Hacker News Reader

This is cool, but something tells me the t/sec rate compared to native is still wildly off-base. How much less optimized is WebGPU compared to the current OpenCL/CUDA implementations?

View on HN

Good point, but like most things if you give it time it will improve exponentially... remember that WebGPU is still in it's early days, after all!

Reply on news.ycombinator.com