Why so much recent focus on core size for executing instructions in parallel but still less emphasis on vectors? SVE2 and AMD bringing AVX-512 to Zen 4 are a step forward, at least, but even AMD seem intent on building a wider core (see: comments about Zen 5's issue width from AMD)
SIMD seems like the only way forward for ST performance, at least with a proper SW and HW push, Dennard scaling is over and frequency has run it's course. Logic shrinks continue, but otoh is throwing these transistor budgets at larger cores for diminished ILP returns worthwhile?