Is 64GB of DDR4 a better long-term choice than 32GB of DDR5?

For most everyday software and gaming, DDR5’s theoretical bandwidth advantage does not translate into a dramatic improvement. Capacity is more important once a workload exceeds the available memory, so 64GB DDR4 can be much better than 32GB DDR5 for virtual machines, development environments, data-heavy tools, and local ML work. That said, the bigger difference is the platform rather than the memory itself. The Ryzen 9 5900XT and AM4 are effectively at the end of their upgrade path, while a DDR5 platform gives you newer CPUs, newer motherboard features, and more room to upgrade later. If 32GB is enough for your current workloads, I’d choose DDR5; if you already know you need more than 32GB, 64GB DDR4 is the more practical choice.

For ML and homelab use, I would prioritize capacity. Running out of RAM can cause severe slowdowns or prevent a model, virtual machine, or container stack from running at all. A modest memory-speed difference is usually far less noticeable than swapping to storage because the system needs more memory. For gaming, the graphics card will usually matter much more than DDR4 versus DDR5. A well-configured DDR4 kit—around 3200–3600 MT/s with reasonable timings—will remain perfectly usable for years. Just don’t describe it as truly future-proof: future processors will increasingly require DDR5, so you’re trading upgrade flexibility for more memory today.

I’d lean toward the DDR5 build if you can make 32GB work initially. The newer processor and platform are likely to provide more useful performance for development and CPU-based workloads than the memory bandwidth alone suggests, and you can add a larger DDR5 kit later if needed. If your ML or lab work already needs 64GB, though, don’t sacrifice capacity just to use the newer standard. In that situation, 64GB DDR4 is the better system for your actual workload, even if the platform has less upgrade headroom.