Xbox Series... M?

Quick follow-up to this: JEDEC just published the LPDDR5X standard, which clocks at no less than 8533MT/s:
https://www.anandtech.com/show/16851/jedec-announces-lpddr5x-at-up-to-8533mbps

With this, Microsoft could make a portable Xbox One S console using just 64bit LPDDR5X, as long as they enabled the same 32MB ESRAM into the SoC.

 

What would be the benefit of going 64bit vs 128bit? Cheaper to make? Better battery life?

 

Less power due to less amount of chips and traces, smaller apu.

Cheaper would depends on how much a LPDDR5X dimm cost.

 

So I did more exploration on what a Xbox handheld would look like.

MS needs a custom 8-9" 1080p LCD / OLED as all common options in those ranges are 1920 x 1200. There are options in the sub 7" and over 10" ranges, but one extreme is too small and the other extreme is too big.

They also need a custom solution for the display MCU to enable VRR.

They can remove the touch capabilities from their display because no games will support it and the OS will need non-trivial rework to support touch. This should solve BOM costs as they don't need a digitizer or a screen with integrated touch.

So summarize the specs:

8 core Zen at 1.8+ghz. SMT disabled in BC mode.
6 WGP (7 total, 1 disabled), 12 CUs at 1-1.3ghz
8GB of LPDDR5 / LPDDR5X, 68GB/s. 12GB if devs really push for it.
32MB LLC
512GB SSD / flash solution with the lowest TDP that delivers at least 600MB/s uncompressed. Perhaps Sata 3?
8-9" 1080p OLED / LCD. No touch capability. VRR range 48-60hz.

Since the screen is already so large, I think the below form factor makes sense. The home button will be moved to the bottom piece so the top can be one piece.



Although looking at it, the expectations would be that the screen is touch capable. Perhaps we remove the power button altogether. The opening and closing of the unit is the power button.

 

IMO, the Series M is either:

1 - Capable of running XB One games in a Switch-like handheld form factor, it has the bare minimum specs to run them and is priced accordingly. For this to work, it'd need to release ASAP;

or

2 - Capable of running Series S games, getting ISO specs but much higher price, with a large tablet form factor closer to the Razer Edge.




IMO it would be a terrible idea for Microsoft to introduce yet another target spec for devs to worry about, considering they're most probably adding a mid-gen upgrade to the Series X in 2023-2024..

 

The R&D and cost of shrinking Xb1 chip down to 6/5nm is likely as expensive as making a custom one since 16nm to 5/6nm might not be a straight port and the APU that I proposed will be dual purposed in a Surface device.

It's also possible you get worse performance AND battery life with the old architecture, especially since it was designed for a 110 Watt console originally.

As for the release date, I think releasing it before the next gen only games arriving en masse makes sense. That means latest Q4 2022 / early 2023. Until then, it should still play the majority of new games natively because the cross gen period will likely last for a while.

As for target spec, it'll be optional for devs unlike the situation with the S.

They can always default to Xb1 or to streaming.

However if you're shipping on XB1, 1X, S, X, PC using the GDK, it should be trivial to add optimizations for the Series M by either scaling up from XB1 version, or scaling down from the S version.

I completely agree that the Series M should be ballpark power wise either in the previous gen, or around the S.

I think it needs to be substantially stronger, not accounting for new architecture, than the Xbox One in order to smooth out the performance issues that games had on Xb1.