Durham, NC, October 10, 2023 – From search result suggestions to text chatbots, NLP powers many everyday applications. Deep learning frameworks for NLP, such as RoBERTa, require intensive compute power to analyze text and make predictions quickly and effectively. Principled Technologies tested the RoBERTa performance of Amazon EC2 M7i instances featuring 4th Gen Intel Xeon Scalable processors and M7g instances featuring AWS Graviton3 processors. For a fuller look at NLP performance an organization might expect from the instances, PT tested three instance sizes with two precision types and two batch sizes.
The report states, “Our RoBERTa test results indicate that AWS EC2 M7i instances enabled by 4th Gen Intel Xeon Scalable processors outperformed M7g instances with AWS Graviton3 processors at different vCPU counts, batch sizes, and precisions, processing up to 10.65 times as many sentences per second. These performance gains led to M7i instances delivering a better value, achieving up to 8.62 times the throughput per dollar. With an instance that can analyze text more quickly, your business could offer a smoother experience on RoBERTa-supported apps or potentially condense RoBERTa workloads onto fewer instances.”