![Roofline model, based on [49]. The x-axis represents the operational or computational intensity (CI) and y-axis represents the attainable performance (AP) or throughput. Computational roof and I/O bandwidth roof limit the achievable AP. On the right (yellow area), the algorithms are compute-bound, while on the left (orange area), they are memory-bound](https://prod-files-secure.s3.us-west-2.amazonaws.com/db2f0ab2-59c6-4eb1-ba0f-e873fbe781f7/285c89ce-00b5-402f-b9e8-23bcd2abdc7a/Untitled.png)
Roofline model, based on [49]. The x-axis represents the operational or computational intensity (CI) and y-axis represents the attainable performance (AP) or throughput. Computational roof and I/O bandwidth roof limit the achievable AP. On the right (yellow area), the algorithms are compute-bound, while on the left (orange area), they are memory-bound
COMET: A Comprehensive Cluster Design Methodology for Distributed Deep Learning Training
[x] how one should work when trying to optimize its deployments, which GPU card should I use, A100, A10g, or 4xA10g.