I am planning to deploy llama 3.3 70b(FP8) Model in my EC2 instance, and I am wondering which would be good for better performance, GPU memory utilization, and operational complexity?

I will be just deploying this single model in my ec2 so can you suggest which I should go with?