Hacker News new | past | comments | ask | show | jobs | submit login
Wave Computing Launches TritonAI 64 Platform for High-Speed Inferencing (insidehpc.com)
24 points by rbanffy 3 days ago | hide | past | web | favorite | 5 comments

Another interesting offering is from Habana Labs: https://habana.ai/

I eagerly wait for MLPerf https://mlperf.org/ Inference benchmark. Both Habana Labs and Wave Computing are MLPerf members.

A little too much "up to" next to the claims, but even if it's half as good as they say, it'll be a solid offering. Strange to see them call 7nm "industry standard" given that very few companies have access to the technology, and they themselves most certainly do not.

I thought all you had to do to get access was to drop piles of money on TSMC lap. Is the access not common?

This would be rather a different order of magnitude of cash than a 10 year old "startup" with 203M in funding, > 100 employees and no revenues would have. From what I understand the current 7nm TMSC capacity is already maxed out by the likes of Apple, NVIDIA, and AMD. And then TMSC also has just screwed up a bunch of wafers through chemical contamination as well.

I am much ignorant about this, but I am curious to know if any comparison can be established between TritonAI and InfiniBand?

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact