At 8U the Supermicro GPU SuperServer 820GP-TNAO is a workhorse server for applications in high-performance computing, AI and Deep Learning. It features NVIDIA’s NVLink with NVSwitch for support of up to 8x HGX A100 SXM4 GPUs. Outfitted with 3rd generation Intel Xeon Scalable processors it can also provide up to 12TB of memory.
With 10x PCIe 4.0 x16 slots the Supermicro GPU SuperServer 820GP-TNAO offers optimal I/O potential to access the 8x HGX SXM4 GPUs in the chassis through remote direct memory access (RDMA). The PCIe slots reside on a separate circuit board from the NVIDIA NVLink board which supports the SXM4 GPU cluster. The NVLink board is placed just behind the fans on the front of the chassis and pulls out like a drawer for easy access. 6x drive 2.5-inch hot-swap drive bays on the pull-out section above provide hybrid storage options. Power is provided by a single 6000W PSU featuring an 80 PLUS titanium level classification.
Outfitted with dual 3rd generation Intel Xeon Scalable processors this system can support 8x to 40x cores per socket for up to 80x cores of processing power with top-of-the-line Platinum CPUs installed. The thermal design power for the CPUs is capped at 270W, which includes most SKUs in the lineup. Only Platinum processors provide up to 40 cores, while Gold processors will provide 32 cores maximum and Silver top out at 20 cores. All 3rd Gen processors provide PCIe 4.0 and 8-channel memory architecture for superior memory density compared to the previous generation.
With both processors installed, the system will support 32x active memory module slots with 16x slots per CPU. 8-channel memory architecture enables 2x DDR4 modules to be installed per channel but only 1 Intel Optane Persistent memory (PMem) series 200 module can be installed per memory channel and must be paired with DRAM modules. ECC DDR4 Registered (RDIMM) or Load-Reduced (LRDIMM) memory modules are supported including 3DS versions of both, which offer greater DRAM density. At full capacity using DCPMM and DRAM the system will support up to 12TB of memory or up to 8TB using just DRAM.
Primarily used for GPU augmented computation and RDMA applications, the Supermicro GPU SuperServer 820GP-TNAO has only 6x 2.52-inch storage bays on the front of the chassis. 4x of the drive bays are dedicated for support of NVMe drives. The bays can support SAS, SATA and NVMe storage options. 2x PCIe 4.0 slots on the system board can be used for NVMe or SATA3 format M.2 drives but only to boot the system.
With the GPUs housed on a separate NVLink board with NVSwitch technology providing GPU-to-GPU interconnect, a separate PCIe switchboard provides up to 10x PCIe 4.0 x16 slots for support of additional CPU-to-GPU interconnect. Also provided are 2x 10GbE BaseT ports for built-in network connectivity.
Compatible with the integrated platform management interface 2.0 (IPMI) the system also provides for virtual media over LAN and KVM-over-LAN support. A dedicated RJ45 port on the back of the system provides access to the baseboard management controller (BMC). The Managing the system remotely requires an out-of-band (OOB) license key. In addition, the system can also be addressed using Supermicro’s proprietary management software SuperDoctor 5 (SD5) plus Watch Dog. Additional tools include the Supermicro Update Manager (SUM) to ensure the latest firmware and BIOS, plus Supermicro Power Manager (SPM) and Intel Node Manager.
Designed for high-performance computing with up to 8x SXM4 GPUs, the Supermicro GPU SuperServer 820GP-TNAO is an extremely capable appliance for AI, AI training and inference plus deep learning. With support for dual 3rd gen Intel Xeon Scalable processors and Optane PMen series 200 modules it can support up to 12TB of memory for significant in-memory processing power.
If you know what you want but can't find the exact configuration you're looking for, have one of our knowledgeable sales staff contact you. Give us a list of the components you would like to incorporate into the system, and the quantities, if more than one. We will get back to you immediately with an official quote.