Arm's New Lumex Platform Drives Double-Digit Performance Gains
a day ago
- #AI
- #Mobile
- #Arm
- Arm introduces Lumex CSS platform for smarter, faster, and more personal AI on consumer devices.
- Lumex features SME2-enabled CPUs delivering up to 5x faster AI performance and significant IPC gains.
- New Mali G1-Ultra GPU offers 2x ray tracing uplift and 20% faster AI inference.
- Developers can leverage KleidiAI integration with major frameworks like PyTorch ExecuTorch and Google LiteRT.
- Lumex supports real-time AI use cases like assistants, translation, and personalization on-device.
- Partners can customize Lumex for various product tiers, balancing performance and efficiency.
- SME2 enables low-latency AI tasks, such as 40% faster LLM responses and neural camera denoising at 120fps.
- Leading companies like Alibaba, Samsung, and Tencent are adopting SME2 for on-device AI.
- Lumex offers architectural flexibility with CPUs like C1-Ultra, C1-Premium, C1-Pro, and C1-Nano for different use cases.
- The platform is optimized for 3nm nodes and integrates deeply with the software stack for seamless AI acceleration.