By Peter Cohen
Businesses looking to hyperscalers for instantly scalable compute capacity for big data number crunching, supercomputing services on demand and more have new options that yield better performance and better efficiency, according to Amazon. AWS has announced new Elastic Compute Cloud (EC2) instances powered by and optimized for their silicon. The new instances yield significant price performance benefits for customers, according to vice president Amazon EC2 at AWS, David Brown.
“AWS’s continued innovation means customers are now getting brand new, game changing instances to run their most important workloads with significantly better price performance than anywhere else,” said Brown.
New C7g instances powered by AWS’s Arm-based Graviton3 processors yield 25 per ent better performance for compute-intensive workloads than the previous-generation chips powering C6g instances while using 60 per cent less energy, said AWS.
“AWS Graviton3 processors also deliver up to 2x higher floating point performance for scientific, machine learning and media encoding workloads, up to 2x faster performance for cryptographic workloads, and up to 3x better performance for machine learning workloads compared to previous generation AWS Graviton2 processors,” said AWS.
Trainium leads EC2 ML price performance
To boost EC2 Machine Learning (ML) performance, AWS has also deployed new Trn1 instances powered by its Trainium processors. The company touted the energy-efficiency of the new cores and their performance stacked against EC2 P4D instances running Nvidia GPUs.
Trn1 instances provide supercomputing-class performance on demand, said AWS. Customers can scale machine learning training on up to 10,000+ Trainium accelerators networked at petabit scale.
“Trn1 instances offer 800 Gbps EFA networking bandwidth (2x higher than the latest EC2 GPU-based instances) and integrate with Amazon FSx for Lustre high performance storage—enabling customers to launch Trn1 instances with EC2 UltraClusters capability,” said AWS.
AWS Nitro SSDs boost big data
AWS’s big data customers and others working with I/O-intensive workloads have a new option to turn to. Im4gn/Is4gen/I4i instances feature new AWS Nitro SSDs. AWS claimed full-stack performance optimization thanks to its controls both the SSD hardware and firmware.
“Im4gn/Is4gen/I4i instances are architected to maximize the storage performance of I/O-intensive workloads. Im4gn/Is4gen/I4i instances offer up to 30 TB of NVMe storage from AWS-designed AWS Nitro SSDs, delivering up to 60 per cent lower I/O latency and 75 per cent lower latency variability compared to previous generation I3 instances to maximize application performance,” said AWS.
The Graviton3 is the latest generation of Graviton Arm-based processors to emerge from Amazon-owned Annapurna Labs. Amazon acquired Annapurna in 2015.
