NGINX is a high-performance HTTP web server that is widely used for cloud web applications. As a web server, the performance of NGINX depends on compute, memory, and network performance. This makes it a great test subject for the overall performance of our processors.

While benchmark testing in our labs, we configured NGINX to serve a static HTML file over HTTPS. The workload generator was configured to run under increasing load from multiple clients, with a strict SLA. In our tests, we found that NGINX performance on Ampere Altra Max was up to 3.2 times faster than equivalent Intel Xeon Icelake 8380 and 1.83 times faster than AMD EPYC Milan 7763 at a socket level.

We also tested NGINX on Ampere Altra instances on Azure, comparing Ampere Altra Dpsv5 instances with comparable x86 instances. We found that Ampere Altra instances are at least 30% better in price-performance compared to Intel instances and 22% compared to AMD instances.

In the past decade, we have seen the power consumption of cloud service providers explode, with Bloomberg estimating in 2020 that cloud computing consumed 1% of the world’s electricity and predicting that this would rise to 8% of the world’s energy consumption by the end of the decade.

Ampere microprocessors provide best-in-industry power efficiency per core, allowing cloud service providers to scale the compute resources they provide to customers, without increasing their energy consumption.

In our testing of NGINX, we found that Ampere Altra Max processors used up to 73% less power at comparable performance to x86 platforms. Up to 69% fewer Ampere Altra Max processors were required to deliver 1 million requests per second. For cloud application developers, choosing Ampere instances means reducing the amount of compute resources, and reducing the total power required to deliver similar performance. You can reduce the carbon footprint of your cloud applications, while maintaining high performance at a much lower cost.

Performance data on Azure Public Preview as of 7/1/2022.

See the detailed workload brief here.

Video encoding is a critical cloud technology used to compress videos to reduce network & storage requirements. To evaluate video encoding performance, we use libx264 which implements the H.264/MPEG-4 AVC standard, the most widely used today.

Ampere Altra Cloud Native Processors lead in performance for popular cloud native applications while running at lower power than legacy x86 processors. With industry leading core count, power efficient SIMD units, fixed CPU frequency giving consistent performance on all cores and an highly energy efficient design are why Ampere Altra Processors are ideal for running video encoding workloads.

To evaluate x264 performance on Azure VMs with Ampere Altra, we ran “vbench: a Benchmark for Video Transcoding in the Cloud” on 16 vCPU VMs on Microsoft Azure, encoding the 15 input videos on each vCPU and measure performance throughput. Ampere Altra CPU based VMs running in Microsoft Azure have an average price-performance advantage of 130% and 115% vs. legacy x86 processors.

Performance data on Azure Public Preview as of 7/1/2022.

See detailed workload brief here.

In-memory cache such as Redis are at the core of many cloud-native applications. Redis is an open source, in-memory, key-value data store that is typically used as a database or a cache. It uses an in-memory dataset, but data can be persisted through periodic writes or appends to the disk. Redis is very fast, and it can deliver high throughput at sub-millisecond latencies. It is a feature-rich, modern, mature caching solution for todays cloud. According to DB engines, it ranks highly in popularity among key value stores in the cloud.

Ampere Altra offers compelling price-performance in Azure cloud. We measured throughput and latency on Ampere Altra for Redis compared to legacy x86 VMs. We used the memtier_benchmark suite from Redis Labs as a load generator and used similar standard client VMs for all tests. A 16vCPU Ampere Altra VM offered 35% better price-performance than competing VMs.

Performance data on Azure Public Preview as of 7/1/2022.

See detailed workload brief here.

MLPerf™ Inference is a benchmark suite consisting of carefully selected AI architectures that represent the cutting edge of today’s artificial intelligence. Therefore, it allows for a comprehensive test of how well a given system performs on a variety of machine learning tasks, including natural language processing, computer vision, recommendation engines, and more. It is a consensus on the best benchmarking techniques, forged by experts in architecture, systems and machine learning.

Ampere Altra VMs offer great performance in assortment of AI workloads, including the ones in MLPerf Inference benchmark. ResNet-50 v1.5 is a popular neural network architecture primarily used in the field of computer vision. This model, trained to perform well on ImageNet class prediction task, is part of the MLPerf Inference suite.

Performance data on Azure Public Preview as of 7/1/2022.

See Detailed workload brief here.