That versatility comes at a cost though - nothing in silicon is free. It clearly couldn't compete with GPUs in their strengths, and couldn't find it's own niche

That versatility comes at a cost though - nothing in silicon is free. It clearly couldn't compete with GPUs in their strengths, and couldn't find it's own niche, which is why it was cancelled. And I'm not sure it'll be that much easier to program - "few large cores" and "GPU waves" are now pretty well supported in the stack with mature tooling - trying to insert a new stack between the two would likely be pretty difficult as it doesn't really fit well with either established paradigm so likely needs something new to show it's benefits.