AMD Takes Lead in AI with Latest MLPerf Submission

AMD is proud to reveal the results of its inaugural submission of MLPerf with AMD Instinct™ MI300X GPUs here at the updated MLCommons site.

There are multiple MLPerf submission types, including the MLPerf Inference: Datacenter, MLPerf Training, and lastly the Llama 2 70B, which is a cutting-edge and large GenAI language model that was introduced in 2024. AMD chose to make its first MLPerf submission on the popular and relevant LLM model, Llama 2 70B. 

The AMD MLPerf Inference v4.1 submission has three entries for Llama 2 70B, which showcase: 

· The best combination of AMD CPU and GPU that are available on the market for AI tasks. 

· The large memory in MI300X (192GB) enables a single MI300X GPU to run the entire Llama 2 70B model, which on most competing GPUs needs to be split between multiple accelerators. 

· And finally, how AMD’s next generation CPU improves performance of AI tasks. 

In the MLPerf 4.1 Inference round, AMD results show that MI300X with ROCm can deliver superb inferencing performance for massive LLM like Llama 2 70B, even compared against submissions from other industry leaders like the Nvidia H100 with the Llama 2 70B LLM model. 

^{𝐒𝐭𝐚𝐲 𝐢𝐧𝐟𝐨𝐫𝐦𝐞𝐝 𝐰𝐢𝐭𝐡 𝐨𝐮𝐫 𝐥𝐚𝐭𝐞𝐬𝐭 𝐮𝐩𝐝𝐚𝐭𝐞𝐬 𝐛𝐲 𝐣𝐨𝐢𝐧𝐢𝐧𝐠 𝐭𝐡𝐞 WhatsApp Channel now! 👈📲}

^{𝑭𝒐𝒍𝒍𝒐𝒘 𝑶𝒖𝒓 𝑺𝒐𝒄𝒊𝒂𝒍 𝑴𝒆𝒅𝒊𝒂 𝑷𝒂𝒈𝒆𝐬} 👉 ^{Facebook, LinkedIn, Twitter, Instagram}