AMD already has me excited over their Ryzen CPUs for consumers and gamers, but the Zen architecture is also going to do big things in the high-performance server market - starting with Naples.
Naples is AMD's high-end server part based on the Zen architecture that AMD is touting as an 'optimized GPU server platform' thanks to its 128 PCIe 3.0 lanes, allowing for 4-6 'direct attached GPUs'. This means it can take 32 x NVMe devices, and 4 x graphics cards - all on a 1U rack. This same 1U rack will sport 2 x InfiniBand EDR interconnects for super-fast data communication between the storage, and server systems.
As for specs, we're looking at up to 32 Zen CPU cores with 64 threads of performance, 64MB of L3 cache, a base clock of 1.4GHz and a turbo clock of 2.8GHz. A dual-CPU version of Naples would rock a massive 64/128 CPU cores/threads with some ultimate performance, especially when you throw Vega-based graphics cards and NVMe-based storage into the mix.
The high-end Naples platform will also support massive 8-channel memory, so they can take mass datasets for use cases like molecular dynamics, rendering, graphics and data analytics, and more. Better yet, there will be support in the 1U and 2U racks for:
- 1U (2 x CPU, 4 x GPU)
- 2U (2 x CPU, 8 x GPU)
- Sled: 2P (multi-CPU, 1-2 x GPU)
- Blade: 2P (multi-CPU, 1-3 x GPU)