One node with 4 sockets can populate more RAM, up to 1024GB RAM (256GB/DIMM). And this 3rd Gen can reach RAM speeds of 3200MHz. So, speed and capacity have improved.
Eight half-length or four double-wide expansion card slots for accelerators or I/O cards fit in GIGABYTE's Intel Xeon Scalable servers.
GIGABYTE has the R-series and G-series to accommodate configurations of GPUs, networking, or other I/O interfaces, while using Intel processors.
GIGABYTE's "Building Block" concept allows for easy swapping of parts. Customers can choose which type of module they want, such as PCIe or SXM.
GIGABYTE has had a long, strong relationship with Intel allowing for GIGABYTE servers to be ready for next gen technology.
Single node computing performance has increased, and GIGABYTE has developed a layout and thermal solution to handle the heat and maintain peak performance at the same time.
For a balanced, dedicated CPU and GPU server, GIGABYTE has a 4 socket R292 series server and a 2 socket G-series servers ready. They combine compute, storage, network and expandability suitable for data analytics and computational science.
A more cost-effective model of HPC computing, hybrid HPC, allows expandability through the cloud. GIGABYTE's G-series and R292 can easily reallocate on-premises nodes for efficiency as well as access on-demand resources in the cloud.
AI flourishes with a high per-core performance, multi-socket, CPU platform with substantial I/O bandwidth to maximize the transfer rate between processors, cache, memory, and GPUs. In addition, GPU accelerators speed up processing through parallel computing. System is also able to scale-out easily.