Want to try AMD’s Milan, Intel’s Ice Lake-SP, and Nvidia’s A100? Here's your chance

(Image credit: Oracle)

When it comes to cloud computing instances, there are currently three leaders — Amazon, Google, and Microsoft — that offer a broad range of instances targeting hundreds of workloads. 

Oracle, a leader in business software, has to work hard to catch up with its rivals, and its Oracle Cloud Infrastructure (OCI) announcing plans to offer instances powered by the latest and the upcoming hardware from market leaders. 

OCI says it now intends to offer instances powered by AMD’s yet-to-be-announced EPYC 7003-series ‘Milan’ processors with up to 64 cores, Intel’s approaching Ice Lake-SP CPUs featuring up to 28 cores, and Ampere’s upcoming Altra chips. Furthermore, next week OCI will make bare metal GPU instances powered by Nvidia’s Ampere GPU generally available. 

Latest hardware for demanding workloads

(Image credit: Oracle)

Oracle Cloud Infrastructure will use AMD’s EPYC ‘Milan’ processors for E4 general-purpose computing instances designed for general-purpose workloads that benefit from high performance cores and/or high core counts on AMD EPYC processors.

(Image credit: Oracle)

By contrast, Intel’s Xeon ‘Ice Lake-SP’ parts will power high-performance compute-focused X9 instances that need over 20,000 cores as well as low latency.

(Image credit: Oracle)

Ampere’s Altra-based instances will be offered with up to 160 cores per instance for those who need maximum multithreaded performance, but not top-notch single-threaded performance. All of these instances will be available next year. 

Oracle does not disclose too many details about its future AMD and Intel instances featuring unannounced CPUs, but it reveals pretty much everything about its Nvidia A100-powered offering.

(Image credit: Oracle)

OCI’s new GPU servers are based on AMD’s existing EPYC CPUs as well as up to eight Nvidia A100 processors. Unlike Nvidia’s new A100 DGX100 boxes, Oracle’s machines are equipped with 2 TB of DDR4 memory. The bare metal instances will be generally available worldwide starting from September 30 at a price of $3.05 per GPU hour, which is the lowest in the industry, according to Oracle. 

One of the reasons why Oracle offers its A100-powered bare metal instances at a low price is because it wants to attract clients who need to run software that runs on bare metal. Customers who use GPUs for AI and HPC applications already use instances from Amazon, Google or Microsoft and stealing these clients from their current providers will take time. By contrast, giving access to the latest hardware to those with bare metal applications ahead of everyone else gives OCI an immediate edge over rivals.

Sources: OracleAnandTech

Anton Shilov is the News Editor at AnandTech, Inc. For more than four years, he has been writing for magazines and websites such as AnandTech, TechRadar, Tom's Guide, Kit Guru, EE Times, Tech & Learning, EE Times Asia, Design & Reuse.