Fujitsu is a Japanese company with extensive experience in the design and production of various electronic components, including processors. The company has developed and produced chips for various applications, including for use in computers, servers, embedded systems and other devices.
Now comes confirmation that Fujitsu engineers are working on a next-generation processor for computer applications artificial intelligencehigh performance computing (HPC) and data center applications. The processor Monaco promises formidable performance, being based on well 150 core Armv9 optimized and also allowing the use of any accelerators.
Fujitsu Monaka, a high-performance ARM processor for the data center and HPC world
In the recent past, Fujitsu has designed and manufactured A64FX, a processor known to be used in the Fugaku supercomputer. Presented in 2020, Fugaku it is one of the most powerful systems in the world based on ARM architecture. Monaka seems to follow in the wake of A64FX even if it will use a much higher number of cores (A64FX uses 48 for each single unit) and will be created using a construction process 2 nm Of TSMC. The cores will be spread across multiple die and accompanied by SRAM and I/O modules.
Although it is possible to draw several similarities with A64FX, Monaka promises to be a processor that does not look at supercomputers at all but rather has the objective of establishing itself in data centers and high-performance computing.
In the context of processors, the term “length of the vector” or “vector size” (vector length/size) refers to the number of elements that can be processed simultaneously in a single vector instruction. For example, if a processor has a vector length of 128 bits, it means that it can perform operations on 128 bits of data in one single instruction. This can correspond to four single-precision floating-point numbers (32 bits each) or two in double precision floating point (64 bits each).
It is known that Monaka will be able to handle one vector size up to 2048 bit, will support RAM memories DDR5 (probably not HBM) and connectivity PCIe 6.0 with CXL 3.0 to connect accelerators and accessory components at high speed.
The overall design of the processor is a 3D chiplet, a disaggregated architecture that allows you to focus on maximum scalabilityexactly as AMD does with its Ryzen and EPYC CPUs.
An ARM processor that also focuses on energy efficiency
L’energy efficiency represents one of the key features of Monaka. In fact, Fujitsu has the ambitious goal of improving the efficiency of competing processors by two times.
The chip is currently in the preparation phase and, according to what has emerged, it will not arrive on the market before 2026-2027, also keeping in mind the extremely miniaturized construction process on which it is based. Fujitsu, however, seems to have estimated how rival processors will evolve in 3 years from now and deduced the first conclusions.
Monaka is intended to offer robust security mechanisms: there will be for example CCA (Confidential Computing Architecture) of ARMv9-A, an isolation technology that is part of ARM’s series of innovations aimed at protect data and the code wherever the computation occurs. CCA in turn rests its foundation on Dynamic TrustZone: it allows you to create a secure execution environment and isolated within a system, where code and data can be protected from external attacks.