AMD's 5 Biggest Server Announcements At 'Next Horizon' Event
AMD's 'Total Commitment To The Data Center'
A day after Intel announced its response to AMD's top-performing server CPU, AMD provided its own rebuttal to the semiconductor giant's advances with major data center announcements at its Next Horizon event, including a reveal of AMD's next-generation CPUs and GPUs.
Held on Tuesday, AMD's Next Horizon event shed light on the Santa Clara, Calif.-based company's much-hyped 7-nanometer processors through the lens of the data center. During the presentation, AMD CEO Lisa Su said the company is making a "total commitment to the data center."
[Related: AMD Stock Price Nosedives As Channel Graphics Sales Slow Down]
Mark Papermaster, AMD's CTO, previously told CRN in August that 7nm chips will increase the company's competition with Intel, which has struggled with multiple delays of its 10nm CPUs that are now set for a holiday 2019 release.
"That's what AMD has always been about — innovation — but now it's got a razor's edge to fight against larger competitors that have been dominating the industry," Papermaster had said.
Here are the five biggest announcements from AMD's Next Horizon event.
AMD Reveals 7nm Zen 2 CPU Architecture
One of the most significant highlights from the event was the official reveal of the successor to AMD's Zen CPU architecture, which serves as the foundation for the company's Ryzen desktop and laptop CPUs, as well as its EPYC server CPUs that launched last year.
AMD said the processor based on its new Zen 2 architecture is the "world's first high-performance x86 7nm CPU." What that means is double the core density over the company's previous top CPU while using half the power with a 1.25x improvement in performance. The first Zen 2 processors are due out sometime next year, according to AMD.
AMD Dishes On Core Count For EPYC Rome CPU
The company also shed new details on its next-generation EPYC CPU based on Zen 2 architecture, code-named Rome. The new EPYC CPU will pack up to 64 cores and 128 threads, with increased instructions per cycle and double the compute performance per socket. Rome is due out in 2019.
AMD said EPYC Rome will be the industry's first PCIe 4.0-capable x86 server processor and provide double the bandwidth per channel. In a demonstration, the company said a single-socket version of the Rome CPU outperformed a top-line Intel Xeon dual processor running the C-Ray benchmark. The CPU will be compatible with first-generation EPYC sockets, the company added.
AMD Drops Details On Forthcoming 7nm GPU
In addition to revealing its next-generation CPU, AMD also provided details on the AMD Radeon Instinct MI60 for data centers, which the company said is "the world's first 7nm GPU." The graphics card is due out later this year and will provide the "world's only hardware virtualized GPU," the company added.
The forthcoming Radeon GPU features two times the density and a 1.25x performance gain over the company's previous top-line graphics card while using 50 percent less power. With 32GB of memory, up to 1 TB/S memory bandwidth speeds and support for PCI 4.0, the product is designed for deep learning, high-performance computing, cloud computing and rendering applications.
AMD Confirms Next-Gen Product Roadmap
AMD provided brief details on the company's next-generation product roadmap. With the company's 7nm Zen 2chips sampling now, the company said Zen 3 architecture, which is based on the company's improved 7nm+ process, is on track for 2020. Meanwhile, Zen 4 is in the "design completion phase."
On the GPU side, the company said the Radeon MI-Next is in development but did not provide any details on when the next-generation graphics card could come out. On the CPU side, the company said its Zen 3 architecture is also on track.
New EPYC-Based AWS Compute Instances
AMD also revealed a new partnership with Amazon Web Services that will allow the cloud service provider to serve three instances of Amazon Elastic Compute Cloud, or EC2, powered by AMD's EPYC server CPUs. These instances will provide "exceptional performance per dollar for general purpose and memory optimized workloads," according to AMD. The EPYC-based R5 and M5 instances are available now while the new T3 instance will become available within the next few weeks.
AMD said EPYC's core density drives the cost savings for the new R5 and T3 instances, which provide a "balance of compute, memory, and networking resources for web and application servers, backend servers for enterprise applications, and test/development environments with seamless application migration." Meanwhile, the company said EPYC's memory bandwidth advantage makes the new R5 instance ideal for in-memory processing, data mining and dynamic data processing.