IBM (International Business Machines) unveiled Power Systems LC, a new line of power servers aimed at the HPC, cloud, and big data analytics markets. IBM claims its newly unveiled three Power Systems LC servers cost 50% less than a comparable x86 server and provide twice the level of performance.
Sumit Gupta, previously of NVIDIA and now VP of OpenPower high-performance computing at IBM, commented a, “This is our re-entry into the HPC market. This is our first OpenPower-based HPC server, and our go-to-market in the HPC space is an accelerated approach, which means not just CPUs but using CPUs and GPUs together. We still obviously offer a CPU-only version for clients who have not migrated their code to take advantage of GPUs yet.”
The Power Systems LC servers are designed and based on technologies of OpenPOWER Foundation. The system S822LC sales literature notes the inclusion of CAPI ( the Coherent Accelerator Processor Interface ) that enables close integration of accelerators with the POWER8 processor.
The FPGA used to support CAPI is important to enterprise customers and potentially to the HPC community. FPGA’s are becoming more important to enterprise customers in Internet search and other areas. The technical literature notes an energy hierarchy with FPGA’s being the most energy efficient.
IBM notes that CAPI on POWER8 provides:
“a high-performance solution for implementation of customer-specific, computation-heavy algorithms on an FPGA. This innovation can replace either application programs running on a core or custom acceleration implementations attached via I/O. CAPI removes the overhead and complexity of the I/O subsystem, allowing an accelerator to operate as part of an application. The IBM solution enables higher system performance with a much smaller programming investment, allowing hybrid computing to be successful across a much broader range of applications.”
IBM provides a specific use case, “CAPI can be much more than an acceleration platform. The IBM Data Engine for NoSQL uses the coherent CAPI interface to revolutionize database storage access, making 56 TB of external flash data function like in-memory storage.”
No word yet on NVlink or GPU performance which is critical to the IBM NVIDIA Volta plus NVlink 2017 Delivery for $325M DOE Procurements.


Leave a Reply