A discrete GPU has fast but limited VRAM (up to 32 GB on consumer cards). Unified-memory machines share one large pool between CPU and GPU, so they can hold much bigger models, but their bandwidth is lower than high-end GPUs, so generation is slower per active parameter.
Unified-memory systems
| System | Chip | Memory | Memory bandwidth | Price | Source |
|---|---|---|---|---|---|
| Mac mini (M6, 32GB) | Apple M6 | 32 GB unified | 170 GB/s | TBD | Apple: Mac mini technical specifications |
| Mac mini (M5 Pro, 64GB) | Apple M5 Pro | 64 GB unified | 307 GB/s | TBD | Apple: Mac mini technical specifications |
| Mac Studio (M5 Max 40-core GPU, 128GB) | Apple M5 Max | 128 GB unified | 614 GB/s | TBD | Apple: Mac Studio technical specifications |
| GMKtec EVO-X2 (128GB / 1TB) | AMD Ryzen AI Max+ 395 | 128 GB unified | 256 GB/s (computed: 256-bit × 8000 MT/s) | $3,499.99 (maker store, accessed 21 Oct 2026) | GMKtec: EVO-X2 product page (128GB + 1TB variant); AMD: Ryzen AI Max+ 395 specifications |
| NVIDIA DGX Spark (128GB) | NVIDIA GB10 Grace Blackwell | 128 GB unified | 273 GB/s | $4,699 (NVIDIA Founders Edition MSRP from Feb 2026; current marketplace price TBD) A 64GB configuration is sold through OEM partners. | NVIDIA: DGX Spark product page and specs; NVIDIA Developer Forums: DGX Spark price change announcement (Feb 2026) |
Discrete GPUs for comparison
| GPU | VRAM | Memory bandwidth | Launch MSRP (US) | Source |
|---|---|---|---|---|
| GeForce RTX 5090 | 32 GB | 1792 GB/s | $1,999 | NVIDIA: GeForce RTX 5090 specs; NVIDIA Newsroom (Jan 2025): RTX 5090/5080/5070 Ti/5070 pricing; Wikipedia: GeForce RTX 50 series (memory bandwidth table) |
| GeForce RTX 4090 | 24 GB | 1008 GB/s | $1,599 Launch MSRP (2022). Current price TBD. | NVIDIA: GeForce RTX 4090 specs; NVIDIA Newsroom (Sep 2022): RTX 40 Series pricing; Wikipedia: GeForce RTX 40 series (memory bandwidth table) |
| GeForce RTX 3090 | 24 GB | 936 GB/s | $1,499 Launch MSRP (2020); sold used today. Used price TBD. | NVIDIA: GeForce RTX 3090 specs; NVIDIA Newsroom (Sep 2020): RTX 30 Series pricing; Wikipedia: GeForce RTX 30 series (memory bandwidth table) |
| GeForce RTX 5080 | 16 GB | 960 GB/s | $999 | NVIDIA: GeForce RTX 5080 specs; NVIDIA Newsroom (Jan 2025): RTX 5090/5080/5070 Ti/5070 pricing; Wikipedia: GeForce RTX 50 series (memory bandwidth table) |
VRAM and bus width from NVIDIA spec pages; bandwidth from Wikipedia series tables (not listed on NVIDIA pages); MSRPs from NVIDIA press releases. Current street prices: TBD.
How much of unified memory can the GPU use? Not all of it: the OS reserves some, and limits differ by platform and settings. Our tools assume about 75% usable on Mac and Ryzen AI Max (96 of 128 GB; Corsair's listing for its Ryzen AI Max+ 395 system says up to 96 GB VRAM) and 120 of 128 GB on DGX Spark. These are planning assumptions, not maker guarantees.