Saturday , 22 September 2018
Home >> P >> Processors >> Lowering TOC with single-socket AMD EPYC processors

Lowering TOC with single-socket AMD EPYC processors

​AMD EPYC - No-compromise single-socket processors

AMD EPYC – No-compromise single-socket processors

One of the most disruptive aspects of AMD’s EPYC processors was the company’s focus on delivering a no-compromise single-socket processor server solution. And the savings that await companies for some types of workloads are huge.

See also: AMD integrates Ryzen PRO and Radeon Vega graphics in next-generation APUs

Earlier this month I spoke to Raghu Nambiar, AMD’s CVP and CTO of Datacenter Ecosystems and Application Engineering, and Dan Bounds, AMD’s Senior Director for Data Center Products on the subject of SDS – Software Defined Storage – and how EPYC is well placed to lower TOC is ways previously not possible.


AMD debuts embedded EPYC and Ryzen processors


AMD debuts embedded EPYC and Ryzen processors


AMD debuts embedded EPYC and Ryzen processors


AMD debuts embedded EPYC and Ryzen processors


AMD debuts embedded EPYC and Ryzen processors



































































So what is a “no-compromise single-socket processor server solution”? Put simply, it means that customers don’t have to make do with arbitrary limitations when it comes to I/O, or memory bandwidth and performance when they choose single-socket over dual-socket servers, which in turn allows customers to make a no-compromise choice for underutilized servers, and benefit from lower power consumption and lower capital expenditure.

Each EPYC processor package can support up to 2TB of DDR4 RAM over eight channels, and has 128 PCIe lanes, which means that the platform has more than two-and-a-half times the I/O density of a processor such as Intel’s Xeon SP Series.

And SDS is an interesting space to consider because it combines all the demands of big data, data analytics, and offline analytics into a single workload.

So what makes EPYC so well suited a solution to the SDS workload? Well, according to Nambiar and Bounds, there are a number of reasons.

Take, for example, how storage and memory are converging in the enterprise space, and EPYC is well placed to take advantage of this. EPYC, with its eight memory channels and 128-lanes of PCI, means that EPYC-powered servers can handle larger memory footprints than those powered by Intel silicon, and there’s also a no need to separate PCI switches to be used to take advantage of NVMe storage.

All this allows for greater memory-to-core, and cores-to-spindle/drive ratio in the server room.

Another factor to consider is how the software licensing landscape has changed. For example, VMware vSphere licensing is based on the number of CPUs in a system, and there are special bundles for single-socket systems. And the savings aren’t measured in a few dollars, but in thousands of dollars.

According to an April 2018 report by Demartek, for a one-node system, a single-socket server can deliver $7,610 in savings on VMware licensing costs for vSAN and vSphere, while for a three-node cluster, these savings rise to $22,830.

And these savings are simply as a result of switching from dual-socket to single-socket. That’s it.


amdepyc1.jpg


amdepyc2.jpg


amdepyc3.jpg


amdepyc4.jpg


amdepyc5.jpg

























































































































































And these figures don’t include savings that come from reduced power consumption and lowered cooling costs.

I also talked with Nambiar and Bounds about what the future has to offer, and it’s clear that there are a lot of changes coming. Here are a few things that are on the server horizon:

  • Multi-chip modules will bring greater flexibility, especially when it comes to storage class memory
  • DDR5 memory which promises to be twice as fast as DDR4
  • Huge increase in NVMe density is set to double yearly for the next few years

The server market hasn’t been this exciting for years. Maybe even decades.

Previous and related coverage:

See also:

close
==[ Click Here 1X ] [ Close ]==