AI Performance per Dollar (Memory Bandwidth per $)
Memory bandwidth per launch MSRP. LLM token generation is largely bandwidth-bound, so GB/s per dollar is the best spec-level value proxy for local LLM work. Computed live from our database — every row's launch price links to its source.
| # | GPU | GB/s per $1000 | LLM tok/s | tok/s per $1k | VRAM | Bandwidth | TDP | Launch MSRP | Year | |
|---|---|---|---|---|---|---|---|---|---|---|
| 1 | Arc A580 | 2860.34 | — | — | 8 GB | 512 GB/s | 185 W | $179 src | 2023 | Check Price on Amazon |
| 2 | Arc B580 | 1831.33 | 41 tok/s | 164.7 | 12 GB | 456 GB/s | 190 W | $249 src | 2024 | Check Price on Amazon |
| 3 | Arc A750 | 1771.63 | — | — | 8 GB | 512 GB/s | 225 W | $289 src | 2022 | Check Price on Amazon |
| 4 | Arc B570 | 1735.16 | — | — | 10 GB | 380 GB/s | 150 W | $219 src | 2025 | Check Price on Amazon |
| 5 | Arc A770 16GB | 1604.58 | — | — | 16 GB | 560 GB/s | 225 W | $349 src | 2022 | Check Price on Amazon |
| 6 | GeForce RTX 5060 | 1498.33 | — | — | 8 GB | 448 GB/s | 145 W | $299 src | 2025 | Check Price on Amazon |
| 7 | GeForce RTX 5050 | 1285.14 | — | — | 8 GB | 320 GB/s | 130 W | $249 src | 2025 | Check Price on Amazon |
| 8 | Radeon RX 7800 XT | 1250.5 | — | — | 16 GB | 624 GB/s | 263 W | $499 src | 2023 | Check Price on Amazon |
| 9 | Radeon RX 7900 XT | 1232.67 | — | — | 20 GB | 800 GB/s | 315 W | $649 src | 2022 | Check Price on Amazon |
| 10 | GeForce RTX 5070 | 1224.04 | — | — | 12 GB | 672 GB/s | 250 W | $549 src | 2025 | Check Price on Amazon |
| 11 | GeForce RTX 5070 Ti | 1196.26 | — | — | 16 GB | 896 GB/s | 300 W | $749 src | 2025 | Check Price on Amazon |
| 12 | Radeon RX 9070 | 1165.76 | — | — | 16 GB | 640 GB/s | 220 W | $549 src | 2025 | Check Price on Amazon |
| 13 | GeForce RTX 3060 12GB | 1094.22 | 38 tok/s | 115.5 | 12 GB | 360 GB/s | 170 W | $329 src | 2021 | Check Price on Amazon |
| 14 | GeForce RTX 3080 10GB | 1087.27 | — | — | 10 GB | 760 GB/s | 320 W | $699 src | 2020 | Check Price on Amazon |
| 15 | Radeon RX 7600 | 1070.63 | — | — | 8 GB | 288 GB/s | 165 W | $269 src | 2023 | Check Price on Amazon |
| 16 | Radeon RX 9070 XT | 1068.45 | — | — | 16 GB | 640 GB/s | 304 W | $599 src | 2025 | Check Price on Amazon |
| 17 | GeForce RTX 5060 Ti 16GB | 1044.29 | — | — | 16 GB | 448 GB/s | 180 W | $429 src | 2025 | Check Price on Amazon |
| 18 | Radeon RX 7700 XT | 1031.03 | — | — | 12 GB | 432 GB/s | 245 W | $419 src | 2023 | Check Price on Amazon |
| 19 | GeForce RTX 5080 | 960.96 | 132 tok/s | 132.1 | 16 GB | 960 GB/s | 360 W | $999 src | 2025 | Check Price on Amazon |
| 20 | Radeon RX 7900 XTX | 960.96 | 76 tok/s | 76.1 | 24 GB | 960 GB/s | 355 W | $999 src | 2022 | Check Price on Amazon |
| 21 | GeForce RTX 4070 | 918.03 | — | — | 12 GB | 504 GB/s | 200 W | $549 src | 2023 | Check Price on Amazon |
| 22 | GeForce RTX 4060 | 909.7 | — | — | 8 GB | 272 GB/s | 115 W | $299 src | 2023 | Check Price on Amazon |
| 23 | GeForce RTX 3070 | 897.8 | — | — | 8 GB | 448 GB/s | 220 W | $499 src | 2020 | Check Price on Amazon |
| 24 | GeForce RTX 5090 | 896.45 | 215 tok/s | 107.6 | 32 GB | 1792 GB/s | 575 W | $1,999 src | 2025 | Check Price on Amazon |
| 25 | Radeon RX 7600 XT | 875.38 | — | — | 16 GB | 288 GB/s | 190 W | $329 src | 2024 | Check Price on Amazon |
| 26 | GeForce RTX 4070 SUPER | 841.4 | 92 tok/s | 153.6 | 12 GB | 504 GB/s | 220 W | $599 src | 2024 | Check Price on Amazon |
| 27 | GeForce RTX 4070 Ti SUPER | 841.05 | — | — | 16 GB | 672 GB/s | 285 W | $799 src | 2024 | Check Price on Amazon |
| 28 | Radeon RX 6800 XT | 788.91 | — | — | 16 GB | 512 GB/s | 300 W | $649 src | 2020 | Check Price on Amazon |
| 29 | GeForce RTX 3080 Ti | 760.63 | — | — | 12 GB | 912 GB/s | 350 W | $1,199 src | 2021 | Check Price on Amazon |
| 30 | GeForce RTX 4080 SUPER | 736.74 | 102 tok/s | 102.1 | 16 GB | 736 GB/s | 320 W | $999 src | 2024 | Check Price on Amazon |
| 31 | GeForce RTX 4060 Ti 8GB | 721.8 | — | — | 8 GB | 288 GB/s | 160 W | $399 src | 2023 | Check Price on Amazon |
| 32 | GeForce RTX 4090 | 630.39 | 125 tok/s | 78.2 | 24 GB | 1008 GB/s | 450 W | $1,599 src | 2022 | Check Price on Amazon |
| 33 | GeForce RTX 3090 | 624.42 | 85 tok/s | 56.7 | 24 GB | 936 GB/s | 350 W | $1,499 src | 2020 | Check Price on Amazon |
| 34 | GeForce RTX 2080 Ti | 616.62 | — | — | 11 GB | 616 GB/s | 250 W | $999 src | 2018 | Check Price on Amazon |
| 35 | GeForce RTX 4080 | 597.83 | — | — | 16 GB | 716.8 GB/s | 320 W | $1,199 src | 2022 | Check Price on Amazon |
| 36 | GeForce RTX 4060 Ti 16GB | 577.15 | — | — | 16 GB | 288 GB/s | 160 W | $499 src | 2023 | Check Price on Amazon |
| 37 | GeForce RTX 3090 Ti | 504.25 | — | — | 24 GB | 1008 GB/s | 450 W | $1,999 src | 2022 | Check Price on Amazon |
| 38 | NVIDIA RTX 5000 Ada | 256 | — | — | 32 GB | 576 GB/s | 250 W | $2,250 src | 2023 | Check Price on Amazon |
| 39 | AMD Radeon Pro W7800 | 230.49 | — | — | 32 GB | 576 GB/s | 260 W | $2,499 src | 2022 | Check Price on Amazon |
| 40 | AMD Radeon Pro W7900 | 216.05 | — | — | 48 GB | 864 GB/s | 295 W | $3,999 src | 2022 | Check Price on Amazon |
| 41 | M3 Ultra (Mac Studio) | 204.8 | — | — | 512 GB | 819 GB/s | 270 W | $3,999 src | 2025 | Check Price on Amazon |
| 42 | M4 Max (MacBook Pro) | 170.68 | — | — | 128 GB | 546 GB/s | — | $3,199 src | 2024 | |
| 43 | RTX A6000 | 170.67 | — | — | 48 GB | 768 GB/s | 300 W | $4,500 src | 2021 | Check Price on Amazon |
| 44 | RTX 6000 Ada | 141.18 | 110 tok/s | 16.2 | 48 GB | 960 GB/s | 300 W | $6,800 src | 2023 | Check Price on Amazon |
Method: scores divide a database spec by launch MSRP (or TDP). Only GPUs with a source-linked launch price appear — 20 GPUs are excluded for lacking sourced pricing. These are launch MSRPs, not current street prices; used-market values differ. See methodology.
LLM columns: measured llama.cpp single-stream tokens/s for Llama-3.1-8B (Q4_K_M, llama.cpp b3500), joined from our sourced benchmark rows — not spec-derived estimates. EST marks estimated rows; "—" means no sourced result yet.
LLM columns: measured llama.cpp single-stream tokens/s for Llama-3.1-8B (Q4_K_M, llama.cpp b3500), joined from our sourced benchmark rows — not spec-derived estimates. EST marks estimated rows; "—" means no sourced result yet.
What to do with these rankings
High VRAM per dollar means more model capacity for the money — start here for LLMs. High bandwidth per dollar means faster token generation per dollar. High bandwidth per watt suits always-on servers and small-form-factor builds.
Then compare finalists head-to-head in the GPU comparison tool, check which models fit each card, or browse all sourced benchmark results.