Loading
Build your rig
Your custom configuration, tuned for your workflow.
Pick your primary workflow. We’ll recommend a complete system. Customize anything after.
Customize your build
Every component below is pre-selected based on your use case. Change only what you want.
Graphics
This determines which AI models you can run and how fast.
VRAM decides which models fit. A 70B model at Q4 needs ~40 GB. If it doesn't fit, it won't run (or runs painfully slow from system RAM).
Bandwidth decides how fast they run. Two cards with the same VRAM can differ 3× in speed based on memory bandwidth alone.
Not enough VRAM for this workflow
Code with AI locally needs at least 16 GB of VRAM. With 12 GB, models will need to offload layers to system RAM, which is significantly slower.
Models come in quality levels, like video resolution. The original (FP16) needs the most VRAM. Quantized versions trade a small amount of quality for much less VRAM.
Most people use Q4 or Q6. A 70B model fits on a single 24 GB card at Q4 instead of needing 80 GB+ at full precision.
Memory
16 GB (1 sticks)
System RAM (not VRAM) is your multitasking headroom. While the GPU runs your model, RAM handles everything else: editor, browser, OS. 32 GB minimum for AI work; 64 GB lets larger models spill over from GPU VRAM without grinding to a halt.
Not enough RAM for this workflow
Code with AI locally needs at least 32 GB of RAM. With 16 GB, your system will struggle to run a browser, editor, and AI model at the same time.
Storage
1.0 TB (1 drives)
AI models are large files. Here's what typical models need on disk:
1 TB minimum, 2 TB recommended. A few models, a code editor, and your OS fill 1 TB fast.
System
B650 + AMD Ryzen 5 7600
Platform
Your system platform. We'll pick the best one for your GPU.
B650
Gigabyte B650I
Processor
Matched to your system platform.
AMD Ryzen 5 7600
AM5 · 65W TDP
CPU Bottleneck Detected
Your selected use case (Code with AI locally) requires a more powerful processor (at least 8 cores). The AMD Ryzen 5 7600 has only 6 cores and will likely bottleneck your system, especially during data preparation or running concurrent models.
Thermal
Keeps your system cool and quiet under heavy AI workloads.
Power
Estimated draw: 385W. Recommended PSU: 600W+
GPU power spikes can exceed rated TDP by 30–50%. 40–50% headroom is the safe target.
Total
$1,145