Hey everyone! I'm looking to build a new PC specifically designed for running AI tasks for an app I've developed. I've set aside around $3,000 but can add more if needed. My current setup features an RTX 4060, an Intel i7 11th gen processor, and 32GB of DDR4 RAM, but I need to upgrade to at least 64GB of DDR5 RAM and a better CPU. The main focus is the graphics card since my app is Nvidia locked. I'm pretty new to building PCs and have mostly worked with prebuilt ones, so I'm not well-versed in things like heatsinks and component compatibility. Also, I prefer a clean look without any RGB lighting. Any build suggestions or tips would be greatly appreciated!
2 Answers
Just a thought—why run the AI locally? The highest you'll typically manage with local computing is around 70 billion parameters. While local running is valuable for learning, if you intend to use it effectively, consider setting up an API with services like Claude or Gemini.
Here's a solid $3k PC build suggestion that keeps the RGB to a minimum. You might want to consider this part list:
- **CPU:** AMD Ryzen 9 9900X - ~$374
- **CPU Cooler:** be quiet! Pure Rock Pro - ~$55
- **Motherboard:** Gigabyte B850 AORUS ELITE WIFI7 - ~$210
- **Memory:** Crucial Pro 64GB (2x32GB) DDR5-6000 - ~$680
- **Storage:** Samsung 990 EVO Plus 2TB SSD - ~$328
- **Graphics Card:** Zotac GAMING SOLID CORE OC GeForce RTX 5080 - ~$1,263
- **Case:** Phanteks XT PRO Mid Tower - ~$50
- **Power Supply:** be quiet! Pure Power 13 1000W - ~$140
This comes up to around $3,110, and everything's compatible with no RGB issues. It should serve your AI processing needs well!
Thanks for putting this together! Looks great and fits my requirements perfectly!
It's not intended for LLMs. The idea is that running locally means no ongoing costs like cloud services. Even the cheapest GPU can cost about $1 an hour just running 24/7, which adds up quick. I also want to rely on my own machine in case of outages.