Nvidia’s Bluefield DPU is designed in Israel and includes an array of hardware accelerators and general-purpose ARM cores that speed networking, security and storage applications in the data centre and for supercomputer designs.
Idan Burstein, a principal architect leading the BlueField design team detailed some of the accelerators, including IPsec, TLS, AES-GCM, RegEx and Public Key Acceleration for security, NVMe-oF, RAID and GPUDirect Storage for storage, RDMA, RoCE, SR-IOV, VXLAN, VirtIO and GPUDirect RDMA for networking, andoffloads for video streaming and time-sensitive communications.
The architecture described at Hot Chips is moving into several of the world’s largest clouds as well as a TOP500 supercomputer and integrated with next-generation firewalls. It will soon be available in systems from several OEMs including Atos, Dell, Fujitsu, Asus, Gigabyte and Supermicro, supported with software from more than a dozen other partners.
- Nvidia stakes its claim to the data centre
- Nvidia to rent out its supercomputers
- UK to look at national security issues in Nvidia-ARM deal
- China set to block Nvidia-ARM deal
Both Intel and AMD are also looking at ways to use hardware accelerators and separate processing chips to boost performance in data centre applications.
The BlueField3, sampling next year, will drive networks up to 400 Gbit/second with the equivalent of 300 x86 cores. This has up to 16 ARMv8.2+ A78 Hercules cores with up to 16 Programmable Datapath Accelerator cores that can handle 256 threads.
The BlueField-4 will deliver an order of magnitude more performance than this with the addition of Nvidia’s AI computing technologies.
Other articles on eeNews Europe
- Intel shows engineering silicon of its biggest ever data centre chip
- VE-Vides Project looks to trustworthy IP and verification
- UK launches Serapis framework for space technology
- Rolf Segger on the chip shortage
- Intel, Synopsys set for trademark battle
- The greening of Onsemi