How does the GPU billing works?
Here is my situation. I want to use a GPU to train a chatbot using machine learning. But I'm just going to train it a couple of times a day. The training takes about 5 minutes on the other server I have using only CPU. How am I going to get billed for that? Will I get charged only for the GPU power I use? Or the server only supports plans on a whole GPU just for me?
Also, the chatbot can use CUDA to process the messages it receives, if I decide to migrate my chatbot to Linode, will I get billed for the whole time the bot is running or just for the GPU processing power it uses?
In short, the GPU billing works based on the processing power you use, the time you are using it or I need to get a GPU dedicated just for me the whole time?
Our GPU plans are all Dedicated plans.
All Linode services, including the GPU, are billed hourly. You're billed for the time you have the service rather than the amount of resources you're using in your selected plan. Here's a description on how our billing works:
This means that you'll get billed for a service as long as it is present on our system, whether or not it's powered on:
Basically, if you create a GPU and keep it for any duration less than an entire month, you'll be billed for the number of hours that it is on our system. If you keep it for the entire month, you'll be billed at the monthly rate, which is capped at the amount listed on our pricing page.
Below, I've put in a link that will take you directly to the page that discusses GPUs and their pricing: