Keep in mind while we wait for M5 reviews that we already have reviews of the A19/Pro. e.g.:
Now these are complicated by the fact that the Pro devices got better cooling so increases for 50-60% in non-ray tracing games probably exaggerates what to expect from M-series devices with the same cooling. But ~40% is not unreasonable for some games.
From the video above, we can see Death Stranding tested on the iPhone, but you'll notice Apple has allowed the wattage to go up significantly, by about 15%. Assuming that's all the new cooling and that Apple won't allow the M5 to be any hotter than the M4 at all and its linear (a lot of assumptions), then the M5 would still be about 40% more performant (the other two games, AC and RE get more like 20-25% above wattage). Further you can watch the video yourself for RT and other tests (and we already have a Blender result for the M5) that shows Apple has improved RT even more than raster performance. This should have no problem translating to the M5.
So again, why are people excited? Is it because the base M5 is now incredible gaming machine capable of playing AAA games at the same level as a discrete GPU? No.
But like for people who buy PC ultrabooks, premium or not, and game on them, improving performance of the base chip by this amount is ... well ... insanely good and not really a minor, "iterative" improvement.
Some days you gotta play a top-5-graphically-intensive-game on the smol iGPU you have in front of you.
Absolutely.
Further for the Apple devices that do compete with discrete GPUs, like the Pro and Max, we can extrapolate what their likely performance figures will be and that's very exciting. Again, this won't turn Apple devices into the most economical gaming machines that are going to finally power the Mac gaming revolution. Heck, to that point many if not most games will have to be emulated and no matter how good Apple's CPUs or GPUs are or how well Apple can translate DX12 into Metal that will always cause a performance hit. And there are times where the native port is ropey enough that emulating the X86 Windows version of the game is actually the better experience.
But if the extrapolation from the A19Pro/base M5 to the Pro and Max pans out, then at least the hardware itself is not the limiting factor. The base will be likely (one of) the best ultrabook-class iGPUs out there and for those who want more serious GPU performance, the Pro will be capable of competing against the 5050/5060 mobile*, and the Max capable of competing against ... well in the video he goes all the way to the 5090 mobile*. Which is technically possible looking at the numbers from Blender and Steel Nomad, etc ... - again, most real world games, even if there is a native Mac port it will likely not have the same performance tuning as the PC version to put it mildly. So I'm not necessarily expecting the same comparators for games as in Blender or various graphics benchmarks which spend a lot more time optimizing for the Mac than most game developers ever will. The performance uplift might be the same, but an Apple device typically has further to go.
*important to remember as well there are a huge variation of differently tuned Nvidia graphics cards - there are 5050s that soundly beat 5060s or up (while drawing an insane amount of power) and there are 5050s that ... simply don't get anywhere close. Same for just about every class of Nvidia chip. So it's about comparing to averages or some reasonable selection.