NVIDIA Tensor Core Programmability, Performance & Precision · HackerPapers discussionshttps://hackerpaper.kaipfister.de/p/1803.04014/Collected HN comments and submissions linking this paper.enMon, 02 Nov 2026 11:16:22 GMTNVIDIA Tensor Core Programmability, Performance & Precision — mentioned by m0zghttps://news.ycombinator.com/item?id=22246330urn:hackerpapers:mention:1803.04014:22246330Wed, 05 Feb 2020 12:57:56 GMT<p>https://arxiv.org/pdf/1803.04014.pdf - no matter what they did, they could not go beyond 83TFLOPs in fp16. And that&#x27;s just matrix multiply. Any kind of deep learning workload is going to be a lot slower than that.</p><p><a href="https://hackerpaper.kaipfister.de/p/1803.04014/">Paper and discussion history</a>. A mention is not an endorsement.</p>cs.DC