Infrastructure

NVIDIA Tesla P100 GPU: Be prepared for the AI revolution

Share this post:

NVIDIA MapD BitfusionWe know you’ve been wanting faster and easier access to GPU computing for AI in the cloud, so today we are excited to be the first major cloud provider to globally offer the NVIDIA Tesla P100 GPU card currently being revved up in our data centers and will be available in the forthcoming weeks.

NVIDIA and IBM have been partnering since 2014 to bring you the latest GPU technology in the cloud including being first to market in 2015 with the NVIDIA Tesla K80 and the Tesla M60 in 2016.

And now our customers can experience even more power to take on their AI and deep learning workloads with the P100s.

The landscape is changing and IBM Cloud and NVIDIA are at the forefront of the revolution.

Where supercomputing was once only something afforded by large corporations, by adding NVIDIA GPUs into the cloud, we are making it easily accessible to all.

We’re seeing deep learning and AI transition from traditionally research oriented computation to workloads dealing with infinite computing needs. The advantages of running high performance computing (HPC) in the cloud with NVIDIA GPUs, span industries and today offers financial services, healthcare, and scientific research the ability to perform better and calculate and analyze data faster. In fact, GPUs are now crossing over into business situations to attack business oriented problems, which essentially helps customers of any size or type answer their most complex big data challenges.

Recently, IBM Cloud, together with MapD and Bitfusion, were able to scale up to 64 Tesla K80 GPUs across 32 servers to filter, query and aggregate a 40-billion-row data set in just 271 milliseconds. That’s a mind-blowing 147 billion rows per SECOND. Imagine the possibilities with the new P100.

Watch how Bitfusion incorporated its software developed to manage deep learning and GPUs to help our customer MapD add GPUs to their cloud environment to accelerate their data analytics.

When you provision your bare metal server with two NVIDIA Tesla P100 GPU cards, you can see 50 times the performance than its predecessor, the Tesla K80. The addition of the accelerator can deliver up to 65 percent more machine learning capabilities giving you higher throughput than traditional virtualized servers.

For those ready to start trying out GPUs in a cloud environment, we currently offer NVIDIA Tesla M60 and K80, designed for high performance acceleration of scientific computation and data analytics. The Tesla M60 (with NVIDIA GRID software) along with the GRID K2, are engineered for professional grade virtualized graphics—all available on various pre-configured bare metal servers with hourly and monthly options.

Check out the configurations and the infographic.

A version of this article originally appeared on the IBM Bluemix blog.

More Infrastructure stories

Solving the productivity crisis: Digital process automation for deep deployments

The way we do work today isn’t working for many employees and employers. Employers have fewer qualified people to perform complex tasks, while employees get bogged down with low-value tasks. What slowly emerges is a productivity crisis. How can companies solve it? Automation is one of the go-to solutions, but automation fixes are still segmented, focusing on […]

Continue reading

How to start scaling automation projects with design thinking

Digital business automation is helping companies across industries improve operational efficiency, drive innovation and better serve customers. Automating work helps businesses offer a better experience at speed and scale. To begin, businesses must evaluate their big picture. What are the complex processes with multiple steps that when done manually are slow, costly and frustrating for […]

Continue reading

5 keys to overcoming cloud migration challenges

Cloud migration challenges continue to bedevil enterprises, despite the fact that the cloud itself has been around for nearly 20 years. Many enterprises have yet to realize the full promise of an abstract, distributed, federated data environment because migrations are still often so difficult. Nobody embarks upon a cloud migration expecting to fail, of course. […]

Continue reading