Regarding Halo iGPUs: https://videocardz.com/newz/intels-...o-rival-planned-questions-amd-igpu-efficiency
This is more important that it reveals what they think internally about halo iGPUs. Another thing is they tried this already, twice, and went nowhere. dGPUs benefit from the volume ecosystem that Halo iGPUs don't get.
What I think is ok to try is basically take the mainstream chip, keep everything same, but 1.5x more compute, which will probably get 20-25% better performance. So let's say for Pantherlake, we'd have an 18 Xe core version.
This is more important that it reveals what they think internally about halo iGPUs. Another thing is they tried this already, twice, and went nowhere. dGPUs benefit from the volume ecosystem that Halo iGPUs don't get.
What I think is ok to try is basically take the mainstream chip, keep everything same, but 1.5x more compute, which will probably get 20-25% better performance. So let's say for Pantherlake, we'd have an 18 Xe core version.
