20x error in your benchmark results
Nobody has claimed this yet.
- Dominant language
- Objective-C
- Stars
- 7.3k
- Forks
- 960
- PR merge metrics
- No merged PRs in 30d
Description
Your work became quite famous, in particular cited in https://www.quora.com/Why-are-so-many-people-willing-to-buy-Mac-Minis-to-run-OpenClaw/answer/Franklin-Veaux
> The ANE is roughly 80× more efficient per FLOP than an A100. Of course, the A100 has 50× more total throughput.
This means that A100 uses 4,000 times more energy, i.e. 13 KW. It's 20-30 times more than the real A100 power usage, and once you use real numbers, A100 is pretty close to ANE. And if you compare with a modern GPU, like 5090, any difference will be gone.
There is no Apple magic; the silicon is the same, and you can't pull 80x more efficiency than the rest of the world (which has pretty similar efficiency). You can use outdated GPUs like A100 to get 2-3x boost compared to modern silicon, but that's all.
Your work is GREAT, but unfortunately, Apple boys can't count their fingers and unconditionally believe everything that looks like Apple magic.
Contributor guide
No contributing guide indexed for this repository
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Start by reviewing the benchmark results and the A100 and 5090 power and throughput assumptions described in the issue. Done means reproducing the comparison and either correcting the reported energy figures or documenting why the existing figures are valid.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- objective-c
- Domain
- machine-learning, performance
- Issue type
- Bug
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 35/100