Agentic AI PC's from NVIDIA and Microsoft - NVIDIA RTX Spark Killing OpenAI

But when you fuse a gpu+cpu .. total different architecture + motherboard layout , heat dissipation, memory requirement, the notebook layout or PC need to be redesign , + a string of software testing with microsoft and provider.. when it is going to be released? this is easily a 5 year project before stable, and they manage to hid it this long and no beta ? Did they use AI agent to rewrite microsoft kernel d??? Oh dear

Eli, thanks for these videos man, really love them. Thanks also for keeping them clean, I can finally watch you with my kids around.

Microsoft abandoning x86 because they want to suck on the teet of Nvidia would be absolutely hilarious.

Oh good, trying to run the clock back to terminal computing. It can only go well. Also, the only rational response to this Dumpster Fire is to let it burn. As for Wintel, some of us called it the WIMPtel stack during the days of that duopoly.

Well that is going to kill itself.

I think MS and NVIDIA partnership is a sign they are shit scared of Chinese chips and Chinese software and trying to not get too far behind. I think they're toast

The idea of a dedicated little PC that does AI stuff is nice. What would be even nicer would be to be able afford a good old normal PC. Also I have zero faith that that little AI PC won't be riddled with spyware, telemetry and some more or less locked down software eco-system.

Saw this news earlier today they're literally just making a really expensive Alexa's , I can't see this product being worth the price

Its probably in nvidias best interest to make local ai accessible to the general user, because nvidias competitors are getting better and providing cheaper alternatives in the datacenter space. One day they will have to sell more consumer level ai hardware to make up for the lost revenue.

I think there needs to be something similar to LPCAMM2 or such that aside PCI but enable plugging in 8GB, 16GB and even other ranges of graphics so that the graphics memory is closer to CPU as it is to memory. This hopefully has a similarly Unified or Unified Access Graphics Memory. I have also experimented with eGPU and those are also very useful approach so we could look at speeding that as well so in future we could have computers that send work to some units for work to be done (which is a lot of what we do with training, inference and ML).