Leading AI Scalability Benchmarks with Microsoft Azure
-
Russ Fellows
Azure Demonstrates MLPerf Benchmark Strength
Businesses understand that AI has rapidly moved from being an interesting phenomenon to a tool that can be leveraged to provide firms with competitive advantages. Companies are now exploring options for developing and deploying AI applications both on premises and in public clouds. This trend is occurring across nearly all vertical segments and enterprise classes, with virtually all businesses affected by the drive to adopt AI implementations.
The ability to experiment, and rapidly grow or shrink infrastructure has always been one of the most compelling benefits of cloud computing. As companies experiment with AI training, fine-tuning and other resource intensive workloads, the benefits of cloud computing are clear. The ability to scale infrastructure to precisely match rapidly changing resource needs is vital.
Futurum Research found that the use of public clouds for AI workloads will grow at a 30% compound annual rate over the next five years. Additionally, Futurum Research found that Azure was the most commonly mentioned cloud vendor in it’s AI cloud research survey.
Our clients see value in the ability to quickly scale their AI investments i without requiring significant up front capital. As opportunities for growth present themselves, cloud AI infrastructure enable users; then afterwards reducing resources is easily achieved.
Microsoft Azure asked Signal65 to analyze their AI portfolio, along with the latest set of MLPerf Training benchmark results. The MLPerf benchmarks provide a way to effectively compare alternative offerings. While this test measures large-scale AI training, the results may be analyzed to understand how various infrastructure options could be utilized for workloads many enterprises are evaluating, including AI fine-tuning and inferencing workloads.
In analyzing Microsoft Azure AI’s capabilities, together with the MLPerf 4.1 training benchmark results, we found the following:
- Microsoft Azure’s MLPerf outperforms a leading cloud competitor by 28%: Using same number of GPUs to run Llama70B fine-tuning
- Microsoft Azure Delivers AI at Scale: Azure is one of three cloud vendors to demonstrate Llama 70B performance at scale, beyond 100 GPUs
- Azure AI has Industry Leading Options: Azure AI optimized software stacks, together with leading hardware options provide the highest performing, cloud computing AI workload options (based upon MLPerf DC Training Results).
- Azure AI has Leading Price / Performance: As a top public cloud provider, Azure provides highly competitive price / performance options for AI workloads.
Research commissioned by:
