I'm putting together a machine learning rig and I'd like some feedback on my build. I'm aiming to run two GeForce RTX 5090 graphics cards, but I'm wondering if that's realistic. For the motherboard, do I really need the X870E model, or would the X670E suffice? I'm looking for a balance between cost and performance, so if you have suggestions for alternatives or see any potential bottlenecks, I'd love to hear them! Here's my current build list:
- **CPU:** AMD Ryzen 9 7950X 16-Core
- **CPU Cooler:** Thermalright Aqua Elite ARGB V6
- **Motherboard:** Gigabyte X870E AORUS ELITE WIFI7
- **Memory:** Corsair Vengeance 96 GB DDR5 6000
- **Storage:** 4 TB each from Crucial and KingSpec
- **Graphics Card:** Two GeForce RTX 5090s (though currently out of stock)
- **Case:** Lian Li O11 Dynamic EVO XL
- **Power Supply:** ASRock Phantom Gaming 1600W
Total cost is around $5962.66. Looking forward to your thoughts!
3 Answers
Overall, this build is a strong contender for professional-grade work! That CPU is great, but have you thought about upgrading to the 9950X? It gives a solid performance boost without breaking the bank. And those coolers are great too! Just a little tip: the O11 case doesn’t come with enough fans, so consider adding a few more to keep airflow optimal. Oh, and why the cheaper SSDs? You might want to invest a bit more there for speed.
Your build looks pretty solid, especially for machine learning! Just a heads up, though, you won't be able to run two 5090s at full x16 speeds since the AM5 platform has limitations there. Each card will only get x8 lanes. It might impact performance depending on your specific ML tasks, so you might want to consider that. I’d suggest looking for a motherboard that can split the lanes effectively—something like the Asus ProArt X670E would be a great fit. Keep in mind the PCIe bandwidth for your training tasks!
Thanks for the info! I wasn't aware that the lane split would affect performance that much. I'll definitely check out the ProArt X670E—it seems to be a good recommendation.
The choice between X670E and X870E really isn't too critical unless you need USB 4.0. What you really need to look for is a motherboard that can support those PCIe lanes effectively. The Asus X670E ProArt is a solid pick if you’re prioritizing PCIe bandwidth. Just keep in mind, for machine learning, the actual performance impact can vary, so investing in a capable board is wise.
Great to know! I’ll definitely lean toward the ProArt since bandwidth is key for ML. Thanks for the recommendation!
The CPU was a bit of an afterthought, but now that you mention it, the 9950X does sound appealing! And yeah, I definitely need to add more fans. As for SSDs, I think I was trying to balance cost; maybe I'll reconsider and get one premium SSD for speed.