Where Compute Reigns | 3rd Gen Intel® Xeon® Scalable Processors - "Ice Lake" GIGABYTE servers and Intel Xeon Scalable Processors deliver incredible 1P/2P performance from edge to data center with incredible gains in I/O throughput and workload demands that need high performing CPUs with large, yet optimized, memory configurations. Enhanced I/0: Support for PCIe 4.0 allows for double the throughput of PCIe 3.0 for fast and large data transfers. Also, 64 lanes per socket with up to 128 lanes in a 2P configuration for fast data transmission from CPU to GPU, accelerator or storage. CPU Performance Boost: Now up to 40 cores per socket with greatly improved IPC on Intel''s 10nm architecture while operating CPUs at 105-270W. Also, an additional UPI lane is added for up to 11.2GT/s for a low latency interconnect between CPUs. Next Gen Memory: Support for DDR4-3200MHz (1DPC or 2DPC) DIMM. 8 channel memory has been added with up to 16 DIMMs per socket delivering up to 4TB of system memory via usage of 256GB DDR4 DIMM. Additionally, 2/4/6/8-way interleaves for optimal configuration. Arrival of Intel Optane PMem 200 Series: Optimized for improvements in latency while increasing system memory up to 6TB per socket as a mix of PMem and DDR. The 200 Series also has support for memory mode to expand capacity and app direct mode to act as fast flash storage. Next Level Security: New instructions and architecture deliver a high level of cryptographic operations for data privacy and protection. Cryptographic accelerators speed up encryption protocols as well as improvements in software guard extensions and memory encryption are added. AI Acceleration: AI inference and training are enhanced with DL Boost Technology for greater business, operational, and security insights. Applications include HPC, AI, and Media & Graphics. Why GIGABYTE G492-HA0 | Accelerator Computing | Dual root complexParallel processing with multiple GPUs and support for dual slot GPUs allows for the highest, dense GPU compute possible. To get even more GPU performance a balanced dual root complex allows for offers the lowest latency as system memory is accessed by GPU memory via PCIe switch, while bypassing the CPU. High-performance | NVLink and GPUDirect utilizedRequiring the ultimate performance in every aspect of a server is possible. GIGABYTE servers have been designed to support the cutting edge, high-end CPUs, GPUs P2P, and GPU-to-NIC bridges while utilizing GPUDirect and PCIe 4.0 for fast throughput and multi-GPU communication.