Feasibility of handheld X86 Xbox One on 7nm.

[Proelite]

Smach Z exists but we're not going there.

I am not choosing to discuss Sony handheld because I feel that they can leverage the Japanese companies that are already creating games for mobile or Nintendo for their Vita 2. If Sony were to make a Vita 2, it's probably going to PowerVR / Tegra based.

Microsoft, with its UWP initiative, would most likely want a device that can play run binaries of released X1 games without any game studio input outside of licensing and future games with minimal coding.

The Xbox One X improved Jaguar core ported to 7nm, running at least 1.75 ghz. How many watts is that? Would it be more power efficient to use 7nm quadcore Ryzen clocked very low?

12 GCN CUs, 16 Rops running at least 863mhz. Again, how many watts is that? Which generation of GCN or Navi would be the most power efficient?

I am assuming the 32MB Esram would be in use in junction with LPDDR4.

Is it possible to reach 68 gb/s with LPDDR4? If not can increased esram cache and new GCN compression methods help offset the loss of main memory bandwidth? Would released X1 games need to be 'remastered' in that case?

Edit: seems like LPDDR4 tops out at 25gb/s. It's not a viable candidate for a simple X1 handheld. LPPDR5 has bank grouping.

 

[HBRU]

it seems Xbox One portable is much more on the reach than a PS4 portable... But maybe in that case Ms can also include a phone and make something really smart... I know there is this project Andromeda they are working on... Windows Mobile for Phone I still use it on secondary phone and is great by my opinion.. but actually dead... No more development.

https://m.hindustantimes.com/tech/p...nch-in-2019/story-TAozz2dJ7dYdk6ZtQLuH5O.html

 

[Allandor]

No, I don't think that would work. in 16nm the xbox one is at ~80W. Even if 7nm would bring it down to 20-30W (not realistic) it would still be much to much to cool for a mobile device. The Switch is <10W and that is with the screen.
Yes you could remove the HDD (SD card-based doesn't need as much as a HDD) and the disc-drive, but that shouldn't really matter (<10W). The bigger power supply is also something that draws power in the xbox one, so if you cut all these components the xbox one board, memory and APU might need something like 60W, shrink that to 7nm, use LPDDR4 and you still have more than 20W.

 

[HBRU]

maybe could be done at 5nm

 

[MrFox]

XB1S and PS4 Slim consume somewhere around 80W while gaming, a little more for PS4 but the difficulty is pretty much in the same ballpark.

The switch consumes 6W at best for 2.5 hours gameplay, including the screen. So the SoC is around 4W while gaming.

7nm is not going to be enough, unless the goal is a large, heavy, and expensive tablet.

I can see half specs to be a possibility, but it requires porting down. Maybe using modern hardware like a quarter of X or Pro specs, not based on the original xb1/ps4. Without esram, with color compression, and with a lot of power management. No crazy power consumption idling in the fricking menu. It needs a new platform and some dev efforts to port.

 

[Shifty Geezer]

However, if the SDK for XB2 is so designed for multiple resolutions from the off (basically just Windows Play Anywhere games), a handheld future console/tablet on the same 1/4 hardware makes sense.

 

[MrFox]

However, if the SDK for XB2 is so designed for multiple resolutions from the off (basically just Windows Play Anywhere games), a handheld future console/tablet on the same 1/4 hardware makes sense.

Yeah it would be pretty easy, the same APIs, same OS, binary compatible. It would be just an additional power target to test, like midgen or PC games.

And add a dick mode for more power like switch did.

EDIT: Dock mode. I will leave the typo there, for art.

 

[Proelite]

Yeah it would be pretty easy, the same APIs, same OS, binary compatible. It would be just an additional power target to test, like midgen or PC games.

And add a dick mode for more power like switch did.

I wish more consoles had dick modes for more power.

 

[tongue_of_colicab]

Are we talking feasible in the sense of cramming the hardware in package that at least looks like it would be portable or are we talking feasible in the sense of a product that is ~300 dollars and people would actually buy?

I'm sure the former won't be that big a problem to make the hardware portable'ish if you look at the kind of hardware they can cram in thin laptops these days.

By the way, do we have any idea how hardware independent the software on this generation consoles is? I can imagine it would be easier to get AMD's latest designs to run at more acceptable power level in a mobile device than the older chips.

 

[HBRU]

I think about an ultra light portable PC / Console / Phone... not gonna be cheap actually... targeting both business and gamers

connectable to a monitor / tv, mouse, gamepad, keyboard, printers... probably they wait 5G to launch...

 

[DmitryKo]

Microsoft, with its UWP initiative, would most likely want a device that can play run binaries of released X1 games ...

This part makes no sence.

Full-featured Xbox SDK games are not built using UWP - what would be the point of refactoring for the UWP platform if the supposed handheld Xbox uses the same spec hardware?

And of course none of these games were designed for a small screen handheld form factor - so it would be like a completely new release on a different platform.

IMHO if Microsoft really wanted to make a handheld Xbox, it would either leverage Windows for ARM64 platform with a Snapdragon 8CX processor, or use a ultra-mobile x64 APU from AMD. So there would be a lot less processing power available in comparison to Xbox One.

As for existing UWP games, they would support ARM/ARM64 with a mouse click in Visual Studio. However most of them were actually designed to run on ARM processors with D3D11 FL 9_3 class GPU typical for Windows 8.1 Mobile phones, which would be about several orders of magnitude slower than Xbox One.

 

[Deleted member 13524]

Mobile XBone would already be possible on 16nm IMO, for a certain size and cost.
E.g. a Surface Pro sized handheld (say around 5mm thicker) could be done, on a SoC with around 20W TDP. Of course using the Switch form factor wouldn't be possible though.

Some people keep using the power consumption at the wall of the Xbone S and the PS4 Slim as an indicator that a mobile version wouldn't be possible.
- Power consumption on an AC device that plugs at the wall will always be far from something that uses DC batteries. More power is required for current conversion. Android STBs usually pull up to 10W at the wall using the same chips that go into 5" smartphones, and these last aren't blowing up every day.
- A handheld would definitely not use a mechanical disk. If the point is to achieve performance on par with a HDD, they could use low-power eMMC or UFS that sip power.
- The XboneS' SoC and PCB wasn't developed with power optimization in mind. It was made for low-cost production so they could sell the console for $250 (eventually $200). There really isn't any way to know how a power-optimized version of the XBoneS' SoC would behave. Looking at the 15W Raven Ridge APUs in 13" laptops is probably much closer to what a power-optimized XBone SoC would be than XBoneS'.
- XBoneS uses 16 (sixteen) GDDR3 2133MT/s chips for 256bit and 60GB/s. A 2018 handheld could use 2 (two) LPDDR4X stacks at 4266MT/s for a 128bit bus and 60GB/s.


Of course all of this would be expensive. I don't know if Microsoft could use 4 Ryzen cores w/ Hyperthreading at 1.75GHz to run the game VM. Maybe they could. The 12CU GPU at 853MHz isn't really an issue because the 2700U has 10CUs already and IIRC they run at about 850MHz on the (properly-cooled) 15W APU variants.
The oddest variable here are the 64MB eDRAM. Intel has been using eDRAM on 15W solutions for a while, although on a separate chip.




So to summarize, can a handheld capable of playing XBone games be made on 7nm? Yes, it can. I'm practically sure of that. And so could a PS4 Go.
But I don't know if making those would be profitable or not. Here's hoping one of those will see the light of day.
If not, SMACH-Z.

 

[DmitryKo]

HBRU, Project Andromeda is not intended for gamers - it's an ARM64 version of Microsoft Surface tablet designed around a novel folding display, and basically remains a business-oriented product.

Mobile XBone would already be possible on 16nm IMO, for a certain size and cost.
I don't know if making those would be profitable or not. Here's hoping one of those will see the light of day.

Before we discuss hardware, there should be a viable business case.

If that case is to run x86 UWP games from the Microsoft Store, these are mostly D3D9-class muliplatform mobile games which do not really require so much processing power, and would be perfectly fine with an ARM processor and graphics. It was just Microsoft trying to expand the reach of their dying Windows Mobile platform to the desktop and Xbox One. (BTW the Microsoft Store also offers desktop software, so not all of these game titles are UWP apps for ARM or x86).

And if the case is to run full-scale Xbox One titles unmodified, then mobile parts with such level of performance are not currently possible at viable price points (and these games would have to be tweaked for the handheld platform anyway.)

So they could just as well introduce a new handheld Xbox platform, based on the full Xbox SDK but targeting ARM64 hardware. Or just keep these Windows for ARM laptops (and the new Surface tablets) coming, and hope game developers would eventually start targeting ARM64.

 

[Pixel]

What is the smallest thinnest laptop with GPU around ~1.31 tflop?

Games are demanding and battery life would be an issue though. What is the point of portability if you want formfactor so small that it only gives you 25 minutes of game time when not hooked up via AC/DC.

So take that laptop and triple or quadruple or quintuple the battery size. Remove the BDrom add flash storage. Would need active cooling. Would be very expensive and alot bigger than the Switch, but nobody denies laptops are portable.

Biggest handheld ever.

 

[DmitryKo]

What is the smallest thinnest laptop with GPU around ~1.31 tflop?

Where can I find laptops with ~1.3 TFLOP graphics from AMD?

That would be either integrated Ryzen 7 2800H (1.8 TFLOPS), discrete R9 M280X, M385X, M470, RX M480 (~1.8-2 TFLOPS), or RX 460/560 (~2 TFLOPS). Unable to find any of the above in real products.

Might as well look for Kaby Lake G (Intel Core i5-8xxxG / i7-8xxxG):
Rx Vega M GL (20 CU 3+ TFLOPs)
Rx Vega M GH (32 CU 5+ TFLOPs)

 

[HBRU]

dont think Intel goes with ARM... https://a.msn.com/r/2/BBSSaMq?m=en-us&referrerID=InAppShare

 

[eastmen]

The Ryzen 7 2700u is 12-25w apu with Ryzen 4 core 8 thread cpu and 10 vega CU's up to 1300mhz .
https://www.anandtech.com/show/12709/the-acer-swift-3-sf31541-review-ryzen-meets-laptop/4

Performance seems to be pretty good actually. Its hard to compare strait to a console since the settings are so tweaked for a console. however Shadow of Mordor at 768 medium quality is at a 40fps average. Bioshock Infinite at 900p very high settings is 41fps. So it could replace the APU in the xbox one quite easily. MS would have to require patches perhaps. Its also a 14nm APU so a 7nm APU should get better performance and draw less power. IF MS paid for amd to make an 8 core jaguar paired up to a vega 10 at 7nm it would come in with even less power. Pair it up with a nice 1080p screen perhaps at 7 or 8 inches and it could be a really nice portable. At that screen size if a pure portable they could even target 900p with low level fsaa instead of 1080p would be hard to notice the difference.

 

[see colon]

I love the idea of this, but I don't think it's feasible. What media would the games come on? Semiconductor memory like Switch games? You might as well have custom hardware if you are requiring people to purchase new physical media. Digital only? That was essentially the idea that sunk XBone sales from the start, even though MS never actually did it. I'm doubtful they would go that route and risk a failure like that again. Custom cart and you can play all of your digital games as well? Maybe, but at that point you are requiring developers, or MS, to retool the digital games to fit on the new device. I don't think you can produce a handheld with a large enough pool of internal storage to just download existing XBone digital games without lower quality assets unless there is some revolutionary breakthrough in compression tech that is fast, lossless and can be handled by the low power chips powering the handheld. And if you are going to have to do work like that on every game, you are going to have to do QA and all of the other things required for any other shipping product, plus new manufacturing for the new physical media. At that point you might as well have custom hardware that can get close enough to your target graphically but with power efficiency in mind, make as many of the games network compatible and with compatible cloud saves, and have custom built, well tuned games available on the system. So, like the Vita. Only better.

TLR - Storage is the problem. Not because the technology to build a 1TB solid state handheld doesn't exist, but because it's too expensive. And 1TB isn't even enough anymore.

 

[Esrever]

I think they can make a very expensive mobile SOC at <10W with the same performance as the xbox one. 1 stack of HBM2/3 at ~600mhz, 4 zen cores at <2ghz, ~12CU GPU. Shrinking everything in the xbox one to 7nm will obviously not be low enough power. Using LPDDR4 will not get the memory bandwidth required and SRAM is probably a dead end and won't be cheaper than just integrating HBM.

 

[eastmen]

I love the idea of this, but I don't think it's feasible. What media would the games come on? Semiconductor memory like Switch games? You might as well have custom hardware if you are requiring people to purchase new physical media. Digital only? That was essentially the idea that sunk XBone sales from the start, even though MS never actually did it. I'm doubtful they would go that route and risk a failure like that again. Custom cart and you can play all of your digital games as well? Maybe, but at that point you are requiring developers, or MS, to retool the digital games to fit on the new device. I don't think you can produce a handheld with a large enough pool of internal storage to just download existing XBone digital games without lower quality assets unless there is some revolutionary breakthrough in compression tech that is fast, lossless and can be handled by the low power chips powering the handheld. And if you are going to have to do work like that on every game, you are going to have to do QA and all of the other things required for any other shipping product, plus new manufacturing for the new physical media. At that point you might as well have custom hardware that can get close enough to your target graphically but with power efficiency in mind, make as many of the games network compatible and with compatible cloud saves, and have custom built, well tuned games available on the system. So, like the Vita. Only better.

TLR - Storage is the problem. Not because the technology to build a 1TB solid state handheld doesn't exist, but because it's too expensive. And 1TB isn't even enough anymore.

DD would be fine. This isn't 2013 anymore and we are talking a portable device . Make the game library play anywhere so if buy it on xbox one , portable or pc you get it on all 3. Micro SD should be fast enough to replace the mechanical drive in the xbox one . Some are hitting 170MB/s reads which would be much much faster. Costs have also gone down. A 100MB/s read 256 gig micro sd card can be as cheap as $40 to the consumer. While expensive 400 gig cards can be had for as low as $80 and 512 gig cards dipping under $150 . I am sure by this holiday all the cards will halve in price. So that coupled with 256 gigs built in perhaps would be more than enough. Esp since the xbox one launched with only 512gigs