DreamBig Semiconductor, Inc. has recently revealed its latest platform, called “MARS,” which aims to revolutionize the semiconductor industry by utilizing open standard Chiplets for a wide range of applications. The platform enables startups and companies of any size to scale-up and scale-out LLM, Generative AI, Automotive, Datacenter, and Edge solutions with optimized performance and energy efficiency.
By leveraging the open standard chiplets available in the “MARS” Chiplet Platform, customers can focus their investment on differentiating their silicon in specific areas to gain a competitive advantage and bring their solutions to market faster and at a lower cost. This is especially significant for the AI training and inference market, where application-specific solutions yield the best performance and energy efficiency.
Sohail Syed, the CEO of DreamBig Semiconductor, highlights how their open chiplet platform is disrupting the industry by enabling customers to combine their specialized hardware chiplets with scalable infrastructure, leading to affordable and efficient modular product development.
The “MARS” Chiplet Platform addresses two major technical challenges facing hardware developers in the AI server and accelerator industry: scaling up compute and scaling out networking. The Chiplet Hub, which features advanced 3D memory architecture, provides direct access to both SRAM and DRAM tiers, allowing for efficient data movement, caching, and processing. Additionally, the RDMA Ethernet Networking Chiplets offer unparalleled scale-out performance and energy efficiency with independent control over data path bandwidth and packet processing rate.
Steve Majors, the SVP of Engineering at DreamBig Semiconductor, emphasizes that customers can now focus on designing innovative AI compute and accelerator technology chiplets tailored to their applications. By utilizing the advanced features of the DreamBig Chiplet Platform, they can achieve maximum performance and energy efficiency.
The Platform Chiplet Hub and Networking Chiplets of the “MARS” platform offer open standard interfaces and support for CPU, AI, Accelerator, IO, and Memory Chiplets. The platform also provides secure boot and management of chiplets, a memory-first architecture, fully associative HW acceleration, HW DMA and RDMA capabilities, algorithmic TCAM HW acceleration, and a virtual PCIe/CXL switch for flexible resource allocation.
DreamBig Semiconductor is currently collaborating with customers on next-generation devices for various use cases, including AI servers and accelerators, high-end datacenters, low-end edge servers, petabyte storage servers, and automotive solutions.
Sehat Sutardja and Weili Dai, Co-founders and Chairman/Chairwoman of DreamBig, express their pride in the company’s achievements and its advancements in high-performance and scalable AI solutions.
To learn more about DreamBig Semiconductor’s “MARS” Chiplet Platform, visitors can attend the technology solution demo at CES 2024, held at The Venetian Expo in the Bellini 2003 Meeting Room.
About DreamBig Semiconductor:
Founded in 2019, DreamBig Semiconductor specializes in developing disruptive Chiplet Platforms for the AI revolution and digital world. Their solutions enable customers to bring next-generation high-performance, energy-conscious, affordable, scalable, and composable semiconductor chiplet solutions to market. The company’s expertise lies in applications such as Large Language Model (LLM), Generative AI, Datacenter, Edge, and Automotive markets.
The source of the article is from the blog bitperfect.pe