There is little question that the GPU age is upon us.
We see it everywhere, from game consoles to supercomputers and now the datacenter, GPUs are permeating more and more of the computing ecosystem.
By boasting order of magnitude performance improvements on key tasks and exhibiting massive cost of ownership advancements these once specialized chips are writing a new chapter in enterprise computing.
The early lines of that chapter have been written by pioneers like Nvidia and IBM Softlayer. Needless to say we are delighted to be a partner with both of them.
With IBM’s announcement today that they will be adding the Tesla M60 to their arsenal of cloud based GPUs IBM has extended their leadership in the space (their blog on the subject is here). These Tesla M60s will speed virtualized desktop applications and broaden access to this supercomputing-class compute power.
Behind this massive surge in GPU interest and deployments is a different approach to computing, one tailor made for the tasks that promise to define the innovation frontier over the coming decade: massive datasets, machine learning, autonomous vehicles, virtual reality and advanced analytics.
These applications have effectively broken the CPU compute model. Consider true “big data” in today’s world - billion plus row datasets. To query or parse these sets takes tens of minutes or even hours. If you didn’t ask the right question the first time you could be waiting days to get to the answer.
Figure 1: MapD vs. various CPU databases
What makes GPU computing and the emerging class of software that optimizes them for enterprise-grade tasks so impactful, is that it fundamentally changes the parameters of what is possible. Problems that organizations once considered too large are now within scope. Problems that were once considered too complex are can now be rendered visually. Problems that took too long to consider efficiently can now be understood instantaneously.
Take our client Verizon.
They are querying 3.2 billion rows of data with millisecond lag - performing tasks in use cases ranging from network operations to tracking the status of software updates on devices. These are order of magnitude performance enhancements versus CPUs and will define the types of experiences that enterprises will come to expect.
The IBM announcement will make those experiences available to far more enterprises through their innovative and economical approach to cloud computing.
This is the year of the GPU and we are delighted to be playing a role alongside innovators like Nvidia and IBM as they expand availability of these game-changing sources of compute.