Although the development of ARM processors has been relatively recent, their use in high performance computing has mushroomed over the past 18 months and now accounts for some 20% of all HPC workloads. Applications like machine learning, artificial intelligence and simulations are driving the demand for higher performance in terms of capacity and efficiency and high performance computing relies on processors that can lift it to the next level.
Within the past 18 months Futjisu chose ARM cores when it started building its Post-K supercomputer in Japan. In the UK, supercomputer Islamabard – which will be used, among other purposes for large-scale simulations – incorporates more than 10,000 64-bit ARM processors at its core.
Seattle-based supercomputer maker Cray is already drafting ARM chips for its existing machines and has just announced that its new Cray Blade will also be based on 64-bit Armv8-A architecture.
Cray Blade will use the Cavium ThunderX2 processors and will feature a full software setting, including the Cray Linux and Cray programming environments, and ARM-optimized compilers, libraries, and tools for running supercomputing workloads.
It uses FORTRAN and C/C++ compilers while the software libraries are designed to support extreme parallel processing. Also, it uses storage and processing virtualization technology.
The supercomputer maker improved its compiler and programming environment to achieve a higher performance from the Cavium ThunderX2 processors. When compared with 135 standard HPC benchmarks, Cray’s compiler showed a better performance in two-thirds of the cases and more than 20% higher performance in one-third of the tests compared with other public domain ARMv8 compilers from LLVM and GNU.
Cray’s chief marketing officer, Fred Kohout, says that the integration of ARM processors into the Cray XC50 will offer customers the world’s most flexible supercomputer, which can be used for big data analytics, deep learning and simulations and a variety of other functions.
Many big names are interested in Cray’s latest development and are collaborating with the Seattle-based company, including several labs in the United States Department of Energy and the GW4 alliance, a coalition of four leading, research-intensive universities in the UK.
Key takeaways:
- High performance computing is looking to ARM processors to lift their computing power to the next level.
- Cray Blade uses Cavium Thunder X and incorporates ARMv8 compilers.
- Two of the world’s top supercomputers currently being built are using ARM at their core.