Google now offers Tesla T4 CPUs from Nvidia on the Google Cloud Platform. The offer is still in the beta phase.
Google was the first cloud party to make Tesla T4 GPUs available on its platform in November, then still in alpha. The range is now being made more widely available.
The Tesla T4 is available in the regions Brazil, India, the Netherlands, Singapore, Tokyo and the USA. Thanks to this global coverage, Google is able to guarantee its customers worldwide a low latency.
Each T4 comes with 16GB of memory and performance up to 260 TOPs. According to Google, the T4 is therefore ideal for inferencing workloads, but can also be used for training machine learning models. The V100 is actually the better choice for the latter, but the T4 does it at a lower price point.
“Our T4 GPUs complement our V100 offer nicely. You can scale up with large VMs up to eight V100 GPUs, scale down with cheaper T4 GPUs, or scale up with T4 or V100 GPUs based on your workload characteristics,” said Chris Kleban, product manager Cloud GPUs at Google, in a blog post.
It is also the first CPU in the data centre with separate ‘RT Cores’, as part of Nvidia’s Turing architecture. These special cores enable real-time ray tracing, a technique for realistically simulating light effects. Google also supports virtual workstations on T4 instances so designers can run their rendering applications from anywhere.
Other Nvidia GPUs in the Google Cloud offering are the K80, P4, P100 and V100.our launch article.