2benchesM2
A15 · M2 · Avalanche / Blizzard · 5 nm · First available 2022-06-24
Every configuration
| Configuration | CPU core layout | Bandwidth | GB7 CPU | GB7 Metal | Prefill Q4_0 tok/s | Decode Q4_0 tok/s |
|---|---|---|---|---|---|---|
| 8 CPU / 8 GPU (binned) | 4P + 4E | 100 GB/s | — | — | 145.9 | 21.7 |
| 8 CPU / 10 GPU (full chip) | 4P + 4E | 100 GB/s | 9,693 | 43,178 | 179.6 | 21.9 |
All benchmarks
| Benchmark | Median |
|---|---|
| Geekbench 7 · Multi-core | 9,845 (9,616–9,964, 8 sources) |
| Geekbench 7 · Metal GPU | 42,996 (39,836–43,671, 4 sources) |
| Geekbench AI · GPU · Single | 6,713 (6,172–7,297, 4 sources) |
| Geekbench AI · GPU · Half | 7,458 (6,856–8,097, 4 sources) |
| Geekbench AI · GPU · Quantized | 6,526 (6,089–7,051, 4 sources) |
| Geekbench AI · Neural Engine · Single | 3,303 (3,220–3,351, 4 sources) |
| Geekbench AI · Neural Engine · Half | 24,294 (23,632–24,675, 4 sources) |
| Geekbench AI · Neural Engine · Quantized | 26,774 (26,071–27,121, 4 sources) |
| Prefill 512 · F16 | 201.3 (201.3–201.3, 1 sources) |
| Decode 128 · F16 | 6.7 (6.7–6.7, 1 sources) |
| Prefill 512 · Q8_0 | 164.3 (147.3–181.4, 2 sources) |
| Decode 128 · Q8_0 | 12.2 (12.2–12.2, 2 sources) |
| Prefill 512 · Q4_0 | 162.7 (145.9–179.6, 2 sources) |
| Decode 128 · Q4_0 | 21.8 (21.7–21.9, 2 sources) |
Similar performance
Nearest measured configurations by Geekbench 7 CPU and Metal. Ratios above 1 are faster.
8 CPU / 10 GPU (full chip)
- M3 8 CPU / 8 GPU (binned), 2023-11: CPU 1.2×, GPU 0.99×
- M3 8 CPU / 10 GPU (full chip), 2023-11: CPU 1.2×, GPU 1.11×
- M4 8 CPU / 8 GPU (binned), 2024-05: CPU 1.38×, GPU 1.09×
- A18 Pro 6 CPU / 5 GPU (binned), 2024-09: CPU 0.75×, GPU 0.72×
- M4 10 CPU / 8 GPU (binned), 2024-05: CPU 1.55×, GPU 1.1×
- M1 8 CPU / 8 GPU (full chip), 2020-11: CPU 0.85×, GPU 0.65×