Snapdragon Summit 2024: Oryon (and AI) Everywhere

At its annual Snapdragon media event in Maui, Qualcomm took the Oryon CPU core architecture it applied to PCs and is bringing it to smartphones and automotive chipsets. The main keynote session bounced back and forth between PCs, smartphones, and partner announcements, so it is worth summarizing the announcements and putting them in context.

Snapdragon X PC Platform – Not the Focus, but not not the focus

At last year’s Snapdragon Summit, Qualcomm announced the Snapdragon X line of Windows PC processors with Oryon CPU cores derived from its Nuvia acquisition and a large, dedicated NPU for on-device AI workloads. Snapdragon X Elite chips started arriving in laptops this summer, and a lower-priced Snapdragon X Plus version was announced at IFA this year that maintains the NPU’s 45 TOPS while cutting back on CPU cores.

As the first potentially disruptive silicon for PCs since Apple moved away from Intel, Techsponential has been extensively covering Snapdragon X announcements and laptops*. In Maui, Qualcomm presented graphs claiming that Snapdragon X Elite still beats Intel’s latest Lunar Lake chips and AMD’s Strix Point on wired performance, with even stronger comparisons when running on battery. While Qualcomm made no further PC announcements at Snapdragon Summit, it did promise major silicaon improvements for next year’s Snapdragon Summit. By previewing its road map for later 2025, Qualcomm avoids Osborning sales for this holiday season, while reassuring software developers and hardware partners that Snapdragon X isn’t a one-off. Those developers have been busy. Qualcomm claims that 90% of users’ time will be spent on native Windows-on-Arm apps, but it also noted that Blender is now native for Snapdragon X with text-to-image NPU rendering. Affinity Photo 2 was already native; it now gains new NPU features as well. Another premier photo processing app, Capture One, is now native for Snapdragon X as well. Finally, Microsoft is adding MIDI 2.0 support soon, and the platform will become friendlier to audio input devices with low latency ASIO drivers coming for generic and 3rd party inputs.

CEO Zoom Roulette

Qualcomm had a parade of tech company CEO’s calling in with prerecorded messages. Rather than play each video in contextually-related sections of the keynote, Qualcomm grouped them together, making it appear less a part of Qualcomm’s narrative and more like relatives who couldn’t make it to the destination wedding. where the presentation where Microsoft CEO Satya Nadella talked about its partnership with Qualcomm for Windows on Arm. OpenAI’s Sam Altman had the feel-good message that, "cloud and on-device AI complement each other." Meta’s Mark Zuckerberg talked about its Llama AI and noted that Qualcomm powers Meta’s XR initiatives (the smart glasses collaboration with Ray-ban, all of Meta’s Quest VR headsets, and the compute puck in the Orion smart glasses demo project). Epic’s Tim Sweeney’s video wasn’t about gaming on Snapdragon, but how the auto industry is using Unreal Engine on Qualcomm’s platform. GM’s Mary Barra talked about using that silicon, while BMW’s Frank Weber continued the theme with ADAS. Then a Best Buy executive talked about marketing Snapdragon X PCs at retail, “very human AI,” and something about believing in yourself.

Not everyone was on video, but here, too, the introductions often derailed Qualcomm’s narrative rather than reinforced it. Samsung’s TM Roh appeared live at the event to talk about AI and Samsung’s Galaxy ecosystem. Xiaomi’s Adam Zeng briefly talked about Xiaomi’s cars, connected devices, and phones – including a teaser of the Xiaomi 15, the first phone with Qualcomm’s new silicon. Microsoft’s Pavan Davuluri talked up Copilot+ PCs on Snapdragon X. Alex Himel from Meta talked about using AI on Ray-Ban Meta smartglasses to help with reminders. Honor’s Dr. Ray Guo discussed co-developing an AI ecosystem with Qualcomm: the HONOR AI Agent can cancel subscriptions for you and learns your habits for reordering coffee (this seems to be China-only, but this is a great use case). He also teased the Honor Magic 8, another early Snapdragon 8 Elite smartphone.

Snapdragon 8 Elite Mobile Platform

The big news of the Summit was the launch of the second-generation Oryon CPU core, which takes the original architecture intended for datacenters that was modified for PCs and optimizes it further for mobile. Qualcomm’s current flagship mobile SoC is the Snapdragon 8 Gen 3, so the natural follow-up should have been Snapdragon 8 Gen 4. Instead, Qualcomm is rebranding to Snapdragon 8 Elite, reflecting the significant boost in performance that the new Oryon cores and other changes bring. The changes really are significant, so the name change is smart.

Gerard Williams, Qualcomm’s SVP Engineering introduced the second gen Oryon CPU as having faster clock speeds, lower power, and the "most area-efficient [i.e., "small"] performance core in the industry." Snapdragon 8 Elite brings 44 – 45% improvements in power efficiency, single-threaded CPU tests, and multi-threaded CPU tests. Qualcomm even claimed mobile performance leadership over Apple, and says that its 2nd gen Oryon CPU mobile performance is faster than many PCs (huge if true). Xiaomi noted that the Snapdragon 8 Elite -based Xiaomi 15 uses 30% less power and runs cooler than the Xiaomi 14.

While the new CPU cores take the spotlight, the Hexagon NPU in the Snapdragon 8 Elite features more cores across scalar (8-core) and vector (6-core) blocks. There is longer context support and more input tokens for complex reasoning. Performance and efficiency are both up by a massive 45%. It’s not all hardware; Qualcomm’s AI Hub allows developers to optimize their models for the Snapdragon 8 Elite, and Qualcomm is partnering with everyone.

As you would expect from a Qualcomm mobile SoC, the integrated connectivity is core to the value proposition. The Snapdragon 8 Elite has 6 carrier aggregation 5G, NB-NTN, and highly integrated Wi-Fi 7, Bluetooth, and UWB.

It's About AI. And the Camera. But Mostly AI

After talking up AI use cases generally, Qualcomm CEO Cristiano Amon tried to make it less abstract. He suggested that generative AI integrated into smartphones means that you'll be able to talk to your banking app on your phone instead of navigating the interface, and then show the app your bill (via the phone’s camera) and ask for the bank to pay it. Other examples included an on-device AI assistant that can help you split a restaurant bill with a voice prompt, translate the menu, convert currency, and describe what you see with connected smart glasses or the phone’s camera. (It’s worth noting that some of these use cases have already been implemented on older and rival chipsets.) With software from ArcSoft, Snapdragon 8 Elite phones gain pet photography features using the NPU for tracking movement and enhancing fur so that animal family members are captured and get their closeup.

Qualcomm has proven that a Snapdragon with Oryon and a large NPU can create competitive Windows laptops. While we will need to test the Snapdragon 8 Elite phones to see how well Qualcomm’s performance and battery life claims hold up, the gains appear to be huge. The additional headroom and efficiency are welcome for today’s phone use cases, and as Google, phone OEMs, and developers start incorporating more on-device and hybrid AI models into the OS and apps, it should allow new AI-driven apps and use cases to emerge. Google, OpenAI, and Meta AI models are technically ahead of Apple Intelligence capabilities today; whether they are able to match Apple’s eventual cross-app and privacy-protecting implementation remains an open question. But at least they won’t be constrained by the silicon.

Automotive

Aside from the brief Meta cameos, Qualcomm’s major investment – and silicon lead – in XR was not a big part of Snapdragon Summit, and the company’s growing IoT business was basically ignored. Automotive, however, was the sole focus of the second day keynote. In the space of a few years, Qualcomm has deeply embedded itself into the automotive supply chain – its partner slide is incredibly impressive – and so packed with logos that it could be an eye chart. Qualcomm is taking the Oryon CPU architecture it first employed in Snapdragon X PCs and is now for smartphones and is now applying it to cars with the Snapdragon Cockpit Elite and Snapdragon Ride Elite for both the consumer-facing aspects and ADAS. The SoC powering these platforms are designed to provide significantly more power than other automotive silicon, including 3x the CPU performance and support for 40 multimodal sensors. It also offers a 3x performance increase in GPU to drive in-vehicle visual experiences on up to 16 high resolution displays including camera feeds, real-time ray tracing, gaming, advanced user interfaces, 3D maps, and a functional safety visual digital twin. Is there AI? Of course there is. The NPU has a 12x performance increase, which lays the groundwork for both digital assistants and AI-powered ADAS and self-driving capabilities.

This obviously won’t come cheap, so it is no surprise that Qualcomm’s first design win is with Mercedes Benz. Mercedes’ Nakul Duggal was on hand to talk about how, "cars will be increasingly defined by software." Mercedes is designing its own OS for its cars, and Qualcomm is providing the silicon it will run on.