Forgot your password?
Close
typodupeerror
Hardware

NVIDIA Unveils New ARM-Based AI/Graphics Superchip Coming to Windows PCs and Laptops (axios.com) 83

Posted by EditorDavid from the in-the-chips dept.
"The company best known for powering the AI boom is coming for the PC," reports Axios.

Nvidia's CEO unveiled a new ARM-based "N1X processor made alongside Microsoft," reports CNBC, that "will be incorporated into a new RTX Spark superchip, debuting in the fall on a fresh line of Windows PCs from Microsoft, Dell, HP, ASUS, Lenovo and MSI."

More details from Engadget: It was only a matter of time before NVIDIA released a powerful system-on-a-chip (SOC) to take on AMD's Ryzen AI Max and Qualcomm's latest Snapdragon X2 chips. At Computex today, NVIDIA unveiled the RTX Spark, a "superchip" meant to give both laptops and small desktops fast AI and graphics performance...

The company says it offers 1 petaflop of AI computing power, and that it has 6,144 Blackwell RTX cores and 20 Mediatek Arm CPU cores. NVIDIA claims it's similar to the RTX 5070 laptop GPU but with much lower power draw. RTX Spark also has an NPU that's fast enough to be part of Microsoft's Copilot+ initiative, which requires a 40 TOPS NPU, but NVIDIA says it's mainly touting the tensor cores as part of the chip's Blackwell GPU for AI performance. RTX Spark's GPU can directly draw on the chip's large pool of unified memory, which can span from 16GB to 128GB, and the chip itself can use anywhere from single-digit wattage up to 80W...

NVIDIA CEO Jensen Huang positions RTX Spark as a complete reinvention of the PC, eventually turning them more into devices meant for AI agents than manual human input... NVIDIA has been working together with Microsoft for "several years" while designing the RTX Spark, according to NVIDIA representatives... In a blog post provided to media, Microsoft head of Windows and devices, Pavan Davuluri, noted that the company optimized Windows 11's workload profile scheduling for the RTX Spark. "Whether you're checking your email or running an agent locally to debug code, the Windows scheduler on RTX Spark will ensure you get the best performance and efficiency out of your CPU," he wrote.

NVIDIA Unveils New ARM-Based AI/Graphics Superchip Coming to Windows PCs and Laptops More | Reply

NVIDIA Unveils New ARM-Based AI/Graphics Superchip Coming to Windows PCs and Laptops

Comments Filter:

  • Excellent (Score:3, Funny)

    by TimothyHollins ( 4720957 ) on Monday June 01, 2026 @07:39AM (#66169176)

    Does this mean we finally get a hardware-supported Clippy?

  • Is this whatever they were teasing? (Score:5, Insightful)

    by stealth_finger ( 1809752 ) on Monday June 01, 2026 @07:49AM (#66169180)
    Seems...underwhelming.

    • Re:Is this whatever they were teasing? (Score:5, Interesting)

      by wildstoo ( 835450 ) on Monday June 01, 2026 @08:56AM (#66169238)

      Well nVidia and Microsoft are excited about it which is a good indicator that consumers should not be.

      On the other hand, depending on how the real-world non-AI performance shakes out, it's just another nail in the coffin for Intel.

      RTX Spark's GPU can directly draw on the chip's large pool of unified memory, which can span from 16GB to 128GB

      Wonder how they're gonna get the RAM (presumably LPDDR5) for their laptops made when virtually every fab has pivoted to making HBM for datacenters.

      • Re:Is this whatever they were teasing? (Score:5, Interesting)

        by thegarbz ( 1787294 ) on Monday June 01, 2026 @11:23AM (#66169436)

        By the time this launches into an actual product the rest of the datacentre bullshit will be over too. OpenAI has already cancelled Stargate (which was intending to consume 40% of Sk Hynix's production). Microsoft cancelled project Nova, along with about 2GW (because we measure datacentres in power consumption these days) of projects across the world. As of right now 50% of AI datacentre projects have either been indefinitely delayed or outright cancelled.

        • Fine by me. Where I work we just had to spend a down payment on a house to put 512GB of registered ECC DDR5 and 128TB of SSD into a box for virtualizing federated storage chaos testing for edge-cluster products. The same build 6 months ago would have been half the price, if not less.

          It's good that our customers have very deep pockets, because we've already had to increase our (thankfully unannounced) list price on our new offering by almost 50% due to RAM and storage cost.

          A world where Micron has a $1T ma

    • Re: (Score:2)

      by tchdab1 ( 164848 )

      Now even if we don't directly use AI, our PC will somehow, under the covers, be needing and using those thirsty death-star data centers.

  • Windows? (Score:5, Interesting)

    by ThurstonMoore ( 605470 ) on Monday June 01, 2026 @08:15AM (#66169196)

    If I'm going to go through the trouble of changing CPU architectures then why in the hell would I stay on Windows?

    • Re: (Score:3, Informative)

      by drinkypoo ( 153816 )

      Because Nvidia can't write a Linux driver worth a fuck to save their lives.

      I have two Zen3 PCs, a desktop with Nvidia graphics and a MiniPC which of course has AMD graphics. The MiniPC is flawless. The desktop has issues resuming from suspend — occasionally gives no graphics, switching to a VT sometimes works, sometimes restarting X will graphics working again but sometimes not, suspend with a graphics-heavy application running and come back and it's usually hung and graphics won't work correctly unti

      • Re:Windows? (Score:5, Informative)

        by Bert64 ( 520050 ) <{moc.eeznerif.todhsals} {ta} {treb}> on Monday June 01, 2026 @08:46AM (#66169224) Homepage

        AMD graphics are stable if you use the open source drivers, but if you need their closed drivers for whatever reason (eg opencl) then they are extremely unstable and much worse than nvidia.

        Of course if you're going to arm then apple is currently about the best option.

        • Re: (Score:2, Interesting)

          by drinkypoo ( 153816 )

          Of course if you're going to arm then apple is currently about the best option.

          Not going to ARM is an even better option right now. Let them get that ecosystem sorted out so that every platform doesn't need a special snowflake bootloader first.

          • Re: (Score:2)

            by gweihir ( 88907 )

            Indeed. And especially not one signed by MicroSlop.

          • Re:Windows? (Score:4, Interesting)

            by MachineShedFred ( 621896 ) on Monday June 01, 2026 @12:01PM (#66169480) Journal

            Nvidia has basically already solved that. We're currently creating an ODM design that uses the Jetson Nano as the SoC complex, and we attach a shitload of IO on a carrier board, including a Marvell 10Gbe switch. We are creating our own EFI and loading essentially a slimmed Debian onto it - there's a few modules we need from Nvidia for their hardware drivers / board support and that's it. Granted, we aren't loading any GPU anything because it's meant to be a low-wattage chassis controller that has *some* compute grunt if needed for self-repair and management.

            It's far better than it used to be, and it's not the weird shit that Raspberry Pi and the like use. There's a real EFI interface with TPM so you can target whatever bootloader you like, and sign whatever keys you like for secure boot.

            • Nvidia has basically already solved that.

              For their hardware. When you want to move to someone else's hardware, will it still be solved?

              • That's why we sign long-term supply agreements when we choose a hardware design. We're building something with a 10-year life, so we write contracts that reflect that hrough supply stocking, production run guarantees, or forward-compatible design.

                • That's why we sign long-term supply agreements when we choose a hardware design.

                  That's great for you, but irrelevant to users buying laptops, which is the topic at hand.

                  • How many laptop users frequently change their bootloaders? Is this a thing people do all the time, or just the first time transitioning from Windows to not-just-Windows?

                    Are we looking for problems to solve? If a bootloader can bootstrap your OS kernel, that's all you need, no? So what are we talking about here? If you can throw Grub on there, and Grub gets you a kernel / rootfs, we're good, right?

                    Nvidia already knows how to make a proper EFI setup that doesn't lock anyone out, and has the requisite EFI

        • Apple isn't the best option though.

          Technically the soc is excellent, can't deny that. Trouble is it's hobbled by the operating system. And ports.

      • Re: (Score:2)

        by gweihir ( 88907 )

        The core reason why I have only bought AMD graphics cards for a long time now.

      • Because Nvidia can't write a Linux driver worth a fuck to save their lives.

        I'll grant that the situation for Linux is worse than with Windows. But, they seem to have a LOT of trouble writing Windows drivers that don't suck donkey dicks. They may be good at hardware, but their software sucks!

        • For what it's worth, Nvidia's drivers have always sucked pretty bad, going back to the RIVA TNT2.

          Their hardware has always what has carried them. There is a reason why they do so many driver releases, and the ones released a year after the initial hardware launch usually have much better performance and stability on that hardware because their drivers suck.

          • For what it's worth, Nvidia's drivers have always sucked pretty bad, going back to the RIVA TNT2.

            Compared to AMD's drivers, and ATI's before that, they have always been far and away superior. AFAICT, AMD still can't do drivers, but at least we have the option of FOSS drivers which work on Linux. There are no Nvidia drivers worth a shit on any platform today, except for CUDA.

    • Re: (Score:2)

      by Locutus ( 9039 )
      Nobody does AI on Linux so it makes so much more sense to keep playing three legged racing with Microsoft tied to you. NOT.
      NVidia has always been tied to MSFT and their Linux software has always looked like it was done by a single NVidia employee after hours in the basement office.
      Besides, Windows is sooo 'yesterday' no matter how much Microsoft pays Qualcomm or NVidia to tie themselves up with.

      LoB

    • Re: (Score:3, Informative)

      by gweihir ( 88907 )

      Well, first you have to be stupid. That covers about 80% of the human race in some form. Second, you must be technologically incompetent. And then, you must be in awe of Microsofts dominance, while having no understanding of the abysmal (and getting worse) product quality they deliver. I would say that for the time being, this may be a product they can actually sell.

    • If I'm going to go through the trouble of changing CPU architectures then why in the hell would I stay on Windows?

      For the same reason you're currently on Windows. With other OSes available for your CPU architecture clearly the architecture has nothing to do with your choice.

      • Re: (Score:2)

        by HiThere ( 15173 )

        There are, or at least were, many applications that were useful and on Windows, but not available on Linux or BSD. Switching off of those can be a significant cost. But if you change the underlying system, those probably won't be available anyway.

    • Because some businesses still require software that only works on Windows, and therefore will require some cooperation from the publisher of Windows?

  • These N1x have conspiciously similar specs to the Nvidia GB10 sans networking. Is GB10 selling a lot worse than expected? I understood that Apple's M5 may be better value for money for most use cases - if you don't need high speed networking.
  • Linux?

  • P as in Personal as in Affordable ? (Score:5, Interesting)

    by greytree ( 7124971 ) on Monday June 01, 2026 @08:55AM (#66169232)
    The P in PC means Personal which means affordable for the average man.

    I suspect these devices will not be.

    Luckily, the open models running on actually affordable hardware are good enough for most uses by now and Nvidia's multi-trillion days are over.
    The AI bust can't come soon enough.

    • Re: (Score:1)

      by Sique ( 173459 )

      The P in PC means Personal which means affordable for the average man.

      Not exactly. Personal originally meant "not shared with another person". Originally, it meant a computer only you have access to, only you install and run software, and only you store and retrieve data.

      • Re: (Score:2)

        by Gilmoure ( 18428 )

        They seem to be positioning these as ubiquious computing devices, i.e. the computing toaster Steve Jobs was questing after, starting way back in the late '70s and early '80s.

        These will not be stand alone computing devices but likely tied to a network and corporate control systems.
        That the corporation(s) and/or gov't will be monitoring everything done on the system goes without saying.

        If anything these will probably be the death of personal computing.

        • Re: (Score:2)

          by Gilmoure ( 18428 )

          Forgot to add: you won't buy them but rent them, like a cable box.

          Most non-computing folks will see them as Alexa-Plus systems that they don't have to worry about updates and other stupid computer geek chores.

    • I'm not sure about that. I remember our first desktop back in 2000~ was not cheap by any means... and had Windows ME... *shivers*

    • The P in PC means Personal which means affordable for the average man.

      Nope. When the term was created the Personal Computer was not affordable for the average man. Not for years, not with the first clones, only a bit later than that.

      I suspect these devices will not be.

      I suspect otherwise. The CPUs will be no more AI than Intel's and Apple's latest CPUs. What we are seeing is Windows trying to deliver a non-Intel Windows to the general PC population. Something we have not really seen since WinNT 4.

  • Hopefully this is the wakeup call after it sells like garbage. Nobody can afford a new computer. Few people can handle multiple computers of that wattage in their environment, be it home or work. Nobody wants AI garbage holding onto all their data and spying on them. I can't wait to see their stock price after this fails.

    • Re: (Score:1)

      by gweihir ( 88907 )

      Hmm. Good point. People would be stupid enough to buy this, but they probably cannot afford to. Incidentally, I just got one, but I am in Europe and our economy is not systematically being destroyed.

      • Re: (Score:2)

        by PCM2 ( 4486 )

        I just got one, but I am in Europe and our economy is not systematically being destroyed.

        How did you "just get" something that has yet to be released?

        • How did you "just get" something that has yet to be released?
          You saw the second part of the grand parent's post and you did not think they were lying? What part of " I am in Europe and our economy is not systematically being destroyed." rings true to you? Unless somehow the Russia-Ukraine conflict is over, the neo-caliphatists have returned home, and the greens have ended their suicide pact against industry.

          • Re: (Score:3)

            by gweihir ( 88907 )

            I see you have no reading ability. I was referring to "new computer", not the one form the story, very obviously.

            I guess the rest of your "analysis" is on "utter crap" level as well. And no, the war in Ukraine is not destroying our economy either. For that you have to get a moron in power that imposes massive tariffs and makes everybody wanting to stop to trade with you at all and removes all stability from trade. Oh, and starts a disastrous war with Iran to misdirect away from him having raped underage gir

    • Few people can handle multiple computers of that wattage in their environment

      Did you even read the summary, or know how ARM works?

      This thing can get down to single-digit watts if the performance is not required at that moment. And it says right there in the summary that it will go up to 80W, which is basically where any desktop gaming / performance chip sits (65 - 120W).

      This will consume LESS watts while delivering more performance per watt, just like their Jetson Nano and Jetson AGX platforms do against x86-64 embedded and low-wattage alternatives (i.e. AMD Ryzen Embedded V3C14 /

      • Re: (Score:2)

        by HiThere ( 15173 )
        OTOH, this is an attached processor, so it's in addition to your current setup. Still doesn't sound likely to blow circuit breakers, but possibly equivalent to running two computers.

  • Is this really a top performing game device, or just a thin client to sell cloud SAAS crap? I'm not always attached to the internet and I'm not interested in renting my compute.

    • Re: (Score:2)

      by Junta ( 36770 )

      It's probably going to be pretty credible, though they are thinking more 'openclaw' and less 'gaming'.

      Though I despise everyone playing around with nVidia's "Superchip" terminology.

      • Re: (Score:2)

        by unrtst ( 777550 )

        It's probably going to be pretty credible, though they are thinking more 'openclaw' and less 'gaming'.

        I'm not even much of a gamer, but every time I see or play one these days, LLM's come to mind. In just about any game, NPC's have enough stats and existing dialog to feed to a LLM and have it generate way better dialog and actions. For example, go to a vendor and ask, "what gun suits my current build the best?" It's one of the few areas I look forward to having lots more (on device) AI stuff - gaming machines are already power hungry hot boxes, and any hallucinations or mistakes would only impact a game. Wh

    • Bet ya it'll be a thin client that network boots, and when you're not using it, the AI compute power gets shared with the world without you getting any kickbacks.

  • Over Hyped so far (Score:5, Informative)

    by whitelabrat ( 469237 ) on Monday June 01, 2026 @09:14AM (#66169262)

    I have a pair of Nvidia GB10 and much of Nvidia's claims are overblown. More importantly Nvidia has rather poor support for this chip (SM 121) at the moment, so unless you are highly skilled at running vLLM or whatever, you will probably be disappointed. Give it some time to cook.

    • The slides that are circulating say the quiet part out loud. Itâ(TM)s one workload (timed compilation) and they position their 20 core part as âoebeatingâ a 10 core M5 and coming âoereally close guys!â to a 16 core M5 pro. Whatâ(TM)s missing? Any LLM benchmarks⦠itâ(TM)s a repeat of the DGX which was quickly repositioned as âoemeant for low cost development for the CUDA stackâ after the M4s inference mopped the floor with it. (MLX/metal does have lim

  • Why not risc-v? (Score:3)

    by aglider ( 2435074 ) on Monday June 01, 2026 @09:23AM (#66169286) Homepage

    Nvidia could both contribute to and use risc-v technology!

  • "Personal Computing Devices" (Score:5, Insightful)

    by Gilmoure ( 18428 ) on Monday June 01, 2026 @10:10AM (#66169354) Journal

    They want to replace PCs with PCDs (Personal Computing Devices) that will have to be tied to the net (i.e. rented like a cable box) to work and monitors everything you do on them.

    Oh yeah, they'll let you plot and goon on the boxes, just so they have dirt on you for control down the line.

    I wonder how long it'll be before real computers are restricted to only licensed (gov't / corporate approved) individuals? Maybe we make it to 2040 but pessimist me says they'll try to start controlling PCs before 2035.

  • NVIDIA CEO Jensen Huang positions RTX Spark as a complete reinvention of the PC

    No, it's not a re-invention. It's a nice upgrade, based on specs alone, but will likely be priced at conventional levels. It honestly just looks like you're catching up to Apple Silicon....which is a very good thing...but no, having a decent GPU at a lower price doesn't fit the definition of "re-invention."...nor revolution...nor do I see any way this will change personal computing. Any updates to how we use computers would have happened regardless. Unless they sell these for less than $100 or some insane price drop, no...this is a modest upgrade for Windows users. If these actually ship in fall...well...computing will be EXACTLY THE SAME a year later. No revolution...No "reinvention"...just an actual spec upgrade for a change, which we haven't even been seeing in chips in the last few years, but used to be the norm, just 10 years ago.

  • ..and send it onto their servers.

  • Side benefit: (Score:4, Funny)

    by Locke2005 ( 849178 ) on Monday June 01, 2026 @04:00PM (#66169922)
    The great thing about running an AI engine locally is that it completely replaces your home heating system on those cold winter nights!

  • The definition of hyperbole (Score:3)

    by toxonix ( 1793960 ) on Monday June 01, 2026 @04:36PM (#66169990)

    "Jensen Huang positions RTX Spark as a complete reinvention of the PC"
    He's comparing this to the leap between rotary phones and cell phones. Or is it flip phones and smart phones? I can't tell.
    They look exactly like the MacBook Pro I'm typing on right now.
    They're touting the fact that they had to rewrite the Windows kernel scheduler for the RTX chip.
    Copilot, in all its forms, stinks. It's not the model's fault (use any model really), it's the tool implementation itself.
    This is not the first laptop to include tensor cores, so again it's not a reinvention of the PC.
    IDK maybe it is. Maybe Jensen is not full of hyperbole. Seems unlikely though.

  • Having engineered a hardware shortage and price-gouging the likes of which we've never seen before, now, NOW, Nvidia want to sell us stuff? For the PCs we can no longer source the parts for, or afford to build? Genius. Pure genius. All so we can host our own Hallupedia (https://halupedia.com/) I assume...

Slashdot Top Deals

There is no time like the present for postponing what you ought to be doing.

Close