Xeon Phi, maybe? Since the entire raison d'être of that chip is doing avx-512 computation. Or if you're a glass half empty kind of person, you'd argue that it doesn't speed up when running scalar code. :)