Google Cloud users now have new virtual machines to experiment with. The company today announced that its N2D VMs, which are built upon second-gen AMD Epyc processors, are now available as a public beta in certain regions.
AMD revealed its second-generation Epyc processors in August 2019. The chips are built using TSMC's 7nm process--which means AMD beat Intel to the latest-and-greatest process for the first time--and are based on the Zen 2 microarchitecture.
Unfortunately neither AMD nor Google Cloud disclosed the exact processor used in the N2D VMs. Instead they focused on highlighting the new offering's performance compared to previous-generation N1 VMs built around various Intel processors.
Google Cloud said in its announcement that the N2D VMs excel in these situations:
- General-purpose workloads that require a balance of compute and memory, like web applications and databases, can benefit from N2D’s performance, price, and features. N2D VMs are designed to provide you with the same features as N2 VMs including local SSD, custom machine types, and transparent maintenance through live migration, while features like large machine types with up to an industry-leading 224 vCPUs, the largest general purpose VM on Compute Engine. At the same time, N2D instances provide savings of up to 13% over comparable N-series instances, and up to a 39% performance improvement on the Coremark benchmark compared to comparable N1 instances1.
- HPC workloads such as crash analysis, financial modeling, rendering and reservoir analysis, will benefit from the N2D machine types configured with 128 and 224 vCPUs, which offer up to 70% higher platform memory bandwidth than comparable N1 instances. This, combined with higher core counts, provides over a 100% performance improvement on a variety of representative benchmarks, including Gromacs and NAMD, compared to n1-standard-96 vCPUs.
Forrest Norrod, the senior vice president and general manager at AMD's Datacenter and Embedded Solutions Business, teased future collaborations between AMD and Google Cloud in the company's announcement of these new N2D VMs. He said:
"AMD and Google have worked together closely on these initial VMs to help ensure Google Cloud customers have a high-performance and cost-effective experience across a variety of workloads, and we will continue to work together to provide that experience this year and beyond.”
The message is clear: AMD isn't just hoping to lure consumers away from Intel. It's going for data center customers, too, and powering the latest VMs from Google Cloud is a pretty damn good way to demonstrate how competitive its chips are.
Google Cloud said the N2D VMs are now available to customers in the us-central1, asia-southeast1 and europe-west4 regions as a public beta. The company said there are "more regions on the way!" but didn't offer additional details about its plans.
Stay on the Cutting Edge
Join the experts who read Tom's Hardware for the inside track on enthusiast PC tech news — and have for over 25 years. We'll send breaking news and in-depth reviews of CPUs, GPUs, AI, maker hardware and more straight to your inbox.
Nathaniel Mott is a freelance news and features writer for Tom's Hardware US, covering breaking news, security, and the silliest aspects of the tech industry.
Considering you can get up to 224 vCPUs they have to be 64 core variants. Then from Google's documentation you see "N2D machine types run on AMD EPYC Rome processors with a base frequency of 2.25 GHz, an all-core turbo frequency of 2.7 GHz, and a single-core turbo frequency of 3.3 GHz. " The only CPU that meets those requirements with 64 cores are the 7742s.Reply
Customers as big as Google can get customized versions of AMD's product stack, like they did with the Vega GPUs they're using in Stadia. Intel also offers custom variants for big customers.jeremyj_83 said:Considering you can get up to 224 vCPUs they have to be 64 core variants. ... The only CPU that meets those requirements with 64 cores are the 7742s.
To get 224 vCPUs, you only need 2x56 cores x2 threads. So, maybe they're using 7-chiplet variants or the slowest core on each of the 8 chiplets has been disabled. Or, they could be reserved for some kind of management functions.