News Dell hints at Nvidia-made chips for Windows AI PCs as soon as next year

[Admin]

Dell and Nvidia chief executives imply that Nvidia's processors for AI PCs will be ready next year.

Dell hints at Nvidia-made chips for Windows AI PCs as soon as next year : Read more

 

[vanadiel007]

"All of the nonplayer character will be chatbots. Creating world will be easier. Instead of instruction driven computing, it [is going to be] intention driven computing, so it will be easier to write programs."

Ugh, they can't even write error free code today, with games being full of bugs and unfinished content at release.
And here's this guy claiming it will be easier...

 

[ThomasKinsley]

MediaTek and Nvidia have a partnership where MediaTek supplies the SoC and Nvidia supplies the graphics and AI. MediaTek has announced that they intend to get into the Windows AI PC market, so this could be it.

 

[Notton]

This is old news.
reuters, of all places, had this story 7 months ago.

https://www.reuters.com/technology/nvidia-make-arm-based-pc-chips-major-new-challenge-intel-2023-10-23/

 

[Metal Messiah.]

MediaTek and Nvidia have a partnership where MediaTek supplies the SoC and Nvidia supplies the graphics and AI. MediaTek has announced that they intend to get into the Windows AI PC market, so this could be it.

I suppose that was only about the MediaTek Dimensity Auto system. A different take on this, since the Mediatek partnership only incorporated NVIDIA's GPU IP, RTX graphcis in the Dimensity SOCs which are responsible for the Auto Cockpit.

https://www.mediatek.com/products/automotive

 

[kealii123]

"Instead of instruction driven computing, it [is going to be] intention driven computing"

Very skeptical. Algorithmic computing is much more deterministic, while AI generated computing is too open ended. Game developers, like all developers, want a known outcome.

 

[bit_user]

Nvidia has everything it needs to build system-on-chips for PCs, including Arm general-purpose cores, GeForce graphics processors, Tensor cores for AI, and various other hardware.

It already does have its own SoCs! The Nintendo Switch is using just one in a long line of them, and there are many newer ones that have been made, so far.

the company could address this market by licensing its IP to third partners, which would reduce risks and increase profit margins but still ensure that Nvidia gets its piece of the pie.

Mediatek is one such licensee. It'd be interesting (but not unheard of) for Nvidia to compete with one of its partners.

 

[bit_user]

I suppose that was only about the MediaTek Dimensity Auto system. A different take on this, since the Mediatek partnership only incorporated NVIDIA's GPU IP, RTX graphcis in the Dimensity SOCs which are responsible for the Auto Cockpit.

https://www.mediatek.com/products/automotive

My take on that is Mediatek decided to tap the automotive market, while waiting for the Windows-on-ARM exclusivity agreement with Qualcomm to expire. This lets them build a higher-spec SoC than they normally do, and potentially get it in the hands of laptop makers to do some preliminary eval/integration work.

 

[NeoMorpheus]

MediaTek and Nvidia have a partnership where MediaTek supplies the SoC and Nvidia supplies the graphics and AI. MediaTek has announced that they intend to get into the Windows AI PC market, so this could be it.

Someone should teach Mediatek the tale of the scorpion.

And sounds like Dell found another company willing to “donate” to their coffers, since it looks like intel can’t continue bribing them.

 

[rluker5]

This sounds like it could just be an AI coprocessor on the motherboard like the old graphics chips AMD and Nvidia used to make. It would not have to be large to adequately supplement a desktop chip that doesn't have a large enough NPU.

Edit: Doesn't Raptor Lake have support for some mysterious AI M.2 that never materialized? It might have been some premature preparation for AI Windows.

On a related note I'm finding the intrusiveness of Microsoft Copilot far more aggravating than any Windows ads. It pops up far too much and sometimes makes my pc feel like my OS is some free phone app. I would pay a fee to get a "premium" CPU or motherboard that lacks the NPU necessary to fuel these interruptions for services I have no desire to use.

 

[Metal Messiah.]

Well, some details emerged from that interview.

NVIDIA might leverage a 3nm process node, reportedly either Intel or TSMC, and an advanced packaging solution to bring Arm-based Cortex X5 BlackHawk CPU cores & the Blackwell RTX GPU cores together.

It's plausible that NVIDIA would use the Cortex X5 BlackHawk CPU cores since they are reportedly partnering with MediaTek to create these SOCs. NVIDIA SOCs are also stated to feature LPDDR6 memory as an on-package design.

TSMC N3P
LPDDR6 memory
Cortex X5 BlackHawk CPU
Blackwell RTX GPU cores

https://twitter.com/x/status/1793469879334687122

View: https://x.com/XpeaGPU/status/1793469879334687122

https://twitter.com/x/status/1793523940968820899

View: https://x.com/Kepler_L2/status/1793523940968820899



Previous Tweets:

https://twitter.com/x/status/1790588407942684944

View: https://x.com/XpeaGPU/status/1790588407942684944

https://twitter.com/x/status/1789825653501833463

View: https://x.com/dnystedt/status/1789825653501833463

 

[abufrejoval]

I wouldn't dare to predict this one.

Sure, they have all the IP assets to build everything from smaller Jetson Nanos to robotic monsters.

And some diversification may be a good idea, since the betting everyhing on monster GPUs might not last another 10 years.

But will they actually have the various control boards let them allocate the resources and will they find the required human talent to build and support these vast potential product ranges with plenty of competitors staking claims?

I see this more in corporate growth vs management culture than an engineering issue and then you need really huge loyal long term volume markets in an area that is becoming very, very crowded. Sury, they might be able to afford outspending the competition for a while like Intel ...tried.

The xPU co-design ability might give them the assets to scale IP blocks into a lot of distinct environments and workloads, but so far Nvidia hasn't exactly shown combinatorial flexibility, just some fractional scale capabilities.

This could be more about trying to keep investors from flocking elsewhere or testing the waters than a long term strategy finally bearing fruit.

Good thing I don't have a penny to invest so I can just lean back and enjoy the show!