- US - English
- China - 简体中文
- India - English
- Japan - 日本語
- Malaysia - English
- Singapore - English
- Taiwan – 繁體中文
When it comes to enabling generative artificial intelligence (AI) experiences at the edge, the high-performance and low-power mobile memory inside of your device is critical. At Micron, we have a track record for delivering the world’s highest bandwidth and lowest power memory to global smartphone manufacturers and chipset vendors to design into the world’s leading smartphones.
We have again pushed the boundaries further with an improved 9.6 Gbps LPDDR5X memory solution, which we are currently sampling to the mobile ecosystem. Benefits include sustained high-bandwidth capabilities for AI-intensive applications and further improved power efficiency. When it comes to battery life, every minute that the end user gains is a win. For our latest 9.6 Gbps memory solution, we deliver an additional 4% power savings compared to our previous generation, which was already delivering best-in-class power.1 To put this power savings into perspective, a leading flagship smartphone vendor claims to deliver 23 hours of video playback. If every component in a handset reduced power by 4%, end users would gain an additional hour of video playback, bringing the total playback potential to 24 hours.2 That would be the equivalent of enjoying two more episodes of your favorite streaming show before needing to connect your device to power.
End users demand power-efficient smartphones
Based on consumer research, smartphone battery life is the top pain point for end users and is a critical area for improvement among smartphone manufacturers. In fact, 71% of smartphone users stated that battery life is the most important feature they consider when purchasing a new phone. Battery life easily outpaces other key features such as durability of the device (61%), camera quality (48%) and 5G connection (24%).
Micron leads in LPDDR5X innovation
Our first-to-market 1ß (1-beta) process node has advanced power-saving capabilities that utilize enhanced dynamic voltage and frequency scaling core (eDVFSC) techniques and second-generation high-k metal gate (HKMG) technology to unlock unprecedented power savings. This advanced technology delivers significant power improvements and the flexibility to deliver workload-customized power and performance. In addition to power savings, the 1β process node provides over 12% higher peak bandwidth.3 These power and bandwidth improvements accelerate efficient and effective AI experiences at the edge.
AI use cases at the edge are accelerated with high-performance, power-efficient memory
As AI use cases and applications permeate flagship smartphones, these advancements are projected to trickle down to high-end and mid-tier devices in the future. Counterpoint Research stated earlier this year that over 1 billion AI smartphones are projected to ship globally by 2027, a true testament to the massive demand and quick adoption of AI for the mobile ecosystem and end users. With the proliferation of new AI use cases and the advancements of foundational models delivering more sophisticated and precise feedback, the importance of high-performance and power-efficient memory becomes even more critical to fuel the AI engines that ensure exceptional end user experiences.
AI at the edge is currently in its infancy, and we see many opportunities and challenges that the mobile ecosystem will need to overcome for us to reach full potential and mass adoption. The benefits are clear: security and privacy of your data, instantaneous responses without relying on the cloud, and lower infrastructure costs. The challenges for hardware, software and architecture are still being explored, and standardizations for foundational models are still being solidified. At Micron, we continue to be at the forefront, delivering solutions that are geared to accelerate AI innovation for today and for the future.
1 Compared to the previously launched 9.6 Gbps LPDDR5X
2 Based off iPhone 15 Pro claim of 23 hours video playback
3 Measured against 1-alpha-based LPDDR5X