Skip to content

Conversation

@starkdmi
Copy link
Contributor

Fixed:

  • M2 Pro ~6.8 (FP32, 19 cores)
  • M3 4.1 (FP32, 10 cores)
  • M3 Pro 6.39-7.4 (FP32, 18 cores)

New:

  • M4 4.6 (FP32, 10 cores)
  • M4 Pro 9.2 (FP32, 20 cores)
  • M4 Max 18.4 (FP32, 40 cores)

Fixed:
- M2 Pro ~6.8 (FP32, 19 cores)
- M3 4.1 (FP32, 10 cores)
- M3 Pro 6.39-7.4 (FP32, 18 cores)

New:
- M4 4.6 (FP32, 10 cores)
- M4 Pro 9.2 (FP32, 20 cores)
- M4 Max 18.4 (FP32, 40 cores)
Copy link
Member

@pcuenca pcuenca left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Looks good to me! Where did you get the data?

},
"Apple M2 Pro": {
tflops: 13.6,
tflops: 6.8,
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Was this wrong?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It's more than M2 Max (13.49), probably FP16? But most of Apple Silicon benchmarks here are FP32 as those are sometimes reported by Apple and more probably are correct.

@starkdmi
Copy link
Contributor Author

Looks good to me! Where did you get the data?

A few benchmarks websites plus Apple references.

Copy link
Member

@pcuenca pcuenca left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Ok for me to make Apple Silicon numbers consistent and use FP32 if those figures come from ~official sources. Let's wait for another review to merge :)

Copy link
Member

@julien-c julien-c left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

looks awesome! Coincidentally, I just received my M4 Max MBP so this is perfect timing:)

@pcuenca pcuenca merged commit 21158d7 into huggingface:main Nov 20, 2024
4 checks passed
@starkdmi starkdmi deleted the patch-1 branch November 20, 2024 15:44
@julien-c
Copy link
Member

@julien-c
Copy link
Member

thx a ton for the contrib!

@pcuenca pcuenca mentioned this pull request Mar 21, 2025
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants