AMD's Naples CPUs Cater To Data Centers

By Victor Thomson , Mar 08, 2017 03:30 AM EST

AMD has announced its upcoming Naples platform. However, the official name of the product name will not be Naples and several name options are still debated internally at AMD.

According to AnandTech, Dr. Lisa Su, CEO of AMD, the launch of Naples CPUs for data centers will be an important move for AMD, as the enterprise market can bring billions of dollars to the bottom line. If AMD will be able to rejoin the x86 enterprise niche and become again a big player in the field, it can take a significant market share from Intel as well as stop some of ARMs ambitions.

Naples is based on AMD Zen architecture. The top end processor will feature a total of 32 cores. The CPU will be capable of a total of 64 threads with simultaneous multi-threading (SMT). This will be paired with a total of 16 DIMMs (up to two DIMMs per channel for eight channels of DDR4 memory). A single Naples CPU will support 128 PCIe 3.0 lanes.

Coming with USB, internal IO for storage and other things, Naples can be considered as a system-on-a-chip (SoC). According to PC Mag, Naples is aimed to compete with Intel's Xeon processors in the PC server market. The CPU will be offered as either a dual processor platform (2P) or a single processor platform (1P). In dual processor mode the system will provide 64 cores and 128 threads.

As part of AMD's Infinity Fabric, each processor in dual mode will use 64 of its PCIe lanes as a communication bus between the processors. Bandwidth is designed to be on the order of PCIe, but the Infinity Fabric uses a custom protocol over these lanes. Each core uses 64 PCIe lanes to talk to the other, allowing each of the CPUs to give 64 lanes to the rest of the system for a total of 128 PCIe 3.0.

With eight channels and two DIMMs per channel, on the memory side the Naples CPU will support up to 2 TB of DRAM per socket, for a total of 4 TB in a single server. A single CPU provides a total memory bandwidth of 170 GB/s.

© 2020 ITECHPOST, All rights reserved. Do not reproduce without permission.
Real Time Analytics