Did somebody try deploy three llama3-8b on four T4?

[liberifatali]

There are 4 T4 cards. But I see that the total GPU memory is < 60GB.