Server build for local inference. 128 gb 3200 or 256 gb 2133mhz RAM?
A user is seeking advice on building a server for local inference, specifically questioning the optimal RAM configuration for their dual RTX 3090 setup. They are debating between 128 GB of 3200 MHz RAM or 256 GB of 2133 MHz RAM, considering the cost and potential benefits for large models like Qwen 3.5 397B. AI