Earlier this week, Intel and Barcelona Supercomputing Center (BSC) said that they would invest 400million euros (about US $426million) to set up a laboratory In addition to developing new processors based on risc-v architecture that can be used to build zettascale supercomputing, the laboratory will also focus on artificial intelligence (AI) and autonomous vehicle.
After its establishment in Barcelona, the research laboratory will receive 400million euros from Intel and the Spanish government within 10 years.
The main purpose of the joint research laboratory is to develop chips based on the open source risc-v instruction level architecture (ISA), which can be used in a wide range of scenarios such as AI accelerators, autonomous vehicle, high performance computing (HPC).
Last year, Intel tried to acquire sifive, the leading developer of risc-v CPU, which is also one of the top sponsors of non-profit risc-v International (about US $21.3 million).
However, the establishment of the laboratory does not mean that Intel will develop risc-v processors for the supercomputing platform of the first generation zettascale here, but that the company intends to make additional investment in risc-v.
In addition, the increase in spending on risc-v processors in the next few years does not mean that Intel will pay less attention to the existing x86 product line.
If all goes well, the engineers of the joint laboratory are expected to design a CPU technology more suitable for zettaflops level supercomputing. As an additional reward, Intel's chip OEM service department is also expected to develop and manufacture related CPU / SOC for partners.
Jeff McVeigh, vice president and general manager of Intel supercomputing, said:
High performance computing is the key to solving the most challenging problems in the world, and we just have such an ambitious goal, that is, to sprint to the zettascale era of HPC.
The establishment of Barcelona Supercomputing Center places our common vision for achieving this goal and emphasizes open and sustainable design methods. Intel is very pleased to reach a new cooperation with Barcelona Supercomputing Center to embark on this journey together.
In fact, as early as last year, Intel has set a goal for itself, that is, by 2027, it is called zettaflops level supercomputing platform for short - which means that the company will improve the performance of supercomputing by a thousand times in about five years.
Obviously, this requires new computing and system architecture, high-speed memory and i/o interfaces, novel manufacturing processes, and complex chip packaging methods.
But before that, Intel will try to combine the company's x86 general kernel with Xe HPC GPU architecture. The first concept tea is Falcon shores, which is currently under development.