Home > Servers > Specialty Servers > White Papers > Deploy and Finetune Llama2 70B Chat on PowerEdge XE9680 with AMD Instinct MI300X > Business challenge
Businesses face a significant hurdle when attempting to deploy large AI models such as the Llama2 70B model from Meta without the substantial memory capacity provided by the PowerEdge XE9680 AMD Instinct MI300X GPUs. The high memory demand of the Llama2 70B Chat Model necessitates either a distribution across multiple GPUs or machines, escalating both hardware and operational expenses. This distribution could also result in performance inefficiencies due to inter-GPU or inter-machine communication overhead. Moreover, the solution’s scalability could be compromised, restricting the ability to manage larger workloads or future growth. Thus, the absence of adequate memory capacity can considerably impede the effective deployment of the Llama2 70b model at scale.