And yet Meta is using MI300X exclusively for all live inference on Llama 405B. C...

		dhruvdh 4 hours ago \| parent \| context \| favorite \| on: AMD Instinct MI325X to Feature 256GB HBM3E Memory,... And yet Meta is using MI300X exclusively for all live inference on Llama 405B. Clearly there are workloads AMD wins at, and just going Nvidia by default for everything without considering AMD is suboptimal.