A user inquired about the feasibility of clustering a Nvidia DGX Spark with an AMD Ryzen AI Max 395 to run a single large language model. Both devices possess 128GB of unified memory, offering a potential combined capacity of approximately 256GB minus operating system overhead. The DGX Spark is equipped with a 200Gbit network interface, whereas the AMD Strix system currently has only 5Gbit Ethernet but includes a PCIe Gen 4x4 slot. The user noted that DeepSeek v4 Flash can fit on two DGX Sparks and wondered if the Strix could serve as an alternative node. To improve connectivity, they proposed adding a Mellanox ConnectX-6 QSFP+28 to the AMD system to achieve higher bandwidth over the link.