I'm planning an upgrade for a system I mainly use for classic machine learning, computer vision, neural networks, and occasional gaming. I don't need to run local language models, and I'm currently using a 1080p monitor with no immediate plans to upgrade it.
My current build is:
- Ryzen 5 1600 with the stock cooler
- MSI B450M PRO-M2 MAX motherboard
- Radeon RX 580 8GB
- 16GB DDR4-2400, plus a spare 2x4GB DDR4-3000 CL16 kit
- SilentiumPC Bronze power supply, with the exact model and wattage still to be confirmed
- Existing storage is sufficient
My budget is around €1,150, and I'm interested in the best performance-to-price ratio. Used components are fine. Staying on AM4 seems attractive, but I'm open to changing platforms if the performance gain justifies the cost.
Would it make sense to install a drop-in CPU upgrade such as a Ryzen 7 5700X, add a modern Radeon card such as an RX 9070 XT, replace the power supply, and upgrade the memory? Or would a different combination be more sensible for machine-learning work?
3 Answers
A Ryzen 9 5900X or 5950X is worth checking on the used market if your projects benefit from additional CPU threads. They can be excellent productivity upgrades on AM4, but they make less sense if the GPU is definitely the bottleneck. For an occasional gamer using 1080p, a 16GB midrange Radeon card may offer better overall value than going straight to the most expensive GPU in the budget.
Replace the power supply before installing a high-end graphics card. The exact model and wattage matter, so check the label rather than relying only on the Bronze rating. Choose a reputable modern unit with enough headroom for the selected GPU; around 650–750W is a reasonable target for many powerful cards. Also, don’t mix the 2x8GB and 2x4GB kits automatically. Four mismatched sticks may run more slowly or be less stable, so a matched 2x16GB kit is the cleaner upgrade.
That makes sense. I already dual-boot Windows and Debian but usually use Windows with WSL because rebooting between systems is inconvenient. I’ll check the PSU model and replace it before upgrading the GPU.
Since your work is GPU-bound, prioritize the graphics card rather than spending heavily on an AM4 gaming-focused CPU. A modern Radeon card with 16GB of VRAM would be a huge improvement over the RX 580, especially for computer-vision workloads that can use more memory. A 5700X is a sensible, affordable drop-in upgrade, but you probably don’t need to pay a premium for an X3D chip unless gaming becomes a bigger priority. Update the motherboard BIOS before changing the processor, and add enough memory to reach 32GB if your software regularly uses large datasets.
The 5500X3D may be difficult to find depending on the country, while the 5700X is much more widely available and has more cores and threads. I’d choose the 5700X if the prices are close.
I’ll compare used Ryzen 9 prices, although I’m not sure I need that much CPU power. My main goal is improving comfort and productivity for studying and working with machine learning.