mlboydaisuke commited on
Commit
c4cdbf7
·
verified ·
1 Parent(s): 87924d3

card: same-day ANE vs GPU A/B on iPhone 17 Pro

Browse files
Files changed (1) hide show
  1. README.md +2 -0
README.md CHANGED
@@ -101,6 +101,8 @@ Gated on an iPhone 17 Pro (iOS 27.0) against the fp32 HF oracle: teacher-forced
101
  method, 512-token prompt, 1024 generated, 5 trials, two back-to-back runs): decode **69.6 / 58.6 tok/s**, prefill 2710 / 2364 tok/s; footprint 1.3 GB (p128 / g256: decode 62.3). The Neural Engine is the power-efficient lane; these are
102
  not a GPU comparison (different protocol from the `int8/` row above).
103
 
 
 
104
  ## Measured
105
 
106
  | | decode | prefill | numerics | size |
 
101
  method, 512-token prompt, 1024 generated, 5 trials, two back-to-back runs): decode **69.6 / 58.6 tok/s**, prefill 2710 / 2364 tok/s; footprint 1.3 GB (p128 / g256: decode 62.3). The Neural Engine is the power-efficient lane; these are
102
  not a GPU comparison (different protocol from the `int8/` row above).
103
 
104
+ Same-day interleaved A/B on the same phone (one app embedding both bundles, 128-token prompt, 256 generated, 5 trials, ANE-GPU-ANE-GPU): Neural Engine 8-bit **76.8 / 76.8 tok/s** decode (3899 / 3865 prefill) vs this repo's `int8/` GPU bundle **67.1 / 64.4** (2359 / 2270) — equal weight bytes per token, so +17 % decode and +70 % prefill on the ANE; GPU cold specialization 3.6 s vs ANE AOT load 0.2 s.
105
+
106
  ## Measured
107
 
108
  | | decode | prefill | numerics | size |