Kathy Gibson reports from Pinnacle Tech Tour – Artificial intelligence (AI) is the word on everyone’s lips, but customers can’t get any value from it without the right hardware stack.
JP De Villiers, senior account manager at SuperMicro, points out that the vendor has the broadest portfolio of IT infrastructure building blocks.
These include cloud-optimised servers, accelerate computing for AI, 5G edge and IoT-embedded devices, storage and networking solutions. Building blocks of the SuperMicro server portfolio include Intel, AMD Epyc and Nvidia.
When it comes to AI, different SKUs address markets including various different AI workloads that target the specific needs of each workload.
GPU-optimised SuperMicro servers, racks and workstations that help customers achieve massive performance increases in large scale AI training workloads, high performance computing (HPC) and AI workloads, enterprise AI inference and training, visualisation and design, content delivery and virtualisation, and AI edge.
De Villiers outlines the SuperMicro SKUs that offer GPU-optimised compute for various workloads. They are scalable for different organisations’ requirements.
The company’s flagship offering is the 8U HGX H100 8-GPU Sys-821GE-TNHR system powered by Intel 4th or 5th Gen Xeon scalable processors. With liquid cooling, this system lets customer double their CPU density.
SuperMicro builds systems in the Nvidia Grace Hopper superchip for giant scale AI and HPC, and the Grace CPU superchip for CPU computing. These systems will massively boost performance in air cooled or liquid cooled systems.
“We are a hardware vendor, but we are also playing more towards total IT solutions,” says De Villiers.
These total IT solutions offer higher value for customers, he points out. They are rack-scale plug and play solutions that are scalabale and optimised for specific use cases.
SuperMicro products have been instrumental in building some major systems around the globe – from end users to cloud providers who are building GPU services for wider use.
As a 100% channel focused company, SuperMicro has the broadest product portfolio, De Villiers says, “We provide what the client needs – not what we have. And there is no hardware or vendor lock-in.”
To get going on AI, partners and customers need to define a targeted strategy, De Villiers says.
“Data is important,” he points out. “You have to ensure there is quality data. Training the right algorithms is the next step.”
Of course, computing power is what will make this all happen.
“Ethical considerations are important: AI must be transparent, and for the good of all,” De Villiers points out.