Final month, we wrote about new details about the follow-up to Microsoft’s HoloLens, and the way the brand new machine would combine an AI co-processor. On the time, we speculated that Intel’s Atom SoCs won’t be lengthy for this world, although we didn’t count on the hammer to drop fairly as rapidly because it has. Intel is killing its Atom X5-Z8100P processor, the chip that powers the Microsoft HoloLens.
This chip isn’t publicly listed on Intel’s web site (Ark.Intel.com; helpful for all CPU identification duties), however has been recognized (PDF, however with safety warning) as a Cherry Path-based half clocked at 1.04GHz. Intel’s resolution to cancel its Atom-based SoCs because it exited the smartphone and pill enterprise has left Microsoft with out a follow-up product, although the truth that Intel is exiting the marketplace for a chip that will solely have had one buyer implies Microsoft and Intel have labored this situation out amongst themselves.
Another excuse to view the scenario as a mutual settlement is Intel hasn’t killed its different Cherry Path Atom SoCs in the identical household:
This raises the query, nevertheless: If Intel isn’t offering a CPU resolution for HoloLens 2 (our title), who’s? We could not know which firm, particularly, is constructing the answer, however there are solely three choices, and solely one among them is smart:
1). Microsoft has paid Intel to construct it a semicustom Atom SoC based mostly on Goldmont.
2). Microsoft paid AMD to construct it a semicustom SoC based mostly on a 14nm Jaguar core.
three). Microsoft contracted with an ARM vendor.
#three is the one practical choice right here. Microsoft could have paid Intel for a customized Atom SoC earlier than, however that was again when Intel was nonetheless dedicated to promoting Atom chips throughout the pill and smartphone markets, together with in Microsoft’s personal Floor three. It’s one factor to be a semicustom buyer for the mildly customized half that’s being bought elsewhere, and one thing else solely to be the solely buyer the corporate expects to have for a complete line. MS can scarcely hope to make its funding again in quantity; there’s no indication that the corporate is planning a broad client rollout for HoloLens 2.
The second choice doesn’t make sense both, for a similar purpose. We all know AMD has ported Jaguar to 14nm, however the SoC designs for the PlayStation Professional and Xbox One X don’t have a mobile-compatible GPU bolted on. And AMD has made solely the faintest of overtures in direction of the pill market because it bought the GPU IP that grew to become Qualcomm’s Adreno practically ten years in the past. Once more, Microsoft isn’t going to wish to foot the complete invoice for a low-cost x86 SoC and related graphics , particularly when AMD, like Intel, has no plans to compete within the pill or smartphone area.
That leaves choice #three. Right here, there are a plethora of choices and loads of corporations who would love a high-profile win in a tool like HoloLens. Each Qualcomm and Creativeness Applied sciences have GPUs that might deal with the processing, particularly since Microsoft is beefing up the position and functionality of its synthetic intelligence processing.
This final level is presumably why the agency wouldn’t flip to an organization like Nvidia. Whereas Nvidia might need some silicon in its arsenal that might be repurposed for this position, strapping Microsoft’s customized AI silicon into the identical that options Nvidia’s AI silicon (by way of the GPU) can be redundant. Apart from, Nvidia has gotten out of tablets and smartphones as properly, and is concentrated on autos, which have rather more strong cooling choices and room for bigger boards than your common headset.
Barring a Microsoft announcement that it’s entering into the CPU enterprise, ARM is the one doubtless match right here.