Last year at the ISC High Performance conference, Cavium was one of the first out of the gate with 64-bit ARM servers for HPC. To learn more about this up and coming processor technology, we caught up with Gopal Hedge, VP/GM Server Processor Group at Cavium Inc.
insideHPC: Cavium is known for being an early adopter of 64-bit ARM for high performance computing. How is ThunderX line being positioned for HPC workloads?
Gopal Hedge: ThunderX has significant differentiation in the 64-bit ARM market as Cavium is the first ARMv8 vendor to deliver dual socket support with full ARMv8.1 implementation and significant advantage in CPU cores with 48 cores per socket. In addition, ThunderX supports large memory capacity (512GB per socket, 1TB in a 2S system) with excellent memory bandwidth and low memory latency. In addition, ThunderX includes multiple 10 GbE / 40GbE network interfaces delivering excellent IO throughput. These features enable ThunderX to deliver the core performance & scale out capability that the HPC market requires.
ThunderX is Ideal for applications with characteristics that include:
- Heavy emphasis on parallel integer codes
- Significant string operations
- Irregular data patterns
- Not easily vectorized or limited vectorization
- Applications that require excellent memory subsystem performance
insideHPC: How is ThunderX important to HPC users?
Gopal Hedge: With ThunderX Cavium is offering an outstanding TCO to an industry that has been seeking architectural innovation without having to trade off features or performance. Cavium has also been aggressively driving a thriving software ecosystem for the HPC community.
insideHPC: Is ThunderX powerful enough for HPC workloads, or does it require acceleration with GPUs or FPGAs?
Gopal Hedge: We have customers who our using ThunderX in both types of configurations – ThunderX only and ThunderX with Nvidia GPUs. As we note above ThunderX is well suited to many HPC applications as a standalone solution. An example is for the emerging HPC cloud deployment for distributed computing and analytics. If customers have codes or configurations that they prefer to leverage GPUs, we support this through our partnership with Nvidia.
insideHPC: What industries do you see as first adopters of ARM for HPC? Do you have any success stories there?
We have had a breadth of deployments across Supercomputing Centers, Government Labs, Research Institutes and Commercial HPC customers worldwide. We will be sharing some of these customer experiences publicly over the coming months.
insideHPC: What will you be showcasing in your exhibit at ISC High Performance in June?
Gopal Hedge: We are currently finalizing our plans for the ISC event however we expect to have a range of ThunderX partners in attendance including:
- System Vendors: Gigabyte, E4, Cray, Lenovo, Penguin
- PCIe adapters: Nvidia, Mellanox
- OS Vendors: Canonical, Red Hat, SUSE
- Compilers, Libraries, Tools Vendors: Allinea, Pathscale, Bright Computing
We will be showcasing a range of system configurations and software demonstrations and solutions at the event.