Testing 70GB LLM on Ryzen AI Max+ Pro 395 – Vulkan vs CPU (HP G1A)

Description: In this video, I demonstrate how I managed to load and run a 70GB Deepseek R1 Llama 70B language model on my HP G1A laptop with the Ryzen AI Max+ Pro 395 CPU. Using the latest Vulkan 1.3.4.1 drivers and llama.cpp, I walk through my memory setup—assigning 120GB to system RAM and 8GB to GPU—and discuss the challenges and performance differences between Vulkan and pure CPU modes. Watch as I monitor GPU and system memory usage, test generation speeds, and share insights on how large models behave on current hardware. The results might surprise you! Key topics covered: Loading massive LLMs on consumer hardware Vulkan vs CPU performance comparison Model quantisation and system configuration Practical limitations and future outlook If you have questions or want to see more large model tests, leave a comment below and subscribe for future updates! Hardware: HP G1A Laptop Ryzen AI Max+ Pro 395 128GB RAM 8GB GPU Software: Vulkan 1.3.4.1 llama.cpp #LLM #RyzenAI #Vulkan #AIbenchmark #Llama70B #HPG1A #MachineLearning #AIhardware