NVIDIA GeForce RTX 5090
Full GPU32 GB VRAM
Estimated speed: ~623.3 tok/s
Bandwidth: 1792 GB/s
Loading GPU recommendations...
Running compatibility calculations across available GPUs.
Loading GPU recommendations...
Running compatibility calculations across available GPUs.
GPU Recommendations
Comparing 90 GPUs for running Phi-3 Mini 3.8B (Q4_K_M, 2.3 GB).
Ranking priority: compatibility verdict first, then estimated speed, then memory bandwidth. Buying links are optional shortcuts if you are ready to upgrade.
Copy a link to this GPU ranking so someone else can see the best upgrade paths for Phi-3 Mini 3.8B.
32 GB VRAM
Estimated speed: ~623.3 tok/s
Bandwidth: 1792 GB/s
256 GB VRAM
Estimated speed: ~379.8 tok/s
Bandwidth: 1092 GB/s
24 GB VRAM
Estimated speed: ~350.6 tok/s
Bandwidth: 1008 GB/s
24 GB VRAM
Estimated speed: ~350.6 tok/s
Bandwidth: 1008 GB/s
16 GB VRAM
Estimated speed: ~333.9 tok/s
Bandwidth: 960 GB/s
24 GB VRAM
Estimated speed: ~333.9 tok/s
Bandwidth: 960 GB/s
24 GB VRAM
Estimated speed: ~325.6 tok/s
Bandwidth: 936 GB/s
12 GB VRAM
Estimated speed: ~317.2 tok/s
Bandwidth: 912 GB/s
12 GB VRAM
Estimated speed: ~317.2 tok/s
Bandwidth: 912 GB/s
16 GB VRAM
Estimated speed: ~311.7 tok/s
Bandwidth: 896 GB/s
20 GB VRAM
Estimated speed: ~278.3 tok/s
Bandwidth: 800 GB/s
128 GB VRAM
Estimated speed: ~278.3 tok/s
Bandwidth: 800 GB/s
192 GB VRAM
Estimated speed: ~278.3 tok/s
Bandwidth: 800 GB/s
192 GB VRAM
Estimated speed: ~278.3 tok/s
Bandwidth: 800 GB/s
10 GB VRAM
Estimated speed: ~264.3 tok/s
Bandwidth: 760 GB/s
16 GB VRAM
Estimated speed: ~256 tok/s
Bandwidth: 736 GB/s
16 GB VRAM
Estimated speed: ~249.4 tok/s
Bandwidth: 717 GB/s
12 GB VRAM
Estimated speed: ~233.7 tok/s
Bandwidth: 672 GB/s
16 GB VRAM
Estimated speed: ~233.7 tok/s
Bandwidth: 672 GB/s
16 GB VRAM
Estimated speed: ~217 tok/s
Bandwidth: 624 GB/s
11 GB VRAM
Estimated speed: ~214.3 tok/s
Bandwidth: 616 GB/s
8 GB VRAM
Estimated speed: ~211.5 tok/s
Bandwidth: 608 GB/s
16 GB VRAM
Estimated speed: ~200.3 tok/s
Bandwidth: 576 GB/s
16 GB VRAM
Estimated speed: ~200.3 tok/s
Bandwidth: 576 GB/s
16 GB VRAM
Estimated speed: ~200.3 tok/s
Bandwidth: 576 GB/s
16 GB VRAM
Estimated speed: ~194.8 tok/s
Bandwidth: 560 GB/s
128 GB VRAM
Estimated speed: ~189.9 tok/s
Bandwidth: 546 GB/s
16 GB VRAM
Estimated speed: ~178.1 tok/s
Bandwidth: 512 GB/s
16 GB VRAM
Estimated speed: ~178.1 tok/s
Bandwidth: 512 GB/s
16 GB VRAM
Estimated speed: ~178.1 tok/s
Bandwidth: 512 GB/s
8 GB VRAM
Estimated speed: ~178.1 tok/s
Bandwidth: 512 GB/s
8 GB VRAM
Estimated speed: ~178.1 tok/s
Bandwidth: 512 GB/s
8 GB VRAM
Estimated speed: ~178.1 tok/s
Bandwidth: 512 GB/s
12 GB VRAM
Estimated speed: ~175.3 tok/s
Bandwidth: 504 GB/s
12 GB VRAM
Estimated speed: ~175.3 tok/s
Bandwidth: 504 GB/s
12 GB VRAM
Estimated speed: ~175.3 tok/s
Bandwidth: 504 GB/s
8 GB VRAM
Estimated speed: ~172.5 tok/s
Bandwidth: 496 GB/s
11 GB VRAM
Estimated speed: ~168.3 tok/s
Bandwidth: 484 GB/s
12 GB VRAM
Estimated speed: ~158.6 tok/s
Bandwidth: 456 GB/s
8 GB VRAM
Estimated speed: ~155.8 tok/s
Bandwidth: 448 GB/s
8 GB VRAM
Estimated speed: ~155.8 tok/s
Bandwidth: 448 GB/s
16 GB VRAM
Estimated speed: ~155.8 tok/s
Bandwidth: 448 GB/s
8 GB VRAM
Estimated speed: ~155.8 tok/s
Bandwidth: 448 GB/s
8 GB VRAM
Estimated speed: ~155.8 tok/s
Bandwidth: 448 GB/s
8 GB VRAM
Estimated speed: ~155.8 tok/s
Bandwidth: 448 GB/s
8 GB VRAM
Estimated speed: ~155.8 tok/s
Bandwidth: 448 GB/s
12 GB VRAM
Estimated speed: ~150.3 tok/s
Bandwidth: 432 GB/s
12 GB VRAM
Estimated speed: ~150.3 tok/s
Bandwidth: 432 GB/s
12 GB VRAM
Estimated speed: ~150.3 tok/s
Bandwidth: 432 GB/s
64 GB VRAM
Estimated speed: ~139.1 tok/s
Bandwidth: 400 GB/s
96 GB VRAM
Estimated speed: ~139.1 tok/s
Bandwidth: 400 GB/s
128 GB VRAM
Estimated speed: ~139.1 tok/s
Bandwidth: 400 GB/s
8 GB VRAM
Estimated speed: ~133.6 tok/s
Bandwidth: 384 GB/s
12 GB VRAM
Estimated speed: ~133.6 tok/s
Bandwidth: 384 GB/s
12 GB VRAM
Estimated speed: ~125.2 tok/s
Bandwidth: 360 GB/s
6 GB VRAM
Estimated speed: ~116.9 tok/s
Bandwidth: 336 GB/s
6 GB VRAM
Estimated speed: ~116.9 tok/s
Bandwidth: 336 GB/s
6 GB VRAM
Estimated speed: ~116.9 tok/s
Bandwidth: 336 GB/s
8 GB VRAM
Estimated speed: ~111.3 tok/s
Bandwidth: 320 GB/s
16 GB VRAM
Estimated speed: ~100.2 tok/s
Bandwidth: 288 GB/s
8 GB VRAM
Estimated speed: ~100.2 tok/s
Bandwidth: 288 GB/s
6 GB VRAM
Estimated speed: ~100.2 tok/s
Bandwidth: 288 GB/s
16 GB VRAM
Estimated speed: ~100.2 tok/s
Bandwidth: 288 GB/s
8 GB VRAM
Estimated speed: ~100.2 tok/s
Bandwidth: 288 GB/s
8 GB VRAM
Estimated speed: ~100.2 tok/s
Bandwidth: 288 GB/s
8 GB VRAM
Estimated speed: ~97.4 tok/s
Bandwidth: 280 GB/s
48 GB VRAM
Estimated speed: ~95 tok/s
Bandwidth: 273 GB/s
8 GB VRAM
Estimated speed: ~94.6 tok/s
Bandwidth: 272 GB/s
8 GB VRAM
Estimated speed: ~89 tok/s
Bandwidth: 256 GB/s
8 GB VRAM
Estimated speed: ~89 tok/s
Bandwidth: 256 GB/s
8 GB VRAM
Estimated speed: ~89 tok/s
Bandwidth: 256 GB/s
8 GB VRAM
Estimated speed: ~89 tok/s
Bandwidth: 256 GB/s
8 GB VRAM
Estimated speed: ~89 tok/s
Bandwidth: 256 GB/s
8 GB VRAM
Estimated speed: ~77.9 tok/s
Bandwidth: 224 GB/s
8 GB VRAM
Estimated speed: ~77.9 tok/s
Bandwidth: 224 GB/s
32 GB VRAM
Estimated speed: ~69.6 tok/s
Bandwidth: 200 GB/s
32 GB VRAM
Estimated speed: ~69.6 tok/s
Bandwidth: 200 GB/s
32 GB VRAM
Estimated speed: ~69.6 tok/s
Bandwidth: 200 GB/s
6 GB VRAM
Estimated speed: ~66.8 tok/s
Bandwidth: 192 GB/s
6 GB VRAM
Estimated speed: ~66.8 tok/s
Bandwidth: 192 GB/s
4 GB VRAM
Estimated speed: ~66.8 tok/s
Bandwidth: 192 GB/s
6 GB VRAM
Estimated speed: ~66.8 tok/s
Bandwidth: 192 GB/s
36 GB VRAM
Estimated speed: ~52.2 tok/s
Bandwidth: 150 GB/s
4 GB VRAM
Estimated speed: ~44.5 tok/s
Bandwidth: 128 GB/s
32 GB VRAM
Estimated speed: ~41.7 tok/s
Bandwidth: 120 GB/s
4 GB VRAM
Estimated speed: ~39 tok/s
Bandwidth: 112 GB/s
24 GB VRAM
Estimated speed: ~34.8 tok/s
Bandwidth: 100 GB/s
24 GB VRAM
Estimated speed: ~34.8 tok/s
Bandwidth: 100 GB/s
16 GB VRAM
Estimated speed: ~23.7 tok/s
Bandwidth: 68 GB/s
3 GB VRAM
Estimated speed: ~46.8 tok/s
Bandwidth: 192 GB/s
| GPU | VRAM | Verdict | Estimated tok/s | Bandwidth | Where to Buy |
|---|---|---|---|---|---|
| NVIDIA GeForce RTX 5090 | 32 GB | Full GPU | ~623.3 | 1792 GB/s | |
| Apple M4 Ultra | 256 GB | Full GPU | ~379.8 | 1092 GB/s | |
| NVIDIA GeForce RTX 4090 | 24 GB | Full GPU | ~350.6 | 1008 GB/s | |
| NVIDIA GeForce RTX 3090 Ti | 24 GB | Full GPU | ~350.6 | 1008 GB/s | |
| NVIDIA GeForce RTX 5080 | 16 GB | Full GPU | ~333.9 | 960 GB/s | |
| AMD Radeon RX 7900 XTX | 24 GB | Full GPU | ~333.9 | 960 GB/s | |
| NVIDIA GeForce RTX 3090 | 24 GB | Full GPU | ~325.6 | 936 GB/s | |
| NVIDIA GeForce RTX 3080 Ti | 12 GB | Full GPU | ~317.2 | 912 GB/s | |
| NVIDIA GeForce RTX 3080 12GB | 12 GB | Full GPU | ~317.2 | 912 GB/s | |
| NVIDIA GeForce RTX 5070 Ti | 16 GB | Full GPU | ~311.7 | 896 GB/s | |
| AMD Radeon RX 7900 XT | 20 GB | Full GPU | ~278.3 | 800 GB/s | |
| Apple M1 Ultra | 128 GB | Full GPU | ~278.3 | 800 GB/s | |
| Apple M2 Ultra | 192 GB | Full GPU | ~278.3 | 800 GB/s | |
| Apple M3 Ultra | 192 GB | Full GPU | ~278.3 | 800 GB/s | |
| NVIDIA GeForce RTX 3080 10GB | 10 GB | Full GPU | ~264.3 | 760 GB/s | |
| NVIDIA GeForce RTX 4080 Super | 16 GB | Full GPU | ~256 | 736 GB/s | |
| NVIDIA GeForce RTX 4080 | 16 GB | Full GPU | ~249.4 | 717 GB/s | |
| NVIDIA GeForce RTX 5070 | 12 GB | Full GPU | ~233.7 | 672 GB/s | |
| NVIDIA GeForce RTX 4070 Ti Super | 16 GB | Full GPU | ~233.7 | 672 GB/s | |
| AMD Radeon RX 7800 XT | 16 GB | Full GPU | ~217 | 624 GB/s | |
| NVIDIA GeForce RTX 2080 Ti | 11 GB | Full GPU | ~214.3 | 616 GB/s | |
| NVIDIA GeForce RTX 3070 Ti | 8 GB | Full GPU | ~211.5 | 608 GB/s | |
| NVIDIA GeForce RTX 4090 Laptop | 16 GB | Full GPU | ~200.3 | 576 GB/s | |
| AMD Radeon RX 7900 GRE | 16 GB | Full GPU | ~200.3 | 576 GB/s | |
| AMD Radeon RX 6950 XT | 16 GB | Full GPU | ~200.3 | 576 GB/s | |
| Intel Arc A770 16GB | 16 GB | Full GPU | ~194.8 | 560 GB/s | |
| Apple M4 Max | 128 GB | Full GPU | ~189.9 | 546 GB/s | |
| AMD Radeon RX 6900 XT | 16 GB | Full GPU | ~178.1 | 512 GB/s | |
| AMD Radeon RX 6800 XT | 16 GB | Full GPU | ~178.1 | 512 GB/s | |
| AMD Radeon RX 6800 | 16 GB | Full GPU | ~178.1 | 512 GB/s | |
| Intel Arc A770 8GB | 8 GB | Full GPU | ~178.1 | 512 GB/s | |
| Intel Arc A750 | 8 GB | Full GPU | ~178.1 | 512 GB/s | |
| Intel Arc A580 | 8 GB | Full GPU | ~178.1 | 512 GB/s | |
| NVIDIA GeForce RTX 4070 Ti | 12 GB | Full GPU | ~175.3 | 504 GB/s | |
| NVIDIA GeForce RTX 4070 Super | 12 GB | Full GPU | ~175.3 | 504 GB/s | |
| NVIDIA GeForce RTX 4070 | 12 GB | Full GPU | ~175.3 | 504 GB/s | |
| NVIDIA GeForce RTX 2080 Super | 8 GB | Full GPU | ~172.5 | 496 GB/s | |
| NVIDIA GeForce GTX 1080 Ti | 11 GB | Full GPU | ~168.3 | 484 GB/s | |
| Intel Arc B580 | 12 GB | Full GPU | ~158.6 | 456 GB/s | |
| NVIDIA GeForce RTX 3070 | 8 GB | Full GPU | ~155.8 | 448 GB/s | |
| NVIDIA GeForce RTX 3060 Ti | 8 GB | Full GPU | ~155.8 | 448 GB/s | |
| NVIDIA GeForce RTX 3080 Laptop | 16 GB | Full GPU | ~155.8 | 448 GB/s | |
| NVIDIA GeForce RTX 2080 | 8 GB | Full GPU | ~155.8 | 448 GB/s | |
| NVIDIA GeForce RTX 2070 Super | 8 GB | Full GPU | ~155.8 | 448 GB/s | |
| NVIDIA GeForce RTX 2070 | 8 GB | Full GPU | ~155.8 | 448 GB/s | |
| NVIDIA GeForce RTX 2060 Super | 8 GB | Full GPU | ~155.8 | 448 GB/s | |
| NVIDIA GeForce RTX 4080 Laptop | 12 GB | Full GPU | ~150.3 | 432 GB/s | |
| AMD Radeon RX 7700 XT | 12 GB | Full GPU | ~150.3 | 432 GB/s | |
| AMD Radeon RX 6750 XT | 12 GB | Full GPU | ~150.3 | 432 GB/s | |
| Apple M1 Max | 64 GB | Full GPU | ~139.1 | 400 GB/s | |
| Apple M2 Max | 96 GB | Full GPU | ~139.1 | 400 GB/s | |
| Apple M3 Max | 128 GB | Full GPU | ~139.1 | 400 GB/s | |
| NVIDIA GeForce RTX 3070 Laptop | 8 GB | Full GPU | ~133.6 | 384 GB/s | |
| AMD Radeon RX 6700 XT | 12 GB | Full GPU | ~133.6 | 384 GB/s | |
| NVIDIA GeForce RTX 3060 | 12 GB | Full GPU | ~125.2 | 360 GB/s | |
| NVIDIA GeForce RTX 3060 Laptop | 6 GB | Full GPU | ~116.9 | 336 GB/s | |
| NVIDIA GeForce RTX 2060 | 6 GB | Full GPU | ~116.9 | 336 GB/s | |
| NVIDIA GeForce GTX 1660 Super | 6 GB | Full GPU | ~116.9 | 336 GB/s | |
| NVIDIA GeForce GTX 1080 | 8 GB | Full GPU | ~111.3 | 320 GB/s | |
| NVIDIA GeForce RTX 4060 Ti 16GB | 16 GB | Full GPU | ~100.2 | 288 GB/s | |
| NVIDIA GeForce RTX 4060 Ti 8GB | 8 GB | Full GPU | ~100.2 | 288 GB/s | |
| NVIDIA GeForce GTX 1660 Ti | 6 GB | Full GPU | ~100.2 | 288 GB/s | |
| AMD Radeon RX 7600 XT | 16 GB | Full GPU | ~100.2 | 288 GB/s | |
| AMD Radeon RX 7600 | 8 GB | Full GPU | ~100.2 | 288 GB/s | |
| AMD Radeon RX 7600M XT | 8 GB | Full GPU | ~100.2 | 288 GB/s | |
| AMD Radeon RX 6650 XT | 8 GB | Full GPU | ~97.4 | 280 GB/s | |
| Apple M4 Pro | 48 GB | Full GPU | ~95 | 273 GB/s | |
| NVIDIA GeForce RTX 4060 | 8 GB | Full GPU | ~94.6 | 272 GB/s | |
| NVIDIA GeForce RTX 4070 Laptop | 8 GB | Full GPU | ~89 | 256 GB/s | |
| NVIDIA GeForce RTX 4060 Laptop | 8 GB | Full GPU | ~89 | 256 GB/s | |
| NVIDIA GeForce GTX 1070 Ti | 8 GB | Full GPU | ~89 | 256 GB/s | |
| NVIDIA GeForce GTX 1070 | 8 GB | Full GPU | ~89 | 256 GB/s | |
| AMD Radeon RX 6600 XT | 8 GB | Full GPU | ~89 | 256 GB/s | |
| NVIDIA GeForce RTX 3050 | 8 GB | Full GPU | ~77.9 | 224 GB/s | |
| AMD Radeon RX 6600 | 8 GB | Full GPU | ~77.9 | 224 GB/s | |
| Apple M1 Pro (10-core GPU) | 32 GB | Full GPU | ~69.6 | 200 GB/s | |
| Apple M1 Pro (16-core GPU) | 32 GB | Full GPU | ~69.6 | 200 GB/s | |
| Apple M2 Pro | 32 GB | Full GPU | ~69.6 | 200 GB/s | |
| NVIDIA GeForce RTX 4050 Laptop | 6 GB | Full GPU | ~66.8 | 192 GB/s | |
| NVIDIA GeForce GTX 1660 | 6 GB | Full GPU | ~66.8 | 192 GB/s | |
| NVIDIA GeForce GTX 1650 Super | 4 GB | Full GPU | ~66.8 | 192 GB/s | |
| NVIDIA GeForce GTX 1060 6GB | 6 GB | Full GPU | ~66.8 | 192 GB/s | |
| Apple M3 Pro | 36 GB | Full GPU | ~52.2 | 150 GB/s | |
| NVIDIA GeForce GTX 1650 | 4 GB | Full GPU | ~44.5 | 128 GB/s | |
| Apple M4 | 32 GB | Full GPU | ~41.7 | 120 GB/s | |
| NVIDIA GeForce GTX 1050 Ti | 4 GB | Full GPU | ~39 | 112 GB/s | |
| Apple M2 | 24 GB | Full GPU | ~34.8 | 100 GB/s | |
| Apple M3 | 24 GB | Full GPU | ~34.8 | 100 GB/s | |
| Apple M1 | 16 GB | Full GPU | ~23.7 | 68 GB/s | |
| NVIDIA GeForce GTX 1060 3GB | 3 GB | Partial GPU | ~46.8 | 192 GB/s |