monster/corridorkey-mrp-mlx
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

docs: M4 mini (7th) completes the matched-width M3->M4->M5 generation ladder

Browse Source 
hdim-56 fallback is textbook here (1.00x x3); padding = 2.0x. Fits the
corrected fused-kernel-quality theory as a prediction, not a fit: 0.50x
fused/unfused -> 2.0x padding win, exactly between M3 Air (0.40x->2.6x) and
M1 Max (0.87x->1.3x). At 10c matched width the ladder is 138.9/48.8/44.0ms at
hdim56 but 54.3/24.4/8.9ms at hdim64 — M5's gains live almost entirely in the
fused kernel, making the head-dim fix most valuable on the newest silicon.
This commit is contained in:
modelbeast 2026-07-17 14:30:57 +10:00
parent e8f1721996
commit dfc2bf4a51
1 changed files with 26 additions and 14 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

40
docs/2026-07-16-m-series-fleet-ablation-results.md
View File

@ -1,18 +1,18 @@
---
title: "M-series fleet ablation: 6 machines, M1 → M5 (8GB mini → 256GB Ultra)"
title: "M-series fleet ablation: 7 machines, M1 → M5 (16GB mini → 256GB Ultra)"
date: 2026-07-16
device: "M1 Max 32GB (24c) · M1 Ultra 128GB (64c) · M3 Air 16GB (10c, fanless) · M3 Ultra 256GB (80c) · M4 Pro mini 24GB · M5 MacBook Pro 16GB (10c)"
device: "M1 Max 32GB (24c) · M1 Ultra 128GB (64c) · M3 Air 16GB (10c, fanless) · M3 Ultra 256GB (80c) · M4 mini 16GB (10c) · M4 Pro mini 24GB · M5 MacBook Pro 16GB (10c)"
script: scripts/bench_optimizations.py (unmodified) + engine-level tiled×compile extension
---
# M-series Fleet Ablation Benchmarks
Same methodology as the 2026-03-09 wave2 doc (`--sweep ablation --resolution 512 1024 2048`,
checkpoint weights, 3 warmup + 10 bench runs), executed across **six machines** spanning the
checkpoint weights, 3 warmup + 10 bench runs), executed across **seven machines** spanning the
Apple Silicon range — **M1 Max 32GB (24c) · M1 Ultra 128GB (64c) · M3 Air 16GB (10c, fanless) ·
M3 Ultra 256GB (80c) · M4 Pro mini 24GB · M5 MacBook Pro 16GB (10c)** — i.e. every current GPU
generation, Studio→fanless laptop, an M1 width ladder (24c→64c) and a matched-width M3-vs-M5
pair. Contributed from a multi-Mac render-farm setup; happy to run follow-ups on any of them.
M3 Ultra 256GB (80c) · M4 mini 16GB (10c) · M4 Pro mini 24GB · M5 MacBook Pro 16GB (10c)** —
i.e. every current GPU generation, Studio→fanless laptop, an M1 width ladder (24c→64c) and a
**matched-width M3→M4→M5 generation ladder (all 10-core)**. Contributed from a multi-Mac render-farm setup; happy to run follow-ups on any of them.
## Headline findings
@ -67,8 +67,14 @@ pair. Contributed from a multi-Mac render-farm setup; happy to run follow-ups on
| M1 Ultra | 64c | 150.5 ms | 53.8 ms | **2.7×** | 0.59× |
| M3 Air | 10c | 138.9 ms | 54.3 ms | **2.6×** | 0.40× |
| M3 Ultra | 80c | 63.2 ms | 28.6 ms | **2.2×** | 0.53× |
| M4 mini | 10c | 48.79 ms | 24.40 ms | **2.0×** | 0.50× |
| **M1 Max** | **24c** | 20.16 ms | 15.79 ms | **1.3×** | **0.87×** |
Note how tightly the padding win tracks the fused/unfused column (0.14×→4.9×,
0.40×→2.6×, 0.50×→2.0×, 0.87×→1.3×) and how poorly it tracks core count. The
M4 mini was measured *after* the correction below and fits it — the theory made a
prediction and held.
We initially read this as "narrower GPU ⇒ bigger win" (M5 10c gains most, M3 Ultra 80c
least). **The M1 Max refutes that**: it is mid-width (24c) yet gains the *least* of all.
The actual predictor is each chip's **fused-kernel quality relative to its own raw matmul
@ -139,16 +145,22 @@ vs 12.2 ms). **How much the fast path is worth is chip-specific, not width-speci
(M1 Max 24c: 0.87×; M5 10c: 0.14×) — see the six-machine table above. (M5 numbers taken
on a machine in active desktop use; treat as indicative, not lab-clean.)
### Generational delta at *matched* GPU width — M3 Air vs M5, both 10-core
### Generational delta at *matched* GPU width — M3 → M4 → M5, all 10-core
The Ultras confound generation with width. These two don't: both are 10-core laptop GPUs,
two generations apart (fused SDPA, hdim 64, fp32):
The Ultras confound generation with width. These three don't: all are 10-core GPUs, one
generation apart each (fused SDPA, hdim 64, fp32):
| seq | M3 Air (10c) | M5 (10c) | **M5 speedup** |
|---|---|---|---|
| 1024 | 2.44 ms | 0.80 ms | **3.1×** |
| 2048 | 17.51 ms | 2.32 ms | **7.5×** |
| 4096 | 55.77 ms | 7.22 ms | **7.7×** |
| seq | M3 Air (10c) | M4 mini (10c) | M5 (10c) | **M5 vs M3** |
|---|---|---|---|---|
| 1024 | 2.44 ms | 1.86 ms | 0.80 ms | **3.1×** |
| 2048 | 17.51 ms | 6.35 ms | 2.32 ms | **7.5×** |
| 4096 | 55.77 ms | 24.47 ms | 7.22 ms | **7.7×** |
At hdim **56** (what the model actually runs today) the same ladder is 138.9 → 48.8 →
44.0 ms: the M4 nearly triples the M3, but the **M5's advantage is concentrated almost
entirely in the fused kernel** — its unpadded step is barely better than the M4's, while
its padded step is ~3× better. i.e. M5's headline MLX gains show up *only if you hit the
fast path*, which makes the head-dim fix especially valuable on the newest silicon.
~7.7× at identical core count — i.e. the M5's gain is architectural, not width. (Apple's own
published MLX figures claim ~3.8× M4→M5 on FLUX image-gen; this attention workload shows more.)