- Xcena introduced the mx1 computational memory with thousands of RISC-V nuclei in FMS 2025
- MX1 offers a processing close to the data by reducing the CPU memory overload and enables the expansion supported by SSD at Petabyte scale
- The product roadmap includes MX1P this year and MX1s in 2026 that supports CXL 3.2
In the recent FMS 2025 event (previously Flash Memory Summit but now called Future of Memory and Storage to better adapt to its expanded approach), the South Korean Startup Xcena removed the wraps of its first product, MX1 Memory.
MX1 is based on PCIE GEN6 and the Compute Express Link 3.2 standard. By calculating directly next to DRAM, the chip can reduce the overload of mobile data from one side between processors and memory.
Known as close data processing, this approach could influence how the servers are designed in the coming years.
Thousands of risc-v nuclei
Xcena says that he has covered the “thousands” of internal RISC-V nuclei in the MX1 to handle workloads, such as vector database operations, analysis and memory consultations.
The expansion of memory backed by SSD allows Petabyte’s scale while adding compression and reliability characteristics.
Servethehome Reports that there are two different models available. The MX1P is expected at the end of this year (Xcena says that the work samples will be made available to the selected partners from October), while the MX1s, with PCIe Gen6 X8 Duals and additional characteristics, are scheduled for its launch in 2026.
Both will take advantage of the bandwidth and the broader flexibility offered by the CXL 3.2 standard.
The product won the “most innovative memory technology” award in FMS 2025, which makes it the second consecutive recognition of the company in the event, after being named “more innovative start” in 2024.
“Computational memory represents an emerging architectural approach that aims to accelerate performance and efficiency, particularly for data intensive.
“We are proud to recognize Xcena with MX1, the first computational memory controller in the world that admits CXL 3.XY enables high performance acceleration for intensive data loads such as AI and Analytics.”
With the hope of promoting interest in its product, Xcena, which was founded in 2022 as Metisx, offers a software development kit that includes drivers, execution time libraries and tools.
The battery is designed to adapt to standard environments so that developers can evaluate and implement MX1 in applications ranging from AI inference to memory analysis.
