Qualcomm has just taken a significant step towards cornering edge AI use cases, especially when it comes to smart watches and other wearable devices like AI pins and AI pendants, by unveiling the all-new Snapdragon Elite Wear chip that offers "best-in-class AI performance and power efficiency."
Qualcomm's Snapdragon Elite Wear chip is capable of running an on-device model with up to 2 billion parameters, with Time To First Token (TTFT) of just 200ms, and the ability to process up to 10 tokens per second
Qualcomm's Snapdragon 8 Elite Wear leverages TSMC's 3nm chip fabrication process and sports the following CPU architecture:
- 1x big core clocked at 2.1GHz
- 4x little/efficiency cores clocked at 1.9GHz
Qualcomm did clarify that these are not its custom Oryon cores, leaving the door open for generic ARM CPU cores. The chip also integrates Qualcomm's Hexagon NPU to "support up to billion‑parameter models at the edge."
Elsewhere, Qualcomm claims that the chip delivers up to 5x improvement in single-core performance (relative to the Snapdragon W5+ Gen 2 Wearable Platform), and up to 7x improvement in the GPU's FPS-related performance.
The Snapdragon Elite Wear also sports six different connectivity-related technologies:
- 5G RedCap (reduced capability)
- Micro-power Wi-Fi
- NB-NTN for satellite connectivity
- Bluetooth 6.0
- GNSS
- UWB
Finally, the new chipset features a silicon-carbon battery that delivers up to 30 percent longer battery life compared to the previous-gen silicon, as well as a 50 percent charge in just 10 minutes.
Follow Wccftech on Google to get more of our news coverage in your feeds.
