Next Generation Hardware Speculation with a Technical Spin [2018]

[DieH@rd]

Successor of PS4 was briefly mentioned in the Bloomberg article about John Kodera:


The other key question for Sony’s PlayStation business is the timing of a successor to the PlayStation 4, which is now five years old. Kodera declined to provide a timeline or specifics for a new console, saying only that a new device is being designed around making it easy for developers to create games and also with an eye on easier hardware manufacturing.

“We’ll be leveraging the knowledge and experience of having built four generations of consoles,” Kodera said.

Kodera’s comments came a day after Sony issued conservative earnings targets across most of its businesses, including a drop in PlayStation operating profits by March 2021 from the current fiscal year. Asked if the decline will be because of higher marketing costs and hardware subsidies related to the introduction of a PlayStation 5, Kodera declined to comment. Console makers usually see shrinking profit when new consoles debut, suggesting that Sony may be planning to roll out the PS5 within three years.
https://www.bloomberg.com/news/arti...tation-chief-hints-at-portable-gaming-revival

 

[goonergaz]

Let's be honest, Sony in the home console space is like Nintendo in the handheld space. Only a complete disaster (PS3 - which dispite launching later ended up selling about the same as it's competition) has stopped Sony completely dominating the 'standard home console' space.

MS know they have to up all aspects which is why they've been working so hard this gen (like Sony last) to undo all the bad.

 

[Lalaland]

Yeah the two sku strategy previously failed because the differential wasn't significant enough (PS3 20GB) or actively made the product worse (X360 no hdd edition). Going for actual substantial differences imposes massive production costs for example the easiest way to maximise manufacturing efficiencies is to do what Intel does with it's CPUs and "bin them" or sort them for perf and then fuse off parts of the CPU to produce an i7/i5 or what have you but that doesn't work here. Sony and MS are selling entire systems so they also need to manufacture m/b and psus and cooling sub systems etc, the way to maximise your economies of scale is to make the same m/b and everything else regardless of target platform but now my low end sku is paying for parts not necessary for it and you run the risk of being too expensive on the low end.

Let's say for example in this scenario I make the rest of the box for $200, now I speak to my suppliers and they tell me they can make a cheaper version for $150 for the low end sku, that sounds great but it makes the whole affair far more risky and expensive. Sure my low end system now costs $150 but the high end costs jump to $250 or $300 depending on how big an impact on volume the low cost system has and this gets to the heart of the issue with a 2 sku strategy, forecasting. Forecasting is the great bane of any sales persons life as you are being asked to guess the unknowable and there are big risks off the back of it, I've seen $500 worth of kit being flogged for $250 as they overegged forecast and we were left with 2,000 PCs stinking up a warehouse. Stock costs so goddamn much I'm not sure I can emphasise this enough other than to point out for most manufacturing these days your warehouse is your trucking fleet, a single sku strategy offers the most manufacturing savings and least complexity so Sony or MS would have to have some very solid numbers on high end demand to risk it and that's before we get into the complexities of marketing it.

 

[anexanhume]

Surely it's not that simple, power usage goes up temperature goes up. Now the more exspensive SKU needs a different power supply and better cooling possibly a bigger console. What if they at the limit of an APU and the only way for more TFlops is to go with discrete parts? That means different motherboard.

Remember PS Pro has twice the TFlops of the base PS4 and some people feel it's not enough. Twice the TFlops of the next gen machines would be what 22-28TFlops. I just don't think they can do that with an extra $100 dollars. Maybe an extra 4 TFlops at Max and does that make much of a difference?

You literally truncated my post so you could make arguments against it with things I mention further down in the post. You also magically changed your price range from $100-200 to just $100. Why?

 

[Xbat]

You literally truncated my post so you could make arguments against it with things I mention further down in the post. You also magically changed your price range from $100-200 to just $100. Why?

Well you said it would absolutely be more powerful if they spend $100-$200 more but you didn't actually say how or how much more powerful other than they can make the chip bigger. How much bigger can they make the chip and how much does everything else they have to do eat into the extra $100-$200.
So how much more TFlops or RAM can they put in for the extra $100-$200?

Yes you did mention the other factors in your post but I feel those factors aren't as simple to just sort out especially during the launch period. So I just reiterated them and there was nothing magical of me suddenly using $100 instead of the $100-$200 range it's just that I forgot.

Thinking about it though a $600 console will be dead on arrival though especially if it's what only 4 more TFlops.

 

[Lalaland]

I wonder if a storage medium has ever been as impactful in a gen as it was during the PS1 and PS2 eras? BD didn't confer much advantage to Sony during the PS3 era and with the internet growing in importance as a distribution method will it ever be such again?

I mean BD100 might be useful in this era of 50GB game + 20GB "patch" d/l but would a BD50 + b/w cost be cheaper, does anyone have any idea of bulk mastering costs of BD100 vs BD50?

 

[Shifty Geezer]

As software sales tend to be both companies bread and butter, it's in both there interests to entice me to buy on there platform. Having the best quality 3rd party titles by having the most powerful hardware is one way to entice. The other is an excellent software lineup only available on that platform.

Not everyone is willing to pay an extra $100 for a marginal improvement in visuals. In fact, far from it when we look at PS2. You cannot be more powerful at the same price point unless you manage to do something magical. All these companies are working with the same tech and the same laws of physics. In making a choice here and now for the whole of the next generation, if you set out to be the most powerful, the only way you can assure that is to go all out, which means mega-power just in case your rival is trying the same most power strategy, and that means an expensive box or a lot of losses. If you instead set out to make an affordable box that provides a great gaming experience and compete on branding, exclusives, etc. you still have a very viable strategy with the risks.

The choice of hardware cannot be made based on a fear of what the rivals are doing. It has to be sensible, balanced, and targeting your market. Mainstream sales requires a mainstream price and so mainstream hardware.

 

[mrcorbo]

Purely playing Devil's Advocate here, but it could be argued that if MS were making decisions in fear of what Sony was doing when designing the XBOne it might have saved them some pain.

 

[Shifty Geezer]

Maybe they did. Maybe they looked at their spy reports at Sony creating just a console, and their data on people's media consumption, and the success of Kinect, and that's what caused them to go the way they did. Or maybe they didn't, and just executed their plan without knowledge of Sony. Maybe they caught wind of PS4Pro and rushed to get a more powerful mid-gen console and the end result maybe lost them crazy oney - we've no idea of the costs to design XB1X, but it's likely as much as creating a new console but it'll make a tiny fraction of the money of a new console.

The timelines required for these products means you can't be reactionary, at least not without risking considerable costs.

 

[Magnum_Force]

Not everyone is willing to pay an extra $100 for a marginal improvement in visuals. In fact, far from it when we look at PS2. You cannot be more powerful at the same price point unless you manage to do something magical. All these companies are working with the same tech and the same laws of physics. In making a choice here and now for the whole of the next generation, if you set out to be the most powerful, the only way you can assure that is to go all out, which means mega-power just in case your rival is trying the same most power strategy, and that means an expensive box or a lot of losses. If you instead set out to make an affordable box that provides a great gaming experience and compete on branding, exclusives, etc. you still have a very viable strategy with the risks.

The choice of hardware cannot be made based on a fear of what the rivals are doing. It has to be sensible, balanced, and targeting your market. Mainstream sales requires a mainstream price and so mainstream hardware.

I've got a feeling I'm not explaining myself clearly - I'm not arguing Sony/Ms should go for a high power monster out of reach of Joe consumer instead of going for a balanced system, I'm suggesting they should do both - a balanced system that fits the bill and what most people would normally expect as well as a top tier product.

If I'm reading people arguments against a 2 tier strategy correctly, the seem to be the following:

• A 2 tier launch would split and confuse the customer base
• You couldn't do a 2 tier launch without sacrificing the performance/price of the lower tier.

I believe the first argument is overstated. We have a 2 tier system at present, it just wasn't a simultaneous launch. We have other devices in the consumer space which have multiple tiers (TVs, Mobile Phones, Cars, dishwashers, washing machines) and Joe consumer seems to manage just fine, especially in this day and age where a review is a Google search away.

The second point is harder to gauge, but if this was planned from the start it would likely be less of an impact and there are a few ways a 2 tier design could be done. A salvage chip for the lower sku, more cut back than what we are used to, and lower clocked, or a separate chip. If they went with discrete cpu & gpu a separate gpu chip might make more sense for the separate tiers. The extra components required for the top tier (stronger power delivery/cooling) would factor into its cost.

 

[Lalaland]

I still think 2 tier is unlikely but I think you're discounting the 2nd tier arriving 4 years after launch too much, that 2nd sku was to harvest margin from people already on the platform it wasn't a mass message to the market on launch. Look at the intensity of the Console Wars as we get close to the launch would any company want to invite that vitriol inside it's own tent?

The second point that I was striving at is that if you don't saddle the low end with unnecessary cost (by standardising on all the parts for the top end SKU) you sacrifice economies of scale for your high end part resulting in your extra $100 or $200 having to be split between shiny go fast features and boring stuff like a new m/b, psu and cooling all things that are necessary for a higher performing part but don't contribute to it's performance thus resulting in a narrowing of the difference between the base and top end SKUs.

This go around will be fun though now that the second SKU is a thing, will it be 2 SKU at launch for either vendor on top of a mid life refresh (for 3 total SKUs)? Or will it be single SKU + midlife like last go around?

 

[Magnum_Force]

You could have a 3rd tier 4 years later .....

Joking of course (or am I...)

In seriousness, I'm not sure we'll see the mid life refresh this time around. The midlife refresh coincided with the move to 16nm. This time around I'm not sure Sony or MS will be willing to bet on 5nm being both ready and cheap enough to be viable.

 

[Shifty Geezer]

If I'm reading people arguments against a 2 tier strategy correctly, the seem to be the following:

• A 2 tier launch would split and confuse the customer base
• You couldn't do a 2 tier launch without sacrificing the performance/price of the lower tier.

Those aren't my arguments. I don't think consumers will be confused, and you needn't sacrifice anything on the low end. The problem is that the cost to create the second tier machine is similar to the cost to create the first machine if it's radically better. You have increased stock problems (how much of which SKU do you make?), and you won't sell to more people than you would without the high-end SKU, so you ending spending to make two different machines to sell to the same number of people. You end up spending more money to make less money, so the reason to go that route has to have far better value to your business somewhere else, like brand identity, which it probably doesn't.

The only solutions I see are either

1. An 'easy win' high level SKU, such as using binned parts, which costs very little to do (in the case of binned parts, it means more profit per silicon) and allows a higher profit margin on the top end - expect the hardcore to pay 20+% more for 10% more BOM cost. Like a car with a bunch of overpriced extras.
   
2. The normal console and a cheap-ass option instead of a premium option, such as a streaming box, which will expand the market over the normal console. That is, if PS5 costs $400 at launch, it'll be limited in sales to those willing to pay that. Add a $150 streaming box and you'll sell to a different audience at the same time, so don't interfere with your primary market.

 

[Magnum_Force]

Another option for the console makers if they go for a 2 tier launch is a less powerful "normal" model at a lower price point to what we are used to - say an ~8TF, 16GB Ram model at £329 or lower. That would allow a bigger gap between the top tier and the normal console, with possibly x2 the tflops.

Normal console @ £329:
4c/8t Zen at 3ghz
~8Tflops (3072 Alu x ~1300mhz)
16GB Ram
120GB ssd
1TB hdd

Top tier @ £499
4c/8t Zen at 3.8ghz
~16Tflops (4096 Alu x 1950mhz)
16GB - 24GB Ram
120GB ssd
2TB hdd

The top tier would need much better cooling and power delivery, but the silicon could be the same potentially.

 

[chris1515]

https://twitter.com/x/status/1037893489156083713

Raytracing demo on Xbox One X 100 sample per pixel.

 

[Tkumpathenurple]

https://twitter.com/x/status/1037893489156083713

Raytracing demo on Xbox One X 100 sample per pixel.

Is that in real-time?

If so, that bodes well for the next generation having some level of ray tracing, even if lacking bespoke hardware.

 

[Shifty Geezer]

1080p == 2 million pixels * 100 rays per pixel == 200,000,000 rays per frame
30 fps would be 6 giga rays per second, using the new Bogomips, which seems a lot, but then it's only tracing spheres and they're the optimal shape for RT - you don't need any geometry tests so no scene space traversal as required with poly meshes. So I expect realtime, but not particularly indicative of anything. Perhaps suggests RT'd shadows using CSG primitives might work well where there is no HW acceleration for mesh testing.

 

[vipa899]

Even ps3 should be able to do some form of ray tracing? All hardware can be classified as RT-ready like this

https://m.slashdot.org/story/82859

 

[Shifty Geezer]

That's one ray per pixel excluding bounces - no soft shadows or any fancy materials. You get up to three rays per pixel with recursive reflections*. At 1080p (I think it's 720p) that's 2 million rays per frame, at some 2 frames per second. 4 million rays per second. Across 3 PS3s.

It's about 1/1,000th the speed we're talking about in RTX and ray tracing hardware. Not really sure why you brought it up. I think the other PS3 demos are faster anyway.

* Not sure how the 'rays per pixel' is counted. Should be two per point because of one ray to the surface and one ray to the light?

 

[chris1515]

1080p == 2 million pixels * 100 rays per pixel == 200,000,000 rays per frame
30 fps would be 6 giga rays per second, using the new Bogomips, which seems a lot, but then it's only tracing spheres and they're the optimal shape for RT - you don't need any geometry tests so no scene space traversal as required with poly meshes. So I expect RT, but not particularly indicative of anything. Perhaps suggests RT'd shadows using CSG primitives might work well where there is no HW acceleration for mesh testing.

From what I seen on twitter many devs thinks next generation shadowing will be raytracing. And it is a good things I hate shadew maps problems.