IBM officials for more than a year have turned to Nvidia's GPU accelerators to boost the system in its cloud computing environment.
The massive tech vendor last year brought Nvidia's Tesla K80 and Tesla K10 GPUs into the IBM Cloud to make it easier and faster to run such complex high-performance computing (HPC) workloads as data analytics and deep learning. IBM officials this week said they now are offering the graphics-chips maker's Tesla M60 GPU in the cloud to enable users to run virtual desktop applications more quickly and more affordably.
Essentially, the performance capabilities enabled by using the Tesla M60 GPU accelerators will enable organizations to run their complex tasks much faster while deploying fewer servers than they have to now. Those workloads can include data analytics, graphics, energy exploration, deep learning and artificial intelligence (AI), according to company officials.
Through the addition of Nvidia GPUs onto the IBM Cloud, "we are one step closer to offering supercomputing performance on a pay-as-you-go basis, which makes this new approach to tackling big data problems accessible to customers of all sizes," Jerry Gutierrez, HPC leader for SoftLayer, IBM's cloud company, said in a statement. "We're at an inflection point in our industry, where GPU technology is opening the door for the next wave of breakthroughs across multiple industries."
Organizations in the HPC and supercomputing fields for almost a decade have increasingly turned to GPU accelerators from Nvidia and Advanced Micro Devices to speed up the performance of their systems while keeping power consumption and costs down. GPUs work with a system's CPUs, running complex workloads offloaded by the CPU that include tasks that can be processed simultaneously rather than sequentially, as on CPUs.
Now these organizations are seeing more options when it comes to acceleration technology. Intel offers x86-based acceleration through its many-core Xeon Phi co-processors and field-programmable gate arrays (FPGAs) from Intel (through its $16.7 billion acquisition of Altera) and others, including Qualcomm, via its partnership with FPGA maker Xilinx.
In the latest Top500 list of the world's fastest supercomputers released in November 2015, 104 of the systems used GPUs or co-processors. Sixty-three of these use Nvidia chips, while three use AMD's Radeon offerings. Twenty-seven use Intel's Xeon Phis, and four use a combination of Nvidia and Intel accelerators.
There are more than 400 applications—in such areas as engineering, science, deep learning and data analytics—that can be accelerated via GPUs, according to Nvidia.
Nvidia introduced the Tesla M60 GPU a year ago at the VMworld 2015 show, introducing a graphics technology that includes 16G of GDDRR5 memory and the ability to run up to 128 instances on a single server and up to 32 concurrent users. By offering the accelerator in the IBM Cloud, Big Blue officials said companies can now take advantage of the performance increases to cut the processing time of applications in such areas as CAD/CAM from days or weeks to hours, when compared with systems running on CPUs only.
"For the first time, businesses can deliver workstation-class graphics-intensive applications from the cloud along with high performance computing," Jim McHugh, vice president and general manager at Nvidia, said in a statement.