14 datasheets · prices in USD
Hardware datasheets
Capacity decides what fits; bandwidth decides how fast it runs. Every device with its memory, bandwidth, power and price, and the largest model it holds.
- Discrete graphics card
- Mac Studio
- Mini PC
- NVIDIA DGX Spark
- Laptop
- Home server
- Multi-GPU tower
All hardware
Cheapest first. Bandwidth per GB of memory is a quick read on how a device balances size against speed.
Device | Memory | Bandwidth | Load power | Price | Largest model, 8k |
|---|---|---|---|---|---|
GeForce RTX 5060 Ti 16 GB NVIDIA · Discrete GPU | 16 GBSourced | 448 GB/sSourced | 250 WSample | $779.99Sourced | gpt-oss-20b |
GeForce RTX 4060 Ti 16 GB NVIDIA · Discrete GPU | 16 GBSourced | 288 GB/sSourced | 230 WSample | $844.37Estimate | gpt-oss-20b |
Radeon RX 7900 XTX 24 GB AMD · Discrete GPU | 24 GBSourced | 960 GB/sSourced | 400 WSample | $899Sample | Qwen3 32B |
GeForce RTX 3090 24 GB (used) NVIDIA · Discrete GPU | 24 GBSourced | 936 GB/sSourced | 420 WSample | $1,200Sourced | Qwen3 32B |
CPU only · Ryzen 9 9950X + 128 GB DDR5 AMD · CPU reference | 128 GBSourced | 90 GB/sSourced | 200 WSample | $1,400Sample | gpt-oss-120b |
2× RTX 3090 (used) 48 GB NVIDIA · Multi-GPU | 48 GBSourced | 936 GB/sSourced | 780 WSample | $2,400Estimate | Llama 3.3 70B Instruct |
Mac Studio M5 Max (base) Apple · Apple unified | 36 GBSourced | 460 GB/sSourced | 150 WSample | $2,499Sourced | Qwen3 32B |
Ryzen AI Max+ 395 mini PC 128 GB AMD · Unified memory | 128 GBSourced | 256 GB/sSourced | 120 WSourced | $3,499.99Sourced | gpt-oss-120b |
GeForce RTX 4090 24 GB NVIDIA · Discrete GPU | 24 GBSourced | 1,008 GB/sSourced | 500 WSample | $4,395Sourced | Qwen3 32B |
NVIDIA DGX Spark FE 128 GB NVIDIA · Unified memory | 128 GBSourced | 273 GB/sSourced | 170 WSample | $4,699Sourced | gpt-oss-120b |
MacBook Pro M4 Max 40c · 128 GB Apple · Apple unified | 128 GBSourced | 546 GB/sSourced | 140 WSample | $4,699Sample | gpt-oss-120b |
Mac Studio M5 Ultra (base) Apple · Apple unified | 96 GBSourced | 1,200 GB/sSourced | 240 WSample | $5,499Sourced | gpt-oss-120b |
GeForce RTX 5090 32 GB NVIDIA · Discrete GPU | 32 GBSourced | 1,792 GB/sSourced | 650 WSample | $7,199.99Sourced | Qwen3 32B |
Mac Studio M5 Ultra · 512 GB Apple · Apple unified | 512 GBSourced | 1,200 GB/sSourced | 260 WSample | $9,999Sample | gpt-oss-120b |