I seriously considered using Grayskull. My desire was driven by the interest in NeRF and Gaussian Splatting, for training whose the Grayskull arhitecture (at le

I seriously considered using Grayskull. My desire was driven by the interest in NeRF and Gaussian Splatting, for training whose the Grayskull arhitecture (at least, on paper) could considerably outperform NVIDIA. I decided against it for three reasons: R1. Grayskull is a chip that they made in 2020 ([1]). They already have 2-3 newer chips generations available internally (Wormhole, Black Hole, Grendel), see [2]. R2. The price of the cards is high. It does not look like a too good value, even if judging by paper specs, and then the reality will always correct that down. R3. The primary Tenstorrent founder recently left the company to start a competitor, Taalas, while taking some of the Tenstorrent talent with him ([3]). I decided to wait until at least Wormhole cards available for purchase (hopefully, at a better price), before committing my time and money to Tenstorrent, even if just for exploration. 1. https://insidehpc.com/2020/05/tenstorrent-ships-grayskull-high-performance-ai-processor/ https://insidehpc.com/2020/05/tenstorrent-ships-grayskull-hi... 2. https://www.semianalysis.com/p/tenstorrent-blackhole-grendel-and https://www.semianalysis.com/p/tenstorrent-blackhole-grendel... 3. https://betakit.com/tenstorrent-founder-reveals-new-ai-chip-startup-taalas-with-50-million-in-funding/ https://betakit.com/tenstorrent-founder-reveals-new-ai-chip-...