UnifabriX

growth-positive

UnifabriX Ltd., based in Haifa, Israel, has been granted a patent for protocol-aware provisioning of resource over CXL fabrics. This innovation focuses on enhancing the efficiency of resource utilization in datacenters by enabling dynamic provisioning of compute, memory, storage, and network resources. The patent describes a Resource Provisioning Unit (RPU) that facilitates protocol translations and mappings, allowing for the creation of large-scale compute-memory fabrics. These fabrics can support workloads such as Large-Language Models and Deep Learning Recommendation Models. The patent application was filed on January 11, 2025, and highlights the potential for optimizing low-latency communication between processes using shared memory.

growth-positive

UNIFABRIX LTD., based in Haifa, has filed a patent application for CXL PROTOCOL TRANSLATIONS AND SWITCHES under the International Patent Classification system managed by WIPO. The patent application, filed on December 15, 2024, and published on June 19, 2025, highlights the companys focus on innovation in memory technology. The inventors listed are Ronen Aharon Hyatt, Gaya Opal Hyatt, and Ethan Sharon Hyatt. This development is a positive step for UNIFABRIX LTD., indicating progress in their research and development efforts.

growth-positive

The article discusses UnifabriXs innovative use of Compute Express Link (CXL) technology to enhance CPU interconnectivity and memory scalability. CXL allows for dynamic memory expansion by connecting external devices to servers via PCIe, effectively adding more NUMA nodes and DRAM at runtime without shutting down the server. This advancement addresses the non-linear performance scaling of CPUs due to memory access latency. The article explores the implications of this technology, such as reduced need for pre-faulting and potential benefits for memory-intensive applications like AI and HPC clusters. UnifabriXs solution represents a significant shift in memory technology, offering new operational possibilities for various computing environments.

growth-positive

UnifabriX has introduced its CXL-connected external MAX memory device, which significantly enhances AI processing performance by addressing the memory bandwidth bottleneck in AI applications. The companys technology aims to bridge the performance gap between GPU server memory and flash storage, offering substantial cost savings and improved memory utilization. By integrating MAX memory, servers can achieve higher on-demand memory capacity and bandwidth, resulting in faster application performance. UnifabriXs solution provides a $1 million capex saving and a $1.5 million total cost of ownership gain, making it a cost-effective option for AI infrastructure.

growth-positive

UnifabriX, an Israel-based company, has secured $11 million in seed funding to advance its Smart Memory Fabric systems, which leverage CXL technology to overcome memory capacity limitations in CPU and GPU server systems. The company has developed a Smart Memory Node and a MAX composable memory device, which are already generating revenue through deployments in data analytics, high-performance computing, and cloud environments. UnifabriX is working directly with end customers and partners, including platform OEMs, to integrate memory pools into their product portfolios. The company plans a new funding round in the second half of 2025 and aims to upgrade its silicon process to align with PCIe Gen 6 advancements.

growth-positive

The article discusses a patent application by Unifabrix Ltd for a system that optimizes memory utilization in distributed computing environments. The system uses Compute Express Link (CXL) to manage and allocate unused DRAM across multiple hosts. A resource composer plays a crucial role in this setup, interfacing with kernel modules to gather data on memory usage and make informed allocation decisions. This technology aims to improve system efficiency and performance by repurposing underutilized memory resources. The patent is currently pending, indicating ongoing research and development efforts by Unifabrix Ltd.

growth-positive

UnifabriX has introduced the UnifabriX MAX, a high-performance memory provisioning solution designed to enhance AI capabilities. The product offers on-demand, workload-aware memory capacity and bandwidth, aiming to accelerate memory performance. The live demo showcases significant performance gains for industry-standard benchmarks, marking a positive development for the company in the AI technology sector.

growth-positive

UnifabriX Ltd, an Israeli company specializing in CXL-based memory solutions for high-performance computing (HPC), has announced the expansion of its Early Access Program (EAP) for its MAX memory and storage solution. MAX is designed to enhance memory bandwidth and capacity for HPC and AI workloads, offering up to a 10x performance boost. The company has already signed agreements with HPC entities for validation and is now extending its EAP to include companies upgrading their data centers for AI workloads. This expansion reflects the growing demand for advanced memory solutions in response to the increasing computational needs of AI and data-intensive applications.

growth-positive

UnifabriX is expanding its Early Access Program (EAP) for its MAX™ memory and storage machine following successful showcases at the HPC User Forum and Intel Innovation. The company has received strong interest and positive feedback from the AI community, highlighting the demand for advanced memory solutions. MAX is designed for high-performance computing (HPC) and AI workloads, offering features like Memory Pooling, Sharing, Tiering, and Acceleration. UnifabriX has signed agreements with leading HPC entities and is now including companies enhancing their data centers for AI workloads in its EAP. The expansion reflects the growing importance of addressing memory bandwidth bottlenecks in the face of increasing computational demands.

growth-positive

UniFabriX, a startup, has exited stealth mode by demonstrating its CXL 3.0-connected Smart Memory device aimed at enhancing data center and high-performance computing (HPC) memory pooling. The demonstration highlights the benefits of CXL solutions in improving HPC performance by bridging the gap between processors and memory. The companys Smart Memory Node, powered by their Resource Processing Unit (RPU), enables memory expansion and pooling, providing multi-core CPUs with necessary memory capacity and bandwidth. The demo involved connecting two Smart Memory Nodes across a CXL 3.0 fabric to multiple servers, showcasing CXL-enabled performance in an HPC benchmarking environment. The startup is funded by angel investors and VC Grove Ventures.

growth-positive

Unifabrix has launched its Smart Memory Node, a memory-pooling system based on the CXL (Compute Express Link) standard, aimed at addressing memory access bottlenecks in data centers. This system is particularly beneficial for high-performance computing (HPC) environments. The Smart Memory Node, equipped with Intels Sapphire Rapids processors, can support up to 32 terabytes of memory and offers advanced features like energy optimization and memory hierarchy management. Unifabrixs solution is already operational with several customers, primarily in the HPC sector. The company collaborates with various server manufacturers and HPC providers to integrate pooled CXL memory into infrastructures. Unifabrix is optimistic about the market potential, projecting significant growth as the industry transitions to CXL-based memory solutions by 2030.

growth-positive

UnifabriX is leveraging CXL (Compute Express Link) technology to enhance server and equipment utilization in data centers, aiming to unlock performance and scalability potential. The company focuses on providing CXL Smart Memory solutions for high-performance computing (HPC) data centers, addressing issues like DRAM bandwidth and capacity limitations. UnifabriXs technology is already operational, having been demonstrated live at the SC22 conference. The company introduced its Resource Processing Unit (RPU) to optimize CPU usage and system-wide bandwidth, promising a new way to structure data centers and solve performance challenges. UnifabriX positions itself as a pioneer in the CXL space, offering solutions that transcend traditional memory limitations.

growth-positive

UniFabriX, an Israeli startup, has developed a Smart Memory Node that utilizes CXL memory expansion technology to double the number of servers in a cloud service providers (CSP) datacenter racks. This technology allows CSPs to improve server/DRAM granularity and reduce electrical power consumption per server, thereby increasing revenue potential. The Smart Memory Node uses standard DDR5 DIMMs and offers memory checkpointing and a mediation layer for hyperscalers. UniFabriX has paying customers in the HPC market and is seeking OEM deals. The company is optimistic about capturing a significant share of the CXL memory market, which is projected to reach a $20 billion total addressable market by 2030.

growth-positive

UnifabriX, an Israeli startup, has developed a prototype called Smart Memory Node, which utilizes CXL technology to enable RAM sharing between servers. This innovation aims to reduce costs and increase flexibility in data centers by allowing servers to access a shared pool of memory. The company is targeting the supercomputer market initially and anticipates significant demand from cloud hosting providers. UnifabriXs founders, who previously worked at Intel, have leveraged their experience with CXL to create this product. The Smart Memory Node is currently in the prototype stage, with plans for future versions to support hybrid modes with DRAM and SSD CXL modules.

growth-positive

Unifabrix, an Israeli startup, showcased its Smart Memory Node at SC22, demonstrating its ability to enhance memory bandwidth and capacity for servers using CXL 3.0 technology. The technology was tested on Sapphire Rapids servers, resulting in a 26% increase in CPU core utilization. This innovation addresses the mismatch between local DRAM capacity and bandwidth, offering significant performance improvements for high-performance computing (HPC) applications. Unifabrixs solution is currently in pilot projects with several customers and is generating revenue. The company is targeting the HPC market initially, with potential expansion into AI/ML and analytics markets. Unifabrixs technology is CPU-agnostic and compatible with AMD, Arm, and RISC-V processors, and it could attract interest from OEMs and memory expansion software suppliers.

growth-positive

UnifabriX has exited stealth mode to unveil its CXL-based Smart Memory Node, which promises to revolutionize HPC performance through memory pooling, sharing, and CXL 3.0 fabric. The company showcased its product at SC22, demonstrating significant performance gains in CPU and memory utilization. UnifabriXs Resource Processing Unit (RPU) enhances CPU utilization and system-wide bandwidth, positioning the company as a leader in CXL technology. The company collaborates with Intel, integrating upcoming 4th Gen Intel Xeon processors and Agilex FPGAs into its solutions. UnifabriX aims to address inefficiencies in large-scale system deployments, enhancing performance across various demanding workloads, including AI and big data.