Agreed there. The only advantage is if you have more performant silicon for a given power envelope that exceeds the compatibility overhead of running through FEX. This is why in early days, Rosetta occasionally did just as well on M1 Mac’s as the Intel based software, with the leap in performance more than overcoming that translation layer impact.
This is true. But I’m more looking at the work they’re putting in on the Frame as a bellwether. It seems odd they’d support a tiny platform for them with an ARM cpu (that matches typical Android phone hardware). It’s possible they just want to increase the compatibility layers in that space in a general sense.
I guess I’m looking at the M5 chip 2x’ing the performance (single core) vs the M1 variant from 5 years ago almost exactly, and over that same frame the best desktop cpus have seen a 1.5x’ing. A mobile chip does better single core than desktop chips from AMD and Intel (caveats here). The perf per watt is absurd.
To be clear, I don’t think this will happen quickly. I expect another 2 years before the next steam deck arrives. I just think it’s interesting that valve is supporting a non x64/x86 architecture in a product category here and have a hard time believing it’s a total dead end for them.
Honestly, more small orders getting cancelled is also awesome, good on you!
I don’t think it’s actually true that in low spec ranges AMD and Intel are competitive. The Apple a19 pro chip has a tdp of 8w while a Z1 extreme from amd has a 15w envelop that goes up to 30w. The A19 crushes the Z1 in single core and is 90idh percent on multi core. The fp32 performance is double the z1 as some indication of gpu horsepower. So let’s just say near the same performance at less than half the tdp. Or another way, same steam deck performance you’ve had (well better actually, steam deck doesn’t have a z1 extreme) at twice the battery life. The A19 Pro is also in a passively cooled device where a Z1 Extreme is actively cooled. Data sources for this: just looking at geekbench and pass mark scores that I could find. Of course there’s instruction translation overhead, and it’s not as clear cut as this (for one, Valve is not likely to poach chip designers from Apple and they seem reticent to create their own hardware), but still a thought worth considering.
Ultimately I don’t care if it’s arm based, I care about the performance of the machine itself (in totality, which the steam deck excels at even still).
So I guess in a long winded way, I’m agreeing with you that they should maximize the performance up to 15w (I would have said 30w for docked access but the steam machine seems to be their goal for the living room). I guess I am just not super convinced legacy chipmakers have what it takes to be competitive, even with a FEX penalty. I think we won’t see a steam deck 2 for another 2 years, and that’s a long time for FEX to mature, drivers to mature, and Valve to line up a low power, extremely strong device.