Google LLC today added a new instance series to its cloud platform that runs on seven-nanometer server processors from Advanced Micro Devices Inc.’s Epyc Rome line.

The instances, branded as the N2D series, are the most powerful general-purpose virtual machines Google has on offer and are promised to be more efficient than existing instances that use Intel Corp. silicon. 

The N2D series runs on unnamed, likely customized AMD Epyc Rome chips with a base frequency of 2.25 gigahertz. Google said that the clock speed can go up to a maximum of 2.7Ghz when all cores are running and 3.3Ghz in single-core mode. The exact number of cores available to a given instance depends on how many virtual central processing units a customer provisions: N2D virtual machines can be configured with as many as 224 vCPUs.

For comparison, Google’s N family of Intel-based general purpose instances offer up to 96 vCPUs. The memory cap in the N2D is significantly higher as well with a maximum per virtual machine of 896 gibibytes, an alternative memory size measure that equals 1.074 gigabytes.

The search giant said its new AMD compute options provide advantages in other areas as well. When running standard enterprise workloads such as web servers and databases that need a balance between performance and memory, N2D instances can be as much as 13% more cost-efficient than a comparable virtual machine from the Intel-based N family. Performance is up to 39% better when measured with the industry standard Coremark benchmark.

Google is also aiming the N2D series at high-performance workloads such as simulations and financial trading applications. Thanks partially to the higher maximum vCPU count, the instances can provide more than double the computing power than their Intel-powered counterparts. 

“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,” Google Cloud project manager Catherine Vallejo wrote in a blog post. “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.”

The N2D series is available in beta. It joins the AMD-powered E2 family of budget instances Google launched in December. 

Google’s decision to base a second line of general-purpose compute instances, and its most powerful one at that, on Epyc Rome processors is a major win for AMD. The chipmaker has set a goal of taking a 10% share of the server CPU market from Intel by the second quarter and the search giant’s CPU orders could it get closer to that goal. Google, one of the world’s largest data center operators, spent about $13 billion in 2019 on cloud infrastructure and offices. 

Image: Google