All the major computing-related companies—Microsoft, Intel, AMD, Apple, and Qualcomm—are talking a lot about artificial intelligence these days, but Nvidia is already having a laugh here. The keynote was mostly about the company’s large-scale hardware for cloud-based AI, as well as details on Nvidia’s (admittedly impressive) progress in making AI training feasible and scalable – currently the Blackwell platform , Blackwell Ultra next year, Rubin to follow – and how costs are coming down as a result.
But even if you forget its data center experience, the company first launched a consumer GPU with AI-accelerated Tensor cores about six years ago. Now it joins the ranks of the “RTX AI PC” brand, but at the forefront: any system (desktop or laptop) with an RTX 2000 series and workstation equivalent or higher GPU is a member (they have Tensor processor). Windows Copilot Plus computers Brand Promotion? Yes, they can have that too – as long as there’s a shiny new CPU and a competent NPU inside.
Nvidia also announced a partnership with Microsoft to provide developers with tools later this year to use RTX GPUs to accelerate their applications through the Copilot Plus programming interface.
This is key, because image and video-generating AI tasks require more power to run on a device than they do on the integrated GPU of the latest CPU chips (and forget about the NPU, which is designed more to take up as little battery as possible, does not process anything except text output generating artificial intelligence). Even standalone GPUs as low as RTX 4050 Mobile can perform up to 194 TOPS (tera operations per second) compared to RTX 4050 Mobile. 45 TOPS Qualcomm hypes up.
I’m here to play games
Nvidia has been talking about its ACE suite of “digital human” technology, and in this case, in order to inject better interactive non-player (aka non-player) characters into games, it showed off its latest at GDC in early April. Version. Now, it’s announced the first game to run ACE NPC on the device, though it’s still a tech demo from Perfect World Games called Legends. (Haters, start your engines.)
If you were paying attention during the Copilot Plus PC launch, you may have heard (or seen) Microsoft show off artificial intelligence “coaches” to help players in Minecraft. Nvidia is launching its own G-Assist project, a framework for developers and the gaming community to create AI gaming coaches (it showed off a tech demo using Ark: Survival Boost) and help automatically optimize performance. In theory, G-Assist can access all statistics in the game, potentially making recommendations specific enough to be useful. As Nvidia pointed out (unintentionally, I think), there are tons of wikis, walkthroughs, and other content available to train them. Get your can of worms opener ready.
The company didn’t announce any consumer hardware, but it did reveal SFF-Ready Guidelines for enthusiast small-form-factor systems: a set of card and case sizes that ensure RTX 4070-class cards or better will fit within them. Card dimensions must be 2.5 slots (50mm) x 304mm x 151mm or less, with an allocated space of 50mm x 312mm x 155mm or less within the case. There are already tons of cards and boxes available; this is just one way of formalizing it. However, I think it needs to have a motherboard guide as well.
Smaller announcements include the news that a second beta version of the Nvidia app (which is more like AMD Adrenalin, adding new features like performance statistics and one-click GPU tuning) will be available for download on June 4, the VLC video player, and DaVinci Resolve now supports RTX video (super-resolution upscaling), and it will open source more RTX Remix tools for modding and content creation applications.