Google Cloud Compute Engine Now Harnesses NVIDIA Tesla P100 GPUs With Up To 10x Performance Uplift
Google has been promising to upgrade its cloud platform with NVIDIA Tesla P100 GPUs, and now it has gone and done it. The upgraded hardware is available to customers in beta form, while NVIDIA's K80 GPUs are now generally available to customers. Google also announced the introduction of sustained use discounts (automatic discounts for running a VM instance for a significant portion of the month) on both the K80 and P100 GPUs.
NVIDIA's P100 GPUs are based on the company's Pascal architecture, the same as found in GeForce GTX 10 series of consumer graphics cards. Google's pitch is that the switch allows customers to increase throughput with fewer instances while saving money. And compared to the K80, Google touts a tenfold bump in workload acceleration on the burlier P100 GPUs.
"With today’s Cloud GPU announcements, GCP takes another step toward being the optimal place for any hardware-accelerated workload. With the addition of NVIDIA P100 GPUs, our primary focus is to help you bring new use cases to life," Google stated in a blog post.
Google's move to more powerful GPUs is a big deal for customers who work with machine learning tasks. It also provides a comparatively affordable way to leverage the power of advanced GPUs—customers can pay for this horsepower based on how long they actually use what Google's cloud platform provides, or receive a potentially significant discount (up to 30 percent) if using these GPUs for a sustained period of time.
The cloud GPUs are offered in passthrough mode to provide customers with bare-metal performance. Each VM can have up to four P100 or eight K80 GPUs, with up to 3TB of optionally attached local SSD storage.
This is not only a big deal for Google and its clients, but also NVIDIA, which is trying to gets its GPUs into as many cloud platforms as possible. Back in April, IBM said it would add P100 GPUs to its Bluemix development cloud for customers with demanding workloads, and then in May Microsoft announced it would use Pascal GPUs in its Azure cloud later this year.
NVIDIA's P100 GPUs are based on the company's Pascal architecture, the same as found in GeForce GTX 10 series of consumer graphics cards. Google's pitch is that the switch allows customers to increase throughput with fewer instances while saving money. And compared to the K80, Google touts a tenfold bump in workload acceleration on the burlier P100 GPUs.
"With today’s Cloud GPU announcements, GCP takes another step toward being the optimal place for any hardware-accelerated workload. With the addition of NVIDIA P100 GPUs, our primary focus is to help you bring new use cases to life," Google stated in a blog post.
Google's move to more powerful GPUs is a big deal for customers who work with machine learning tasks. It also provides a comparatively affordable way to leverage the power of advanced GPUs—customers can pay for this horsepower based on how long they actually use what Google's cloud platform provides, or receive a potentially significant discount (up to 30 percent) if using these GPUs for a sustained period of time.
The cloud GPUs are offered in passthrough mode to provide customers with bare-metal performance. Each VM can have up to four P100 or eight K80 GPUs, with up to 3TB of optionally attached local SSD storage.
This is not only a big deal for Google and its clients, but also NVIDIA, which is trying to gets its GPUs into as many cloud platforms as possible. Back in April, IBM said it would add P100 GPUs to its Bluemix development cloud for customers with demanding workloads, and then in May Microsoft announced it would use Pascal GPUs in its Azure cloud later this year.
