Microsoft was rumored to have upset some of its silicone partners during its Qualcomm heavy presentation of Copilot + PCs a few weeks ago, but it may be those same jilted collaborators that help stress test the company’s AI visioned future.
While Microsoft publicly wraps its arms around Qualcomm’s Snapdragon X Elite processors for its initial Copilot + PC initiative, AMD and NVIDIA recently unveiled their own collaboration that should corner the power user ang gaming sectors of Windows customer base.
During the opening day of Computex, NVIDIA teased its RTX AI PC line of laptops that will come from partners Asus and MSI soon.
The RTX AI PCs are being marketed as the gaming wing of Microsoft’s new AI-laden Windows 11 push, with the new PCs being outfitted with GeForce RTX 4070 GPUs and AMD’s newest Strix Points CPUs (Ryzen AI 300 Series) that boast 50 trillion of Operations (TOPS).
Despite being loaded up with CPUs that are more than capable of meeting Microsoft’s new NPU support of 40 TOPS minimum requirements to run its suite of AI features for its upcoming Windows 11 update, these new AMD and NVIDIA powered PCs will not support the Copilot + feautres out of the box.
Instead, NVIDIA followed up their tease with a blog post clarifying that a supported update will upgrade its new lineup with the Windows AI feature set.
“These Windows 11 AI PCs will receive a free update to Copilot+ PC experiences when available,”
Jordan Dodge, SHIELD, GeForce NOW NVIDIA Corp.
It’s unclear why Microsoft and its other silicone partners appear not to be on the same rollout schedule for brand and marketing support of Windows 11’s new AI push, but it has said chip makers that aren’t using ARM architectures will still be able to support all its Copilot + platform features in the future.
Applying Occam’s Razor theory, it may just be easier and faster to apply the Windows Copilot Runtime to a SoC that has an NPU onboard than to reroute processes through CPUs, GPUs and NPUs.
Another theory is that these more powerful PCs will be used to run larger language models (LLMs) where Microsoft may have wanted to display its small language models (SLMs) operations to highlight portability and battery life. GPUs are better suited at running LLMs when connected to a power source while Microsoft’s SMLs are designed to keep AI functions on-device while also taking battery consumption into consideration.
While NVIDIA’s GPUs have more than enough to power TOPS much higher than the minimum requirements set by Microsoft, releasing Copilot + PCs this summer may be a situation of getting something out of the door that works consistently and measured more than offering a wider range of performance comparisons.
Microsoft and NVIDIA are already working on getting the Windows Copilot Runtime out to developers to help them tap into NVIDIA GPUs which can offer up to 1000 TOPS, for the power they’ll need to run the Copilot + platform on supported devices.
For its part, AMD announced its own Ryzen AI 300 series chips with a new Zen 5 CPU microarchitecture that runs using the two cores of RDNA 3.5 and XDNA 2. AMD’s RDNA 3.5 will handle graphics portions of CPU processes while the XDNA 2 core will be dedicated to handling all of the AI workloads with support of 50 TOPS.
AMD’s XDNA 2 core marks the company’s ability to jump almost five times the performance it offered with its third generation solutions and, if proven, outpaces Microsoft flagship partnership with Qualcomm and its X Elite chips, handily.
Part of the performance jump can be credited to the increase in cores, moving up from 8 to 12 this time around as well as 12 CUs to 18.
AMD will offer two versions of its AI 300 Series chips that includes the Ryzen AI 9 HX 370, with twelve cores and twenty-four thread support, and the Ryzen AI9 365 that is just step down with ten cores and twenty thread support. Both AI 300 Series chips will come with those coveted 50 TOPS NPUs and 800M Radeon graphics.
Similar to the way ultrabooks and gaming laptops were defiined by their overall performane, I think this new wave of Copiliot + PCs will shake out in very likely order, with ARM NPUs being reservered for the average battery conscious Windows users while researchers, scientists, traders, gamers, artists, and other power users move toward the beefier CPU+GPU combos.
