Google had announced three months ago that they will be launching its support for high-end GPU’s for machine learning and for other workloads. As per the commitment, yesterday Google officially announced that their GPUs will be available on the Google cloud platform for developers also. It also supports more than 8 GPUs attached to customized VMs, that allows developers to optimize the performance of the applications.
Now Google Cloud platform’s performance will increase by the addition of NVIDIA Tesla K80 GPUs. By using the gcloud command-line tool developers can now spin up NVIDIA GPU-based Virtual Machines that will be available in three GCP regions: us-east1, asia-east1, and europe-west. Creating the GPU Virtual machines using Cloud console will be available in the coming week.
With this new feature, Google wants to help developers who frequently spin up clusters of high-end machines to boost up their machine learning frameworks. This Google cloud GPU’s are combined with Google’s Machine Learning service and with its different storage platforms and databases.
Like other infrastructures, the price of GPUs are priced high and it is charged as per minute basis and a minimum is 10 minutes. Every K80 GPU with VM is costed $0.700 /hour /GPU and For Europe and Asia, it is costing $0.770/hour /GPU. It will be billed based on the usage. This helps in spinning large clusters of GPU machines freely for training deep learning and machine learning with zero cost.