The other issue is that unified memory might be nice for original games. But Mac games are going to be ports - and the assets are going to be built around 8, 16, and 32 gigabyte GPUs.
if Apple pegged non macOS gaming as a top priority there would be 3rd party GPU drivers . There are none.
There would also be more open/portable graphics API support. All Apple has a deprecated OpenGL and an officially ignored Vulkan.
I highly doubt Apple is going to contort and twist their GPU architecture into dramatically reducing the costs of porting older Windows games. I don't think legacy Windows gaming apps is the primary target for the large footprint Unified memory. I think that comment about gamers developers having to wrap their head around it was for future games and current new development on gaming engines.
The ported stuff that makes no account for the new platform will run worse. The folks who put in the work to do fully adapted ports will run better. Is Apple going to cry tears when the folks who pay attention at WWDC do better than the folks who completely ignore it? Probably not.
Has Apple been throwing money like drunken sailors at folks who will do any port. Yeah. Are they going to continue to do long period when the port warmed over code with large impedance mismatch with the Metal API ? Probably not.
Apple put lots of effort into the Rosetta 2 port to help Intel macOS code come across unchanged. For the graphics stack on macOS on M-series transparent and unchanged really hasn't been the main message. They want changes so that the new graphics code runs well completely up and down the Apple GPU configuration stack.
"... And I don't think we're going to fool anybody by saying that overnight we're going to make Mac a great gaming platform. We're going to take a long view on this." ..."
there is a long term strategy here. Apple has enough money that 'long' can be 3-4 years. Nor is this necessarily total consumption of the entire gaming market performance in 3-4 years. when he was talking about gaming it was really more so about ,pre gaming on the whole Mac product line ; not some narrow niche on the upper 1% end.
Mostly likely Apple is looking to connect the already relatively large gaming revenues on iOS/iPad to that of the macOS. Not coming up with a 'gaming PC market killer' system or product line.
Not to mention, even if you had 96 gigs worth of assets, you don't have the GPU power to render them quickly enough for a game. On top of the bandwidth issues.
How much of a 80GB super complex model is going to be actually visible from a fixed viewport perspective? Apple's major approach is to "trim off " as much as they can before getting eyeball deep in the rendering.
Metal 3 has features for asset loading straight from the SSD . If use 40GB for effective caching then are actually alleviating bandwidth ( if trying to point at the RAM to die bandwidth cap). that isn't disconnected from new, future growth features on the Windows side that has a similar DirectStorage API.
In both load times and FPS
www.tomshardware.com
both are a more 'unified' path to loading managing assets. Longer term it will likely remove the notion of managing duplicates in the "CPU RAM" on both operating systems (and associated filesystem caches).
M2 tweak up the cache higher. (probably likely increased the internal bus aggregate bandwidth). The media en/decode got better along with the GPU performance bump.