Nobody cares! It's all about the integrated graphics performance of Apple Silicon.
You asked why nobody cares. It is because it isn't happening.
The only way they can increase generalized graphics performance by significant margins is to actually dedicate more of their power and silicon budget to gaming performance. This is simply not possible for mobile phone chips - because the batteries can only get so large before they are no longer able to be taken on planes, and the phones can only get so hot until people can no longer hold them and they start to have to throttle to keep from components from failing.
Again. We're talking about the iPhone GPU, not the Mac Pro.
If you want to limit "Apple Silicon" to just the iPhone GPU, then no. There is zero chance there are going to be significant increases to catch up with desktop-class or console-class GPU.
You need to stop pretending as if there's any competition for Apple. We're not talking about the PC gaming market, we're talking about gaming on Apple Silicon. iOS, iPadOS and macOS.
There is no competition in the market of Apple Silicon running on Apple devices, because only Apple can manufacture them
There are no trade offs. Apple invests in their own chip technology and the customers reap the benefits for free.
Architecturally there are a ton of trade offs, some of which I listed.
Of course it [PC gaming market] is too small, but luckily Apple owns the largest gaming market that ever was, the AppStore. And the same Metal engine runs on everything.
The App Store does not have PC gaming titles, both because of porting difficulty and because the systems do not have sufficient horsepower.
I'll need a citation for Apple owning the largest gaming market that ever was, given that there are more Android devices than iPhones by a fair margin.
Nobody cares about existing games. And that very special task is to create realism through natural lighting. That's useful for every virtual world.
Lots of people care about high quality, existing games.
Very few people care about "metaverse" virtual worlds.
The next process node shrink will add a few billion more transistors to the chip in need for a good purpose. We already have a Neural Engine, which most people don't need. A Gaming Engine is a no-brainer. Of course you add it if you can.
Process shrinks do not "add" transistors. Adding transistors is a cost and yield and equation. Newer processes shrink the die, but typically come with lower yield and higher per-area costs. It is the improved power performance (and in CPU cases, reduced infrastructure needed for signal travel time) that allows them to scale more cores or have a higher clock.
It will be interesting when you get that there is no differentiation anymore between an iPhone SE and the Mac mini with M1 Ultra. They are meant to be the same platform, with the same basic chip architecture. A few more cores here and there make no difference.
There is no M1 Ultra Mac mini, and there never will be.
If you are talking about the Mac Studio with M1 Ultra, then the difference is that there is a different balance of performance vs efficiency, a much wider memory architecture, additional infrastructure for Rosetta, additions of PC hardware infrastructure (e.g USB4, SSD, thunderbolt, HDMI) , 8x the RAM and 4-6x the number of cores.
There are other features which are harder to tell if they are present in the A15 vs M1 - such as virtualization/paravirtualization support, and GPU backward compatibility for non-tile families in metal.
The max TDP of the iPhone SE processor and GPU is 6W. The Ultra is 60W. It's literally an order of magnitude difference to support the additional features and processing power.
What about a feature for every iPhone do you not get?
The pictures are only prettier, because ray tracing support makes the necessary calculations so much faster that you can actually switch the feature on. You could always calculate pretty pictures on any computer, ray tracing support on the chip is what makes the frame rates acceptable.
The frame rates are not acceptable with zero ray tracing. Adding some zero cost hardware ray tracing might make the shading look
nicer, but it won't improve the texel count. Best case scenario - the current 10 fps for a demanding title will be maintained, but now those ticks will be pretty enough to frame.
