speed-bench: add Mac Studio M1 Ultra 64GB streaming numbers#350
Open
ozgursoy wants to merge 1 commit into
Open
speed-bench: add Mac Studio M1 Ultra 64GB streaming numbers#350ozgursoy wants to merge 1 commit into
ozgursoy wants to merge 1 commit into
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Real-hardware 64GB streaming datapoint. The "Flash on 64GB MacBooks" section
currently only documents a 128GB machine with 64GB locked away, this is an
actual 64GB box.
Machine: Mac Studio M1 Ultra, 64GB, macOS 26.5
Model: Flash Q2 (DeepSeek-V4-Flash-IQ2XXS-w2Q2K...imatrix, ~81GB)
Command:
./ds4-bench -m ds4flash.gguf
--ssd-streaming --ssd-streaming-cache-experts 32GB
--ctx-start 2048 --ctx-max 32768 --step-incr 2048 --gen-tokens 128
--csv speed-bench/m1_ultra_64gb_stream.csv
Results (2K-32K ctx): prefill ~108-118 t/s, generation ~5 t/s, both roughly
flat as context grows. Decode is SSD-bound: it does not scale with the Ultra
GPU, so generation stays close to what smaller 64GB Apple Silicon reaches.
Bonus, simulated 32GB (--ssd-streaming --simulate-used-memory 32GB): with only
~32GB available the expert cache can't stay resident for the 81GB model, so
decode collapses to ~0.17 t/s (prefill ~60 t/s) - not practically usable. Can
add a CSV for it if useful.