how to market yet.

    So just buy intel crap until they do.

  • Not the first thing that popped into my mind!
  • by jythie ( 914043 )
    Really, anything with MMX instructions is an 'AI PC'
  • You may remember when speakers were advertised as "digital ready"......

  • 32 GB should be STANDARD. For an LLM? No for browsing and email. Let's be real.
    • Mutt users point and laugh at you.

      For work, I had to run Slack though, and it OOMed on 128GB. I've then upgraded to 256GB memory and that seems enough -- at least until the Chrome engine underneath gets upgraded. Which suggests there might be something mildly wrong going on with modern browsers.

      • by RedK ( 112790 )

        Or something wrong with your Slack.

        That is absolutely not normal behavior at all.

        • "My" Slack is exactly the binary provided by Salesforce, that I can not (easily) modify or debug. I probably shouldn't have ran it as a non-sandboxed-away process, as it's untrusted code provided by a potentially malicious actor (at the very least, most proprietary software includes adware and spyware -- a practice that has been somehow normalized these days).

          But in any case, I have no means to debug what is wrong with Slack or why it OOMed -- any complaints should be addressed exclusively to Salesforce di

  • (Customer) ”Hey, I was wondering which model is your cool new AI PC.”

    (Intel) ”Yes.”

    (Customer) ”Oh, cool. For a minute there I thought I was going to have to pay some premium to get a special one. Now that I know ALL of your hardware has it, well that pretty much confirms you’re ALL in on the marketing hype and bullshit. Nothing special to see here. At all.”

  • Intel can say whatever they want, but, will Microsoft require VNNI Support? What about DP4a? Or perhaps they just want a machinme that can run DirectML and OpenCL?

    Whatever they say is needed for a PC to run Win12 is what we will come to call AI PC...

    And the decisions MS makes regarding Windows HW needs tend to influence Linux (in particular, the desktop side) in the long term, for better or worse.

  • There are plenty of use cases for AI that will work a lot better if the primary GPU is still available for its original purpose. It was a bit like pushing the physics engine out to the GPU. Great, but now what's going to smash out frames?

    Games will be the obvious reason you'd want your GPU unencumbered, and probably the primary "killer app", though surely not the only one.

  • Does "AI PC" stand for "Advanced Intel" PC? Can it run an Cortana-like LLM?

    "Don't send that John .. you'll get yourself fired!"

"It's the best thing since professional golfers on 'ludes." -- Rick Obidiah

Working...