I'm building my first PC and would appreciate feedback before spending a significant amount of money. The current parts list is: AMD 9950X3D, GeForce RTX 5080 Windforce OC, 64GB Corsair DDR5-6000 CL40 memory, MSI X870 Tomahawk WiFi motherboard, Arctic Freezer III Pro or Peerless Assassin cooler, Lian Li Lancool 207 case, an MSI MAG 1000W or 850W power supply, and a 2TB TLC WD SSD. The PC costs about $3,500, and I'm also planning to buy a MacBook Air, bringing the combined total to roughly $4,700.
I'm a computer science master's student. My idea is to use the MacBook Air as a portable computer for campus and connect remotely to the desktop when I need more processing power. I'll mainly use the desktop for software engineering, AI training and inference, and gaming at 1440p.
I originally planned to spend around $2,000, but several upgrades gradually pushed the build to $3,500. This is a major investment for me, so I'd like to know whether the parts are well balanced, whether I'm overspending, and whether the system should remain useful for at least five years.
3 Answers
The memory kit is usable, but DDR5-6000 with CL30 would generally be a better target for gaming than CL40, especially for improving frame-time consistency. That said, on an X3D processor the real-world difference may be fairly small. If you already bought the CL40 kit at a reasonable price, it probably isn’t worth replacing it just to chase benchmark numbers. You could investigate whether it can run tighter timings safely, but stability testing is important.
The system should still be perfectly usable after five years and may last much longer, but nobody can guarantee how demanding future games and AI software will become. The CPU, motherboard, memory capacity, storage, and power supply give you a strong base. The GPU and its VRAM are more likely to determine when you need an upgrade, particularly for AI workloads.
Also, moving from a $2,000 budget to $3,500 is not a small series of upgrades—it is a 75% increase. Make a list of the workloads you actually need to run locally and cut back on parts that do not directly improve those tasks. Spending more can be justified, but only if you are using the extra performance rather than buying every upgrade available.
The biggest question is what you mean by “AI.” An RTX 5080 can be excellent for gaming, but its VRAM may become the limiting factor for local LLM work and larger training projects. Around 24GB of VRAM is much more comfortable, so cards such as a 3090, 4090, or 5090 may make more sense if local AI is a serious priority. Nvidia is also generally the safer choice because many AI tools have better CUDA support.
If AI is only occasional experimentation, you could keep the 5080 and use cloud services or subscriptions for larger models. If gaming is the main use, you may also be able to save money with a less expensive GPU. For maximum flexibility, a small secondary GPU and a virtualized setup could let you run separate Windows and Linux environments, although that adds complexity.
I agree that VRAM matters more than raw gaming performance for local AI. If the goal is mostly AI rather than gaming, I’d prioritize a 24GB-or-more Nvidia card and treat the 5080 as primarily a gaming GPU.
The build is probably fine for general development and 1440p gaming, but the AI requirements should be settled before buying. If the machine-learning work is mostly coursework, remote servers or cloud GPUs may be more practical than trying to build for every possible model locally.