And you are assuming they can fix it because?
What if it's a hardware issue?
Continued...
I know how bad Intel is with GPUs. Here's few examples why I think driver overhead is a significant problem:
1. Anyone remember GMA X3000? I owned it. They were fashionably late on the driver side, the one supporting hardware T&L launched something like 10 months later. And people were expecting great results. And in some cases it did, like with Half Life 2: Lost Coast. But in other cases it was
slower, because it was later found out due to them skimping out on transistors and having an anemic geometry engine.
Ok, so some is due to hardware, but not having feature level support until 10 months later showed serious deficiency on the software team.
Another possibility on why the geometry sucked: Older Intel graphics scaled great with CPU, because it relied on it for geometry. Maybe that mentality carried on to X3000. Why? To sell more CPUs. The old adage Microsoft does everything to sell more Windows, Intel does everything to sell more CPUs.
2. Even with the modern ones like Iris Xe, videos show that the drivers put too much emphasis on the CPU. AMD CPU runs at 1.4GHz, while Intel runs the CPU at over 3GHz. Why? Maybe they themselves know it's necessary.
3. Pat's admission that performing bad is due to them thinking just porting the iGPU driver code base would have been enough. Tom Peterson said that it wasn't, because the dGPU being much faster exposed driver-level bottlenecks.
4. Lunarlake. LNL performs much better relatively on games compared to Time Spy versus Meteorlake. So if utilization wasn't addressed on Xe2, then it should show similar issues. Yet it's noticeably better.
5. Rumors that they were working on a driver to address not only DX11 deficiencies, but DX12 as well. And I don't mean specific games, but a general one.
See, the thing is, they don't have the promised VR support, they killed ARC Control, they eliminated video capture portion, and the DX11 non-whitelist driver isn't there, which was also promised.