AI Drives Adoption of Accelerated Computing Architectures Sponsored Content by Dell Technologies
With the global artificial intelligence (AI) software market growing by 47% in 2022, thousands of IT environments are adopting accelerated computing infrastructure to handle the demands of not only AI, but also machine learning (ML) and high performance computing (HPC) workloads. They’re gravitating to a variety of server accelerators to provide faster parallel compute performance for the world’s most demanding applications.
The Benefits of Real-time Responsiveness
Just as consumers expect near instantaneous responses from virtual assistants like Siri and Alexa that mimic the speed of human thought, businesses expect ultra-high performance from systems that detect credit card fraud, provide network services, recommend products and services, and deliver products to your door. The growing adoption of advanced, data-intensive technologies is greatly increasing the strain on server CPUs.
Enter Dell PowerEdge Servers with GPUs for accelerated parallel computing. These electronic circuits are designed for parallel processing, working with CPUs to accelerate applications. Due to their architecture, GPUs can also process large amounts of data simultaneously, however, the applications must be written to take advantage of the acceleration. For example, if all the software code knows is to drive one a one lane road, it won’t use all the other available lanes. The good news is the coding is easier than ever, opening applications up to new (faster) capabilities.
Customer Success Stories
From financial transactions to video editing, medical imaging, fluid simulations, and enterprise applications, companies are using accelerated computing, AI, and ML to do a wide variety of jobs. Here are some real-world results from the use of accelerated computing around the world.
- Making content recommendations 30 billion times a day across 4 billion Web pages and processing 150,000 requests per second
- Inspecting 124 moving railcars and transmit the data analysis in eight minutes, the time it would take to inspect one railcar manually
- Processing 300 images in one body scan and animate healthcare processes like lung ventilation and oxygen and blood flow in 5 minutes from patient entry to exit
- Training an AI model for radiology 187 times faster than using a non-accelerated computing environment and reduce training time for a neural translation model from months to hours
- Supporting a neural modeling AI application that maps 500 hours of MRI brain activity per subject and 20 TB of data used for benchmarks
- Enabling a hyperconverged infrastructure for virtual desktop infrastructure for tens of thousands of students and faculty members at a university system
- Powering sustainable, commercial-scale indoor farming using millions of data points from sensors and computer vision systems for 12-day growing cycles and 390% higher crop yields.
Hardware Speed Factors
Today’s workloads demand technology to flawlessly drive workload operations. For example, Dell PowerEdge servers are built to take advantage of the latest technological advances, including a wide range of GPUs to handle various types of workloads. Accelerated PowerEdge allows AI and ML applications to leverage a parallel processing environment that optimizes the intensive processing portions of applications on the data plane, while CPUs run control plane code.
Another consideration in accelerated computing hardware is heat. With state of the art Dell infrastructure designed for high throughput, thermal designs address heat-producing components. Other server features are front-to-back air-cooled designs and multi-vector cooling pathways.
A Growing Performance Differential
Comparing chips, GPUs currently beat CPUs 10 to one on bandwidth and floating point operations, according to computational physicist Vincent Natoli. While the gap is widening, GPUs still don’t function without CPUs ─ they need each other and work together.
According to the IEEE, “Accelerated architectures such as GPUs…have been proven to increase the performance of many algorithms compared to their CPU counterparts and are widely available in local, campus-wide and national infrastructures.” As for business benefits, according to analysis by Gartner, businesses with accelerated computing, integrated datasets, and AI solutions designed to inform decision-making will generate at least three times more value than companies that don’t use those solutions by 2025.
For more information on AI in business, read “Winning in the Age of AI – How organizations are accelerating intelligent outcomes everywhere.”