The next Snapdragon flagship chipset could enable AI agents on your phone. Qualcomm is gearing up to launch its new flagship chips on September 22 during the annual Snapdragon Summit. The company has been teasing the Snapdragon Elite Gen 6 chips for quite some time now, and the latest piece of info details the new Hexagon NPU.
New Hexagon NPU built to enable on-device AI
The new NPU architecture and the Element Accelerator module. | Image by Qualcomm
Today, Qualcomm teased the new Hexagon NPU inside the upcoming Snapdragon 8 Gen 6 chipsets. The silicon is built to enable on-device AI—and especially AI agents—and uses two key ingredients in order to achieve that.
The first one is something called the "Element Accelerator," and no, it doesn't smash elementary particles together. It's a specialized unit that works in tandem with the other three blocks inside the NPU—the scalar, vector, and tensor units.
According to Qualcomm, the Element Accelerator is designed to speed up transformer inference, “helping agents respond faster, reason more efficiently, and deliver richer experiences.”
The second part of the equation is the amount of memory inside the NPU. Qualcomm says that the four accelerator units—tensor, vector, scalar, and the new Element—now have more shared memory, which will reduce dependence on the phone’s RAM.
Recommended For You
The first 5 GHz smartphone chip
Qualcomm is breaching the 5GHz limit with the upcoming Snapdragon 8 Elite Gen 6 chips. | Image by Qualcomm
Another big milestone that Qualcomm is openly bragging about is that the new Snapdragon chipset will be the first smartphone processor to breach the 5 GHz mark.
Qualcomm says that this feat has been achieved through a new and fully custom design, rather than by perfecting the manufacturing process.
For reference, the Snapdragon 8 Elite Gen 5 for Galaxy is able to reach 4.47 GHz.
The upgraded GPU is expected to bring PC- and console-level gaming to smartphones with improved frame-upscaling tech, along with much faster parallel processing of complex shaders, high throughput, and an 18 MB dedicated on-chip memory buffer.
Would you prefer your AI agents to run on-device or in the cloud?
4 Votes
The new Mixture-of-Experts model
The new Snapdragon chips are built and optimized for Mixture of Experts machine learning architecture. | Image by NVIDIA
Another big and complicated term has been thrown into the Snapdragon equation. Qualcomm says that these new chips will support AI architecture built and optimized for something called Mixture-of-Experts (MoE) models.
This is a machine learning architecture that divides a model into smaller sub-networks ("experts") and uses a gate to activate only a relevant subset of experts for each input query. In plain English, this will reduce compute and memory demands when you use AI on your phone by using a dedicated part of the AI module depending on the task.
Qualcomm boasts that the architecture will allow models with up to 30B parameters to run locally on your phone.
The era of on-device AI agents is nigh
The era of on-device AI agents is nigh. | Image by PhoneArena
Ever since ChatGPT and all other LLMs started to appear on smartphones, companies have been talking about "on-device AI." Granted, some features were able to run on the device, but a quick disconnect from the internet revealed a gaping hole in those promises.
Of course, you need an internet connection for the AI agent to gather data and stay relevant, but keeping the computing on-device means no sensitive information will be sent to servers and data centers online.
On one hand, it's much more secure, and on the other, AI agents should work much faster, eliminating network lag entirely. The Snapdragon 8 Elite Gen 6 Pro has already made headlines through leaked Geekbench scores, and the chip is shaping up to be a powerhouse. Whether or not it will be able to challenge Apple's new A20 silicon remains to be seen.
Mint Mobile is now allowing you to get whichever plan you like for either three, six, or 12 months for just $15/mo. If you go for the six-month unlimited service, for instance, you'll now have to pay just $90 upfront instead of $210.
Mariyan, a tech enthusiast with a background in Nuclear Physics and Journalism, brings a unique perspective to PhoneArena. His childhood curiosity for gadgets evolved into a professional passion for technology, leading him to the role of Editor-in-Chief at PCWorld Bulgaria before joining PhoneArena. Mariyan's interests range from mainstream Android and iPhone debates to fringe technologies like graphene batteries and nanotechnology. Off-duty, he enjoys playing his electric guitar, practicing Japanese, and revisiting his love for video games and Haruki Murakami's works.
A discussion is a place, where people can voice their opinion, no matter if it
is positive, neutral or negative. However, when posting, one must stay true to the topic, and not just share some
random thoughts, which are not directly related to the matter.
Things that are NOT allowed:
Off-topic talk - you must stick to the subject of discussion
Offensive, hate speech - if you want to say something, say it politely
Spam/Advertisements - these posts are deleted
Multiple accounts - one person can have only one account
Impersonations and offensive nicknames - these accounts get banned
To help keep our community safe and free from spam, we apply temporary limits to newly created accounts:
New accounts created within the last 24 hours may experience restrictions on how frequently they can
post or comment.
These limits are in place as a precaution and will automatically lift.
Moderation is done by humans. We try to be as objective as possible and moderate with zero bias. If you think a
post should be moderated - please, report it.
Have a question about the rules or why you have been moderated/limited/banned? Please,
contact us.
Things that are NOT allowed:
To help keep our community safe and free from spam, we apply temporary limits to newly created accounts: