ARM today unveiled its next generation CPU and GPU designs and is also presenting the Lumax Compute Subsistum (CSS). Earlier, MediaTek and Samsung’s choice will license CPU and GPU designs and combine a chipset together. Lumex CSS is a turny solution.
To be clear, Arm is not going to start selling chips. Instead, it has prepared the production-taire implementation for 3Nm semiconductor nodes of several founders.
In their words of ARM, its silicon and OEM partners will be able to use the implementation as “flexible building blocks, so they can focus on discrimination at the CPU and GPU cluster level.”
| Cpu | Main advantage | Performance and efficiency benefits | Ideal use cases |
|---|---|---|---|
| C1-AULTRA | Major peak performance | +25% single-thread performance Dual-digit IPC receives benefits from year to year |
Big-model estimates, computational photography, content creation, generative AI |
| C1-Premium | C1-AULTRA Performance with more field efficiency | 35% small area compared to C1-AULTRA | Sub-Flagship Mobile Segment, Voice Assistant, Multitasking |
| C1-Pr | Constant efficiency | +16% continuous performance | Video playback, streaming estimates |
| C1-Nano | Extremely powerful | +26% efficiency, using low area | Wearbals, the smallest form factor |
Therefore, there is no fixed design, however, the new parts are highly adaptable. The new C1-DSU enables the design of 1 to 14 CPU core, with three of the four options, C1-ALTRA, C1-Premium, C1-PRO and C1-NANO we discuss in other articles. For GPU, Mali-G1 1 to 24 shades scales.
Lumex css platform
The ARM has a “secret sauce” to make Lumax better than custom chipset designs that have come before. The new system keeps the system-level cash (SLC) for interconnect L1 chipset and has a 71% leakage decrease compared to standard RAM designs, which reduces passive power consumption.
Additionally, the new Memory Management Unit (MMU) L1 enables safe and cost-efficient virtue (usually used to run several OS simultaneously on the same computer).
The ARM for Lumex CSS has power efficient and safe “connective tissue”
The C1 CPU Compute Cluster sixs receive average 30% higher performance in six industry benchmarks. For applications such as gaming and video streaming, this is 15% faster on average. For other workloads, things like video playback, web browsing and social media, this average is 12% more efficient. This is compared to the previous designs of ARM.
Focusing on the top-end hardware, the Arm C1-AULTRA CPU offers a double-digit instructions for improving the Cortex-X 925 per cycle (IPC). Mali-G1 is 20% in Ultra GPU linearization and is twice on raw tracing works compared to the Immortalis-G925.
The newly introduced scalable Matrix Extension 2 (SME 2) is in the heart of ARM’s push for high on-device AI performance-Navy CPU is 5x up rapidly and 3x can be more efficient than previous designs. Additionally, the G1 GPU is 20% faster on estimates compared to the previous generation.
Arm’s new hardware largely improves on-device AI
Here are some quotes of major industry partners, which features the choice of Samsung, Honor and Google:
“In Samsung, we are excited to continue our cooperation with ARM by taking advantage of ARM’s compute subcutam platform to develop the next generation of major mobile products. This partnership enables us to carry forward the limitations of on-device AI, which provides smart, sharp and more efficient experience for our users.” Vice President and Vice President of SOC IP Development Team in Samsung Electronics, Nose Seong
“In honor, our mission is to bring a premium experience for more users, especially through our upper mid-range smartphone. By taking advantage of ARM Lumax CSS platform, we are able to distribute smooth performance, intelligent AI features and excellent power efficiency that enhance everyday mobile experiences.” Respect
“SM2-Enhanced Hardware enables more advanced AI models, such as GEMMA 3, to run on a wide range of devices directly. As the SME2 is continuing on the scale, it will basically capable of deploying mobile developers in the ecosystem. Iliyan Malchev will be able to deploy the next generation of AI in the ecosystem.
