Oracle is adding the NVIDIA Tesla P100 to its cloud server stable. The voracious appetite for cloud GPU power notches another conquest as Oracle adds NVIDIA’s fastest Pascal-based technology to its cloud and X7 server platform. Recently we covered that Google Cloud Platform Now has NVIDIA Tesla P100 GPU Nodes so one can see Oracle’s move as a competitive response for its on-prem customers.
NVIDIA Tesla P100 at Oracle
As part of Oracle’s US-East Ashburn region and soon the EU-Germany Frankurt region one can utilize two NVIDIA Tesla P100 GPUs per instance. That allows over 21 TFLOPS of single-precision performance per instance.
NVIDIA stated it is delivering the NVIDIA Tesla P100 via its X7 server platform that was just released. We recently covered the Oracle X7 server launch. These servers feature the latest generation Intel Xeon Scalable series processors.
Server manufacturers are increasingly seeing demand and potential for major revenue bumps with the AI/ machine learning trend. Each NVIDIA Tesla P100 GPU sells for more than a single socket server average selling price, making it a lucrative market. Selling these expensive GPUs in a cloud platform allows companies to attain higher utilization and better monetize expensive hardware.
NVIDIA Volta in Oracle’s Future
Beyond the NVIDIA Tesla P100 GPU announcement, there are a few clear hints at where Oracle is going with the technology. First, the company is using the P100 which has the next-generation drop-in upgrade of the NVIDIA Tesla V100. Second, the company said it is working on adding Volta support:
“Oracle is also working closely with NVIDIA to provide the next generation of GPUs based on the Volta Architecture in both bare metal and virtual machine compute instances soon, allowing for up to 8 NVIDIA GPUs, all inter-connected with NVIDIA NVLink.”
As one can see, Oracle is now moving full-swing into the AI and deep learning cloud instance services with the new NVIDIA Tesla P100 and soon V100 GPUs. While this NVIDIA Tesla P100 announcement is cool, the 8x V100 is going to be amazing with its Tensor compute units.