Its X13 range includes more than 50 server and storage systems in 15 families that are both faster and more efficient than Supermicro’s existing gen 3 Xeon lineup, we’re told.
Supermicro president and CEO Charles Liang said: “Our systems with the new Intel 4th Generation Xeon Scalable Processors are able to deliver new highs in both overall performance and performance per watt. In addition to the processor, every major sub-system has been dramatically upgraded w/ up to 2X improvement in memory performance with DDR5, 2X better I/O bandwidth w/ 80 lanes of PCIe Gen 5 I/O that will support higher performance accelerator cards, and 400 Gbps networking and improved manageability and security.”
The X13 systems will also support coming technologies like CXL 1.1 to create a unified and coherent memory space between CPUs and accelerators. They also support:
- DDR5-4800 MTps memory with up to 1.5x the performance and capacity of DDR4.
- PCIe 5.0 for double the I/O bandwidth of the previous generation and 80 lanes per CPU.
- EDSFF E1.S and E3.S storage form factors, set to provide more speed, capacity, and density for hot and warm-tier storage applications.
Intel’s Sapphire Rapids CPUs include four accelerator engines, one of which, the Data Streaming Accelerator, reduces CPU overhead on storage, networking, and data-processing-intensive tasks.
We’re highlighting Supermicro’s storage-related systems in its 15 new X13 product families:
- BigTwin systems provide better density, performance, and serviceability with dual Sapphire Rapids processors per node and hot-swappable tool-less design – for cloud, storage, and media workloads.
- FatTwin with multi-node 4U twin architecture with 8 or 4 nodes (single processor per node). Front-accessible service design allows cold-aisle serviceability, with highly configurable systems optimised for data centre infrastructure with compute and storage density and options. The Fat Twin supports all-hybrid hot-swappable NVMe/SAS/SATA hybrid drive bays with up to 6 drives per node (8-node) and up to 8 drives per node (4-node). It’s suited for running EDA applications.
- CloudDC with 2 or 6 PCIe 5.0 slots and dual AIOM slots (PCIe 5.0; OCP 3.0 compliant) for maximum data throughput. They have tool-less brackets, hot-swap drive trays and redundant power supplies for more efficient maintenance in data centers.
- Petascale Storage all-flash NVMe systems with EDSFF drives, allowing unprecedented capacity and performance in a single 1U chassis. The first in a coming lineup of X13 storage systems, this latest E1.S server supports 9.5mm and 15mm EDSFF media, now shipping from all the industry-leading flash vendors.
Bearing in mind our view that we are going to see greater integration between storage and AI/ML-focussed servers, we also note these two products:
- GPU Servers with PCIe GPUs – Optimized for AI, Deep Learning, HPC, and high-end graphics professionals, providing maximum acceleration, flexibility, high performance, and balanced solutions. Supermicro GPU-optimized systems support advanced accelerators and deliver dramatic performance gains and cost savings. These systems are designed for HPC, AI/ML, rendering, and VDI workloads.
- Universal GPU Servers – The X13 Universal GPU systems are open, modular, standards-based servers that provide superior performance and serviceability with dual processors and a hot-swappable, toolless design. GPU options include the latest PCIe, OAM, and NVIDIA SXM technology. These GPU servers are ideal for workloads needing AI training performance, HPC, and Big Data Analytics.
We think all x86 server vendors will quickly refresh their storage server lines with systems featuring Sapphire Rapids and PCIe gen 5 support. There should also be a storage industry-wide refresh of SAN, file and object array controllers in the next two or three quarters to add Sapphire Rapids processors, PCIe gen 5 connectivity, faster DRAM and EDSFF media. The result should be a significant increase in data capacity and access speeds.