Analyse This: The Next Gen Consoles (Part 6)

So, as we head further into the year with worries surrounding the spread and control of the COVID-19 variant of coronavirus as a backdrop to the convoluted next generation console battle that is playing out between Microsoft and SONY. I mentioned previously that it looked like Microsoft were working towards an earlier release date than SONY and whilst that's still the case, potential (and almost guaranteed) supply shortages for components and final console builds are likely to play a part in both companies' plans.

However, time marches on and waits for no one... so Microsoft have issued a low-key article on their Xbox website that details pretty much all the "confirmed" leaked information that has been circulating over the last two months. So let's have a look at it....

There's nothing new here and there's a reason why this has been a low-key announcement and that's because everybody and their dog has had their "industry sources" confirm the majority of these details. It doesn't hurt MS to confirm them and it keeps the eye of industry and the consumer focussed on their SX console. This ties in quite well to my thoughts on their plans to release early (I'm still going with August) and allows them to focus on games at E3 instead of the hardware. In fact, I predict we'll have a big hardware reveal before E3, allowing the company to focus entirely on selling the concept of the console to the consumer instead of people spending valuable headlines and column inches (or whatever hellish name you wish to give to eternal-scrolling website designs) on the internal hardware - which, if we're to be honest, doesn't matter to the vast majority of consumers.

There is also the danger that E3 could be cancelled or reduced (or Microsoft's presence) in some way in the same way that SONY have cancelled attendances at GDC and PAX East. That sort of planned hardware reveal is probabalistically more likely due to the ongoing coronavirus scare*.

*Please note, I'm not saying the the coronavirus is not a thing to be worried about but the reactions of SONY are over the top compared to the actual potential for harm to their employees. I say this in my capacity as a scientist (though not an expert in this particular field). 

Now with more "X"s!

So what does this new blurt of information really tell us? Not a lot, really. They confirm the 12 TFLOPs of GPU processing power, the ability to use Variable Rate Shading (VRS) and hardware-accelerated raytracing (or, more accurately, path tracing). The SSD storage, "quick resume" for multiple games (though not detailing how many this can be up to), 120 fps potential and HDMI 2.1, as well as backwards compatibility are all either re-confirmed or confirmed from rumour.

Is there anything that is new? Yes, surprisingly!

First off the direct confirmation that the GPU is utilising a full RDNA 2 architecture - previously, it was rumoured to potentially mixed like the PS5's GPU is rumoured to be.

Secondly, MS outline how backwards compatibility will work. This essentially boils down to the limited and curated backwards compatibility available on the Xbox One and One X consoles, whereby you can insert the disc of your game of choice and the console downloads a "compatible" version of the game in order to run it. You can only do this for titles that have gone through the process of being enabled in this capability. What this means is that the SX is backwards compatible with the Xbox One series of consoles (that's not confusing at all, Microsoft!) without any further improvements in the library (though I assume they will keep working on adding titles).

Next, Microsoft reaffirms the mandate that all games released in the upcoming period will work across the XBO(X) and SX series of consoles - you will only need to buy a game once and it will work on all consoles from the original XBO to the new SX. This is a double-edged blade, in my opinion: on the one hand, it's good for the consumer in the sense that no one has to re-purchase a title for it to work on the newer console in six months time and it's good for the publishers in terms of having a large audience to target and sell to. 

However, this move does have associated negatives for developers and publishers. First off, it means that there's no (or little) impetus for consumers to move to the new console hardware. While we don't know the exact worldwide sales numbers of the Xbox One/S and One X consoles (since Microsoft stopped reporting them out of embarassment several years ago, instead only referring to console growth or percentage comparisons year-on-year), it's safe to say that the X hasn't achieved the same sell-through as the One/S did partly due to pricing differences with estimates back in 2018 looking at around 5 millions consoles sold in its first year. 

Comparatively speaking the Switch sold around 17 million units in its first year, the PS4 sold around 17.5-18 million units as well. Again, we don't know how many PS4 Pro models were sold after its release since SONY also engaged "shame mode" and didn't specify how many units were sold separately from the PS4 base model. However, early statements from SONY indicated slightly less than 1-in-5 sales were for the Pro. Let's assume that sell-through number across the whole first year of sales in order to not overestimate. That puts the first year of PS4 Pro sales at around 2.1 million units (averaging over the total for the year from Oct 1, 2016 to Sept 30 2017). That's not great...

Out of the total installed base of PS4 consoles, the Pro is a tiny fraction. So what reason do developers have to spend the extra resources targetting either the PS4 Pro? Yes, sure, the One X is a greater fraction of the total installed One userbase than the Pro but the One X exists in an environment where developers are targetting the One/S and base PS4 models. There's no incentive to spend vital resources on the newer, more performant hardware.

Faster is better...?

Microsoft's decision puts this into even clearer relief. Just as I predicted back in 2016, the mandate for the SX compatibility with the incredibly underpowered One/S will hurt games development because developers are no longer able to target a single performance envelope. Every console developer is now a PC developer and worse still, games cannot be optimised with an eye towards the new hardware - the promised improvements in game design cannot be implemented because they cannot feasibly be back-ported to the almost 8 years old hardware.

I'm not talking about texture resolutions, shaders, lighting, object geometry, rendered frames per second or screen resolution, I'm speaking of increased level size, SSD bandwidth and AI/computed simulation calculations.

In the same way that level design was hampered by the need to target the PS3 and Xbox 360, level design will be constrained by the need to target the XBO hardware. Yes, that means the 2020 equivalent of loading elevators or loading transitions or loading doors*. Furthermore, developers can't even rely on a given amount of loading time to enable suitable loading screens that can distract the player or impart useful information. For example, a large proportion of games use this down time to provide players with tips or hints... or even provide nice little vignettes which provide context for the player character's place in the world.

*Yes, I used the Mass Effect series for two of those, be thankful I didn't use it for all three examples! I can't count the number of hours I watched the Normandy flying through slipspace (or whatever it is called)...

We also cannot have games that could be created whose design could make use of the faster loading speeds. One common example Paul from RedGamingTech likes to use is a Superman game (though it could just as easily be the Flash or whatever counterpart he has in the Marvel universe) where the player is able to cover vast distances very quickly. This type of game mechanic could not be utilised or even implemented under this mandate. This was shown off in a now famous Spider man city-scrolling demo and to a lesser extent, some game loading times.

Getting to that last limitation, which developer is going to spend the resources creating a demanding world simulation with interacting elements and/or complex AI for the player to play against or interact with when you can't scale that back to work on the original XBO without severely compromising the systems? For years, developers and publishers have been complaining about the lack of CPU budget in the console hardware and now, right when developers would have a chance to get away from those limitations (if they so choose), they cannot.

Ironically, the reason the Pro and XBOX have not sold so well is because the cheaper option is just as viable and most consumers take the least expensive viable option. This would indicate that people would probably purchase the Xbox One X (if it came down in price) instead of a much more expensive SX which causes the traditional "chicken and egg" problem of console generations to be compounded.

Usually, the first games in a generational leap are released on multiple platforms to cover not only costs but to also build momentum for the new hardware. There are always exceptions that specifically target the new hardware, ignoring the larger install base of the prior generation which are able to more fully harness the features of the new console. Then as the install base of the new console increases, more and more games drop support for the prior generation. However, if consumers are not purchasing the newer, more expensive consoles because of stated support for the current consoles, then the sell rate will likely stall, meaning that publishers and developers will delay the switch over to the newer, more capable hardware.

This could also delay the reduction in cost of the new hardware as economies of scale will also take longer to kick in, meaning that the new hardware won't have price cuts as soon or potentially as often as the prior generation did, leading to further slowdown in growth of install base.

The second negative for publishers and developers from this move is that it is more expensive to target (and test for) multiple hardware revisions. This will increase development costs for games across the board - especially damaging for the indie and small/medium-sized developers. At the same time, they will no longer be getting consumers "double dipping" - whereby a consumer purhcases the game twice across the two generations. I'm not sure how big of a slice of the pie this can be (probably quite small) but it was a slice, nonetheless! That means less revenue for a particular title... VentureBeat outline a few good reasons why this move ultimately makes sense for Microsoft (and it's their business, after all!) but that's just the perspective from their point of view.

I pointed this out last time but I thought it would feature in the PS5...

Finally, I previously detailed how AMD's GPU architecture is able to do raytracing in order to provide more accurate audio simulation, even on current generation consoles. This is alluded to in the post from Microsoft wherein "realistic acoustics" are mentioned to be specifically enabled by the ray tracing technology in the GPU. This is the first instance I've seen where audio is marketed as a big thing for the SX whereas SONY have been promoting their custom audio capabilities in the PS5 since the beginning. 

The way this is spoken about indicates that Microsoft will be utilising the integrated raytracing capabilities of the graphical compute units in the RDNA 2 GPU on the APU - similar to the way it was described in the patent from AMD that was unearthed last year. Whereas SONY consistently lists its audio processing unit as a separate piece of silicon from the CPU and GPU (though on the same APU die). This might be where the constant rumours of the PS5's off-die or off-GPU raytracing capabilities come from - basically a misunderstanding of what's happening since any raytraced 3D audio capability on the PS5 seems to lie separately from the GPU.

In fact, I specifically said that it was entirely possible for Microsoft to enable raytraced audio through the driver stack as the hardware supports it and they've now announced that they've done that through DirectX.

Optional ray tracing for each compute unit in the RDNA GPU...

This brings me back to my previous thoughts about the parity of the two consoles. I posited that a weaker GPU (in terms of compute units) on the PS5 with dedicated ray tracing hardware could perform as strongly as a more powerful GPU on the SX where the resources were divided between various duties. Of course, this means that there can be scenarios were the SX can dedicate those resources more fully (or dynamically) to a specific task, in which case it would be able to perform better than the PS5 could, given the same workload that focussed on a singular performance envelope (e.g. no ray tracing or only ray-traced graphics).

It really does seem like the rumours that the consoles are basically identical in terms of capabilities were truer than we could ever have imagined. I think the big division between the consoles will be the games and that means, specifically, whether SONY will mandate forward compatibility across PS4, Pro and PS5 as Microsoft have done and how much backwards compatibility SONY have in the PS5. I'd love to be able to just put my PS4 game discs in the PS5 to play those games at better frame rates/consistency or higher resolution and I'd love it even more if my PS2 and PS3 games were also playable without issue. That would keep me and many other gamers in the Playstation ecosystem.

There is one, final, thought I had regarding this announcement and Microsoft's mandate for multiple console compatibility - Lockhart. People have been looking at Lockhart (assumed to be Xbox Series S) as the low-end next generation model. However, I believe that Lockhart is in development as a cheaper refresh of the One X. It must be getting difficult to source and/or lower the APU in Scarlett - not to mention the custom cooling solution they had to implement in order to make it work. Lockhart (or Series S) could not only be a cheaper alternative than producing the One X but it would have a more powerful CPU and comparable GPU, only it would have next gen features.

I also think that Lockhart could be using a lower binned Series X APU. The SX has such a large APU die size that their yields will suffer for it. Just as AMD, NVidia and Intel all use binning to increase yields and segment their products, it would make sense that Microsoft would try and do the same for the SX APU, fusing-off inoperative parts of the GPU portion of the APU and then selling it at a cheaper price for less performance.

Anyway, I'm sure we won't have long to wait for more tidbits from either company as I thought last time, it appears Microsoft is accelerating their schedule in order to get out ahead of SONY...