Provision of artificial intelligence hardware and software

Descriptions

The system to be procured is state of the art GPU-powered servers that enhance and massively accelerate our data-driven AI research for medicines manufacturing. The system shall have: - GPUs: 8x NVIDIA A100 Tensor Core GPUs - CPU Specifications: 6912 CUDA cores & 432 TF32 Tensor Cores per GPU - GPU Memory: 80GB per GPU - 640GB total - GPU Interconnet: 6x NVSwitch - Host CPUs: 2x AMD EPYC 7742, total 128 cores / 256 threads - System Memory: 2TB ECC Reg DDR4 - System Drives: 2x 1.92TB NVMe SSDs - Storage Drivers: 8x 3.84TB NVMe SSDs - Networking: 8x single-port NVIDIA ConnectX-6 VPI 200Gb/s Infiniband. 2x NVIDIA ConnectX-6 VPI 200Gb/s Ethernet or 8x single-port NVIDIA ConnectX-7 VPI 200Gb/s Infiniband. 2x NVIDIA ConnectX-7 VPI 200Gb/s Ethernet - Operatin system: Ubuntu Linux - Power Requirement: 6.5kW - Operating Temperature Range: 5ºC to 30ºC (41ºF to 86ºF) The System (DGX with NVIDIA A100 AI System )to be provided by Scan Computers International Limited meets all of these requirements and is built to a high standard which will in turn maximise it’s up-time and lifespan at University of Strathclyde which will directly the engage the research projects the system will support. The system is required as soon as possible to support the research timelines required by the University: Nvidia, as key hardware component suppliers, work closely with Scan who are the providers of the assembled GPU servers we require for our research. Nvidia prioritise hardware delivery to Scan and as such, Scan are in a position to deliver the final product in an 8-week timeframe. This is exceptionally fast given current global hardware shortages which are causing large companies and even governments to wait up to 12 months for hardware. This is an essential elements of this purchase. Uniqueness of Scan offering: Scan will provide after care support not only in terms of support for the commissioning of the hardware but in terms of familiarisation with software platforms which will be required by the research teams at Strathclyde. Via Scan, we will have access to Nvidia’s leading AI experts. This will allow us to unlock the full potential of the hardware in a timely manner which is crucial for delivery on a number of significant, large value (10mGBP) research programmes. Scan’s ability to assemble the GPU servers and support of the software that runs on top makes them uniquely placed to deliver our bespoke AI, modelling and simulation platform and within the timelines required of our funded research projects, which are already live and in need of the proposed compute resource.