I've just ordered the most expensive PC I've built so far, centered around an RTX 5090. My main use will be heavy AI workloads and model inference, where the card's 32GB of VRAM and compute performance are important to me. I'll also use it for occasional 4K gaming, but gaming performance isn't the primary goal.
The parts are:
- MSI MAG X870E TOMAHAWK WIFI
- AMD Ryzen 9 7900X
- Corsair HX1500i ATX 3.1 PCIe 5.1 1500W 80 Plus Platinum modular PSU
- 64GB DDR5-6000 CL30 Kingston Fury Beast RGB EXPO, 2x32GB
- Noctua NH-D15 chromax.black
- Fractal Design North XL
- Kioxia Exceria Plus G4 2TB PCIe 5.0 NVMe SSD
- ASUS ROG Astral RTX 5090 OC
The 1500W power supply was cheaper than the 1200W model when I bought it. I'm still choosing a monitor and am currently considering the ASUS ROG Strix XG27UCG because I want a 4K display with reasonably accurate colors.
Does anything look poorly balanced or unsuitable for this workload? I'm especially interested in feedback on the CPU, memory capacity, cooling, storage, power supply, and monitor choice.
5 Answers
For the monitor, also compare newer 4K OLED or tandem-OLED models if you can accept their pricing and productivity tradeoffs. They can provide excellent gaming performance and image quality, but check text clarity, burn-in coverage, brightness behavior, and color calibration before buying. If color accuracy is more important than refresh rate, professional displays from manufacturers such as Eizo may be a better fit than a gaming-focused screen. You may also want more than 2TB of storage eventually, since AI models, datasets, and project files fill drives quickly.
The main part that stands out is the Ryzen 9 7900X. It should be perfectly usable for 4K gaming and many workstation tasks, but the rest of the system is extremely high-end. If your workloads scale well with CPU cores, a 16-core Ryzen 9 or another stronger productivity-oriented CPU would make the build more balanced. That said, if the higher-tier CPU would force you to compromise on the monitor, cooling, or other parts, keeping the 7900X is a reasonable budget decision.
That was my thinking. The 7900X covered my current needs while leaving money for the display and other components. I can upgrade the CPU later if my workloads justify it.
64GB is a sensible starting point for AI work, especially when system memory is used to support or offload models beyond the GPU’s VRAM. Whether it is enough depends heavily on the models and how much multitasking you do. Since the board has four memory slots, moving to 128GB later should be straightforward, although you should check the motherboard’s memory support list and stability at DDR5-6000.
I tested my usual workflow and 64GB seems fine for now, but I’m expecting that I may expand it if larger models or heavier multitasking become part of the routine.
The 1500W PSU is more than you strictly need, but it is not a problem if it was cheaper and has the proper modern GPU power connector. The important things are using the native cable, making sure it is fully seated, and avoiding sharp bends near the connector. A UPS is also worth considering if power interruptions could ruin long-running inference, renders, or training jobs.
The NH-D15 is a good, reliable air cooler, though it is expensive compared with some newer alternatives. It should handle the 7900X if the case has good airflow, and avoiding liquid cooling removes pump and leak concerns. Just verify RAM and case clearance, since the front fan can interfere with taller memory modules. The North XL gives you plenty of room for airflow and the large graphics card.
I’m tied to the Adobe ecosystem as well, so I’ll need to balance professional color behavior with gaming features. I’ll compare the newer OLED options and plan to add more storage if the initial 2TB becomes restrictive.