At Virtual SC20, Dr. Rene Meyer, VP of technology at AMAX Information Technologies, spent time with us explaining developments and trends driving new levels of HPC-class performance in support of advanced AI solutions.
AMAX is focused on application-tailored cloud, data center, open architecture platforms, HPC, deep learning and OEM server manufacturing solutions. The company’s customers range from Fortune 1000 companies seeking greater efficiencies for data center cost savings to software startups in need of a manufacturing partner to design and launch products.
Much of AMAX’s work these days is focused on integrating server solutions to meet rising AI/machine/deep learning demand – and that means adopting the latest server gear, including next-generation bus technology.
What we see currently in the industry…is the transition from a PCIe Gen 3.0 to a PCIe Gen 4.0 interface driven by the new AMD Epyc CPUs. And that enables a whole new ecosystem, which sees acceleration on the compute side, especially the GPU compute side, since you open up the bandwidth between the CPU and the GPU. But this also enables a faster 200-Gig fabric and fastest storage solutions on the NVMe side. So what we are seeing, and what our customers are asking for, is (Nvidia) DGX A100 compute power combined with a network fabric typically provided by Mellanox combined with intermediate storage solutions, where you would store your data lakes for AI training applications.
