Talking with Dave Salvator, Senior Product Marketing Manager at NVIDIA

In the world of artificial intelligence (AI), the use of GPUs for acceleration has seen a rapid rise in recent years. NVIDIA, known for its GPUs in the gaming space, has been at the forefront of this acceleration, revolutionising the data centre and AI infrastructure.

Meet Dave Salvatore, Senior Product Marketing Manager at NVIDIA

In a recent podcast episode of “From Here to AI,” DeszBlanchfield, the host, sat down with Dave Salvatore, Senior Product Marketing Manager at NVIDIA, to discuss the trends and challenges in the AI era. Dave’s role at NVIDIA involves working with account teams and customers to guide them towards the right solutions in the data centre and AI space. With a background in journalism and over a decade of experience in the industry, Dave brings a unique perspective to the conversation.

The Evolution of GPU Acceleration

Dave explains the evolution of GPU acceleration, starting from the early days of 2D and 3D graphics to the emergence of deep learning and AI. NVIDIA’s programmable GPU and CUDA programming environment have played a pivotal role in this evolution, enabling developers to create their own effects and accelerate high-performance computing applications.

The Big Bang of Deep Learning

About six years ago, the ImageNet competition at Stanford marked a significant turning point in the AI landscape. A new type of neural network called a deep neural network, specifically AlexNet, outperformed the competition, leading to a surge in GPU-accelerated submissions. This event sparked a wave of innovation and experimentation in the AI domain, with researchers pushing the envelope and leveraging GPUs for cutting-edge research.

The Impact of GPU-Equipped Servers

The use of GPU-equipped servers has revolutionised data centres, allowing for equivalent performance in fewer nodes. This has significant implications for floor space, rack space, power budget, total cost of acquisition, and total cost of ownership. Additionally, NVIDIA has collaborated with companies like Facebook to develop server design recommendations, such as the HGX platform, to accelerate time to market and provide more choices for IT decision makers.

Designing Infrastructure for AI

The transition to GPU-accelerated data centres has prompted a shift in infrastructure design and architecture. Companies like IBM have implemented NVLink in their Power9 series, removing barriers to data movement and enabling high bandwidth and I/O capabilities. This design focus on moving data quickly and efficiently has enhanced the performance and value of AI infrastructure.

The Future of AI Acceleration

Looking ahead, the trends in GPU acceleration for AI are pointing towards explosive growth in accelerated computing data centres. Use cases are evolving rapidly, encompassing not only training but also inference. Industries across the board are exploring ways to leverage AI for business advantage, with applications ranging from image-based networks to speech and natural language processing.

Optimising Data Centre Infrastructure

For organisations embarking on the AI journey, optimising data centre infrastructure is crucial. GPU-equipped servers offer numerous advantages, including space and energy savings, as well as improved performance. Implementing server design recommendations and leveraging technologies like NVLink can further enhance the efficiency and capabilities of AI infrastructure.

In conclusion, the landscape of AI acceleration is evolving at a rapid pace, with GPUs playing a central role in driving innovation and performance gains. As the demand for AI continues to grow across industries, the focus on optimising data centre infrastructure and leveraging cutting-edge technologies will be paramount in unlocking the full potential of AI.

With NVIDIA at the forefront of this transformation, the future of AI acceleration looks promising, with endless possibilities for groundbreaking research and real-world applications.