To be honest, it's not that bad if performance increases 1:1 according to the number of Xe-cores. It could be worse, for example on AMD iGPUs there is hardly any difference between 8 and 12 CUs. They only have like 15-20% improvement per 50% more cores.
That being said... the "improvement" on Panther Lake is still somehow a joke, because the news media mostly neglected the fact that it has 50% more Xe-cores.
Next thing is, that the Arc 140T iGPU (which is still Alchemist with Xe1 cores) already achieved 4348 points in TimeSpy Graphics Score, according to a recent ETA PRIME video (with timestamp) youtu.be/Z_-omwSu3Ek?si=F4eHHeco4lShc1s9&t=249
You could bet, that if you add +50% more Xe cores on Alchemist, it would also end up with a 6000 points Timespy score. Now it's clear why the Intel presentation also did not show any results with Panther Lake only having 8 Xe-cores. Because then everybody would have seen that there is almost no difference compared to Alchemist, and as such almost no difference between Xe3 and Xe1. The only good thing about this is, that if you still have a computer with "only" Xe1-cores, you don't have to feel "outdated", because the raw graphics performance is still similar.
You people are easily impressed. It has 50% more cores, and is on a newer architecture in Xe3 so should be able to do a lot better IMO, which means either it's bandwidth starved or Xe3 is a weak sauce update.
The just-launched Intel Panther Lake flagship, the Core Ultra X9 388H, has been tested with 3DMark Time Spy Graphics to judge the performance of the onboard graphics. Panther Lake's next-gen Celestial iGPU with 12 Xe cores is reportedly up to 50% better than the Lunar Lake Arc 140V.