• Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.

(SONY/MICROSOFT Hardware & Services) We WILL Get Mid-Gen Refreshes, But Not What You Might Think...

What type of mid-gen refreshes do you think we'll get this time (select all that apply)?

  • Pro-level models

    Votes: 137 46.8%
  • Slim models

    Votes: 166 56.7%
  • Service-orientated expansions and launches

    Votes: 30 10.2%
  • Nothing; same PS5/XSS/XSX hardware all gen

    Votes: 37 12.6%

  • Total voters
    293
  • Poll closed .

RoadHazard

Gold Member
Holy shit that's a lot of text, Jeff Rigby.

But why would you need 2TF for a streaming box? All it should really need is a decent hardware video decoder, the UI etc can run perfectly fine on weak mobile hardware.
 
It's not really me ignoring the component scarcity issues, but more so being optimistic about certain other developments. Sony have plans to open up a new fab plant for example, and more smartphone manufacturers are trying to eat up 3nm (and apparently, 4nm) wafers which could free up some 6nm and 5nm space.

But let's say Sony & MS stay on 7nm anyway for those scarcity reasons, fair enough. Maybe they just split some allocations between 7nm and 6nm, since the only benefit with 6nm is cheaper costs really.

Sadly it's not that clear cut anymore. The cost of porting designs to radically dissimilar process nodes today is obscene, and given N5/6 are EUV while N7P isn't the incompatible libraries means these two processes are radically different.

They would be willing to eat the costs if there was any promise that N6 might provide them with the volumes they need, but given the greater competition on the more bleeding edge node, it's highly unlikely to be the case until TSMC's other customers have moved on to high-performance N3 and N5 nodes.

And Sony doesn't fab chips for their consoles. They sold off all their fabs except their optical sensor fabs, a long time ago. So Sony opening up a new fab doesn't help them at all.

They're at the mercy of TSMC and to a lesser extent Samsung, whenever their 7nm process ever catches up to TSMC's N7P node density. GloFo is a joke. TSMC has plans to build a huge fab in Arizona in the US, but it will be 2024 before they're online and ramped up to volume production.

The bigger issues with component scarcity encompass more than just the wafers and chips from the fabs. The bigger problem is RAM, PCBs and substrates. There's no clear solution on the horizon to address those.

Also just want to mention real quick that "mid-gen refresh" the way I've used it isn't meant to imply Pro-level systems, because in that case I do agree we will likely not get them this gen. But moreso thinking, if they keep the same Zen 2 CPU designs but switch to RDNA 3 or 4 (whichever would be the more recent) while maintaining the same TF, pixel fillrate, texture fillrate, geometry rate as the original systems but just implementing an RDNA 3/4 design with some hardware acceleration for certain new features. Certain architectural changes in the GPU alone could help with getting decent power consumption reductions while still keeping the same physical footprint for the chip, and you can get a bit of extra performance potential on top of that as well.

Architectural updates won't net you any meaningful performances advantages for a mid-gen refresh. I think you're forgetting how woefully neglected the specific hardware features on the PS4 Pro were. The mid-gen consoles are designed to play the same console games as the base consoles exclusively. The idea is that devs build the same game binary and it runs on both boxes, meaning any new microarchitectural updates on the mid-gen box GPU will be invisible to the game by definition; since the games will run exclusively in effectively BC mode.

In order to take advantage of any new microarchitectural feature in hypothetical mid-gen boxes, devs will have to code for them specifically, which defeats the entire design objective of the mid-gen console.
 

Sega Orphan

Banned
That method makes 0 sense when the biggest problems with the current consoles is poor RT and ML performance compared with PC, and Bandwidth limits. The next power consoles need to be on a new arch with Gddr7
That will come with the full next gen reset with PS6 etc. Mid gen will keep existing tech just boosted, if there is one.
 

Louay

Member
I have to say OP i missed your Booty profile picture back in 2019/early 2020. the time when fake leaks spreading left and right about next gen Specs.
 

Lupin25

Member
Special/anniversary editions. Yes… maybe even slim models, but I’d even doubt that with chip shortages.

No one has seen one in a store, but we’re already thinking about pro models?
Regardless of we would pay as high-tide enthusiasts, we should probably consider the fact that casuals very likely won’t keep up if Sony hasn’t either.

That’s unless this gen ends up becoming longer than anticipated.

The PS5 has been out for over a year, and we’ll be more than halfway through its lifecycle before shortages begin recouping. I doubt we’ll get them.
 
Last edited:
No one is going to increase die cost by using IC and then use more expensive gddr7 memory and make it run at gddr6 speeds to save power use. Why would you not just use the extra bandwidth of gddr7 instead of dicking around with a much too small amount of IC to be much use above 1080p?

Don't even the smaller IC-supplied RDNA2 GPUs have a notable benefit to them in terms of overall performance? IIRC IC isn't just about cache capacity but also cache bandwidth and enforcing better data locality to cut down on hits across the memory bus.

A small block of IC wouldn't increase the cost too much; you might be right about GDDR7 not matching that design as well, but if that's the case they could just stick with GDDR6 and you probably have an overall setup using less power than replacing with GDDR7 modules running higher clocks. Plus, you can better control supply needs and pricing since you don't have to gamble with if there'll be enough of the necessary RAM modules at those high-perf clocks for your system.

What power increase would you need to vindicate a mid gen pro model? I will use Sony here as I think their situation is more easy to guess. Similarly to the pro over base PS4 Sony will most likely go for the butterfly GPU design of two sets of 36CU giving them 72CUs, possibly with a slight speed bump and keeping the variable frequency. That would be around 22tflops, and a slight increase in CPU clocks to around 4.0ghz. A 2TB SSD of the same speed, with the same amount of RAM but at a higher bandwidth and an extra two gig of side RAM for the OS.
The question is will the tech be ready for that in a couple of years time?

Well if we're just looking at raw power, it's already here. 6900 XT provides ~ 23 TF of perf on an 80 CU spec; there are 21 Gbps GDDR6X modules available now and faster GDDR7 ones set to come by the end of the year and into early 2023, etc.

But the bigger question might be if they even do a mid-gen refresh focusing on massive increased power; IMO I don't think they will, but keeping the same general power perf profile of the PS5 on an RDNA 3 or 4 GPU spec would bring new features that can bring some helpful performance benefits while still staying in current power consumption targets or (most likely) lower, thanks to architectural improvements as well as a potential smaller node, using more power-efficient internal NAND storage, improved packaging designs or some combination of all of these things.

I do wonder if Sony is looking at yields and node sizes in light of the chip shortage and considering a slim as a higher priority than a pro.

My thinking is that they could crank availability if they kept the die size the same but shrank the node.

It's definitely one option; anything they could do to make a PS5 slim that also reduces a lot of power consumption/heat (meaning less complex cooling = cheaper BOMs) also means increasing profit margins on the hardware, and with demand for current PS5s not being met they'd probably rather want to do their best to satisfy *that* demand while maximizing profit margins before creating even MORE demand via a Pro model that requires splintering of resources.

Following mainstream CPU and GPUs will ensure that there is a continued shortage.

Staying on 7nm will ease up supply as AMD moves to 5nm.

It will for sure, but they can also spread out more supply to 6nm if need be, to increase overall supply.

To run console games natively on mobile they should be PS1/PSP/PS2 via emulators, or to run PS4/PS5 games without porting required on a mobile hardware that should be 1:1 PS4/PS5 hardware and as of now doesn't exit.

Not quite yet, sure. But in another year or two? I think it will be possible. Steam Deck is already getting towards that level of performance, and if it compromised on a few things such as RAM capacity and storage they maybe could've shifted that budget to a higher-performance GPU.

One of the few caveats for a PS portable providing 1:1 PS4 base perf is it'd need an 8-core CPU with 8 threads. There are some Zen 2 mobile designs with that, but not many.

They won't make a device that needs dedicated games and development teams. Remember they also have to support PS5 and PSVR2, and that will also release games for PC and mobile.

True; my idea for a PS portable is that it basically provides a big boon for the brand in Japan and other Asian countries, and the appeal would be providing the entire PS4 library "on the go" with its portability. As a portable design they can go for larger profit margins on the hardware versus just manufacturing more regular PS4s, and it could provide a legacy platform for smaller devs to target with software that also is 100% compatible with PS5.

If they want to do PC ports, they just follow the same path they currently do.

The only more or less option I see here would be to make their own PC handheld/hybrid or embrace Steamdeck or a similar one. It would run PC games, not PS4, PS5 or dedicated games. I'm pretty sure that once Steamdeck starts to get a good install base we'll see many other companies releasing their own PC handheld/hybrid, not only the existing Aya and similar. I'm thinking about companies like Razer, Asus, Samsung and similar. Plus I'm also pretty sure that Apple will do the same but in their case using iOS instead of Windows or SteamOs.

For better vertical integration and to have an additional revenue stream off hardware they would preferably make their own portable, but it doesn't have to be a PC-style device as in using general PC components. Again, they can just adapt the base PS4 design to a smaller node process; 18 CUs would already be half the count of the Pro and PS5, and GCN CUs on 16nm are 65% smaller than RDNA2 CUs on 7nm. So an 18 CU GCN design on 7nm (or maybe even 5nm) would bring a much smaller physical footprint, much lower power consumption (than a base PS4) and still provide a 1:1 perf target.

They could embrace the Steam Deck, but at that point they might as well just keep doing what they're already doing and it'd render the point of a PS portable (especially in terms of what it could do for markets like Japan) nulled. For this to work, IMO Sony has to provide the hardware themselves and control the integration process. Again, its main point is just to provide a seamless "on the go" portable option for digital PS4 games, no porting required. And since it's using the PS4 architecture (with some modern QoL benefits, i.e microSD card for storage instead of platter HDDs), devs who plan to stick with cross-gen for the longer haul anyway will have a reinvigorated PS4 market they can sell to and those games will seamlessly work on console PS4s, PS4 Pros, and PS5s as well.

They won't use RDNA3 or 4 for a mid gen refresh. It would be too expensive, too different and in a context where pretty likely they will still be handling chips shortage issues, for consoles I think RDNA 3 or 4 would make sense for the next gen(PS6, I think MS won't make another console but instead will continue moving their focus from being a console platform holder to a full multiplatform -including rival consoles- 3rd party publisher), not for a mid gen refresh. I think smartphone companies will slow don the release of new models because of the chips shortages.

RDNA 3 or 4 for 10th-gen? I dunno; both of those will be old hat by then (personally don't see 10th-gen happening until 2028), so they'd be using a newer architecture.

As for them being too much for a mid-gen refresh, well remember the PS4 Pro and Xbox One X used then-recent architectures instead of the outdated GCN 1.0 of the base systems. The market factors at play MIGHT make it more expensive to go with 3 or 4, true, but it's no big loss if that's the case. It'd just mean them missing out on newer features. To be fair, I think my idea of mid-gen refreshes might be different than most: I'm not thinking of systems with 2x/3x the raw power of the current systems, but just supporting newer architecture features while maintaining all the same key perf metrics (TF, pixel fillrate, texture fillrate, geometry culling rate, etc.).

That thought on Microsoft is definitely interesting...there's a decent possibility it happens, although it might be a loaded idea for some to entertain ATM. I still think they'll do something for 10th-gen, but hardware won't be their priority focus and just an option (also likely one they position as being sold at a solid profit margin from Day 1 like a PC, so a much higher MSRP than typical of a console).

Japan is a small portion of the global gaming market for consoles, they won't remove hardware, marketing, 3rd party moneyhatting and gamedev resources of PS5, PSVR2, PC and mobile to make a handheld that would require dedicated games. Doesn't make sense.

That's just it, though; this ISN'T a portable for dedicated games! Its selling point would be providing the entire PS4 library on-the-go for portable gaming, and for devs who are still focusing on cross-gen, could give their PS4-tier crossgen games a larger market to sell to.

That's the real purpose of the device IMO, so it's not like teams are splitting up resources, especially Sony 1P ones.

If something, they will release more mobile games, which is by far the biggest gaming market on Japan. Or to push harder mobile related services like cloud gaming, remote play and so on. But not only because of Japan, but because they are also big (bigger than console or PC) in the rest of the world.

What they need to sell in Japan, like in the rest of the world, is to get more chips to be able to produce an ship more PS4s and PS5s (and in the near future PSVR2). An issue that doesn't depend on them. This issue would be even worse with an additional device.

They can def do that, but with their own portable they can also have their mobile teams incentivized to develop PS4 versions of those games that then also seamlessly play on a new portable device with base PS4 perf levels. So that game now has an even larger market to appeal to, for more sales and revenue. By this point those still on the PS4 hardware would mainly be more casual types, meanwhile a new portable with that level of specs could get hardcore and core gamers on PS5 to invest into it if it provides a way to natively play digital PS4 games on the go (and stream PS5 games on it).

I don't think getting more PS5s in places like Japan specifically necessarily translates to more revenue from that region, because apparently actual software sales for PS5 there haven't been very good. Possibly, more of the types of games needed to appeal to that market could be served by providing a portable for locations like that which target a lower performance threshold, but enough of one to still be fully compatible with PS4 systems already in those regions, so this could create a space for smaller or more niche teams with a focus on specialized content for those regions to provide that content on.

I think with PCIe 6.0 we might just have a half-assed next gen instead. But doubt it. PS5 Pro? Likely. Xbox Series XXX? Can't see it with S around.

Also, do you get text orgasm? Love your effort here, really, but...

Robot Didnt Read GIF


PS: Before you get upset, see what happened when tried to post quoting you:


OK ngl this made me laugh pretty hard. Also: the pain of running into the text limit and having to parse over everything to fit it in the OP :S

I don't know how PCIe 6.0 necessarily influences a path for 10th-gen; unless they add a ton of CXL features to 6.0 then there's little 6.0 will have that hurts 10th-gen systems performance-wise IMHO.

Also yes I do get text orgasms from large passages. Best orgasms ever.
 

hlm666

Member
Don't even the smaller IC-supplied RDNA2 GPUs have a notable benefit to them in terms of overall performance? IIRC IC isn't just about cache capacity but also cache bandwidth and enforcing better data locality to cut down on hits across the memory bus.

A small block of IC wouldn't increase the cost too much; you might be right about GDDR7 not matching that design as well, but if that's the case they could just stick with GDDR6 and you probably have an overall setup using less power than replacing with GDDR7 modules running higher clocks. Plus, you can better control supply needs and pricing since you don't have to gamble with if there'll be enough of the necessary RAM modules at those high-perf clocks for your system.
When the cache gets too small the hit rate reduces and it starts to get less effective than straight up better memory bandwidth. It's visible in most 6000 gpu benchmarks vs 3000 gpus as the resolution increases. The simple view on this is look at 6600 vs 2060 at 1080p and 1440p, the 32mb IC does ok at 1080p but 1440p it isn't enough and the 2060's increased real bandwidth pulls it ahead. Now if we think about these systems coming out in 2-3 years and games get better visually as expected that small IC cache looks pointless unless everything is 1080p I guess.

RT is also heavy on bandwidth and it's probably why AMD sponsored titles only go for shadows because they are more coherent and probably have a better cache hit rate than other effects, while nvidia chose to go with pushing gddr6 to gddr6x for more straight up bandwidth. We will get the answers to that this year though.

Here's what Infinity cache does when you take the clock speed advantage away from the 6700xt and bench it against a 5700xt without any. I honestly think there are better uses for the die space in your scenario.

 

Sega Orphan

Banned
Well if we're just looking at raw power, it's already here. 6900 XT provides ~ 23 TF of perf on an 80 CU spec; there are 21 Gbps GDDR6X modules available now and faster GDDR7 ones set to come by the end of the year and into early 2023, etc.

But the bigger question might be if they even do a mid-gen refresh focusing on massive increased power; IMO I don't think they will, but keeping the same general power perf profile of the PS5 on an RDNA 3 or 4 GPU spec would bring new features that can bring some helpful performance benefits while still staying in current power consumption targets or (most likely) lower, thanks to architectural improvements as well as a potential smaller node, using more power-efficient internal NAND storage, improved packaging designs or some combination of all of these things.
What I was meaning is when would the tech be ready to have that sort of power in a console form. Like, if they use 5nm fab, could they double the GPU size and still keep a Chip at the same size and heat as the current one in the PS5?
Heat and Price are the two biggest limiting factors for a console.
The XSX, and especially the XSS, were quite conservative clock wise for RDNA chips, and I though that with the XSX and XSS being so quiet that MS might have tried to get a few lazy GHz extra from them prior to launch. Fuck, I still think they should put a patch put to increase the XSS GPU speed to around 1800-1900, as I think the extra few flops would help and shouldn't brick the console or make it too noisy.
 

rofif

Gold Member
PRO models can suck a dick.
Smaller and refined models organically ? sure.
Artificial 20% faster pro models which segregate the market and add even more options to new games which should not exist on console? f that
edit... and I will buy it if it releases. fml
 
Last edited:
- Changing architecture mid-gen is disruptive. most certainly the same RDN2+.
- Higher CPU clock isn't disruptive. possible for a refresh, lily for a Pro model.
- a refresh needs to have the same GPU clock for compatibility. PlayStation 4/Pro/5 compatibility is hardware dependent and not software dependent. the Xbox uses a higher layer of API that makes multi-hardware compatibility easy. Sony need to map the hardware/API years in advance and commit to it.
- A pro would most likely just double the PS5 CU count and result in 20TFLOPS..

CPU architecture change mid-gen is disruptive, yes, but GPU I'm not as sure. PS4 Pro and One X both changed their GPU architectures going to Polaris from GCN 1.0 (or was it 2.0?). Future RDNA designs should still be pretty backwards-compatible with earlier ones, and shifting the GPU design to a more recent one could be cheaper than sticking with the older RDNA2 (AMD might incentivize them to switch to newer RDNA).

Your points are valid especially WRT to a Pro-like mid-gen refresh, but personally I don't see that happening this gen. But if circumstances permit, I don't see straight-up normal "slim" models happening, either. Something closer to a cross between the two, think something like later 3DS refreshes that upgraded certain specifications.

Umm.. you can't just add hardware RT without changing the chip. And that will question why they couldn't do that in the first place. I see nothing being added to the current hardware at all. In fact, I only see various flavors like "slim" or perhaps larger HDD sizes.

It could certainly end up going that way. But insofar as hardware RT, ML etc., this type of refresh IMO would be more like keeping the same TF perf but adding in additional hardware silicon for those RT features and whatnot.

Well given that scarcity of certain parts is going to extend to 2024 and maybe beyond, it would make sense that they'd redesign the hardware to be as simple as possible and cut down on how many chips each console needs. I have no idea how practical it would be to do that

For PS5 at least one approach would be to simplify the I/O complex even further, and redesign it to default to M.2 SSD storage devices with a much smaller onboard NAND module for storing certain critical OS files and a restore image.

Since the internal and expansion SSD can't simultaneously use the I/O as-is anyway, why not just make the internal drive a specific-capacity drive in M.2 format that can then be swapped up with bigger (and potentially faster) drives if the user wishes?

I don't know what the basis for these speculations is, but I don't think Sony or Microsoft are ever going to launch mid-gen consoles that only have 30% higher FLOPs throughput than the original models.
Especially considering the fact that RDNA3 is expected to bring a massive increase in throughput with twice the compute shaders per WGP compared to RDNA2.

Well if they're going to make Pro-level refreshes then yes I'd agree. However IMO they aren't going to do that this generation.

Wait, I could have sworn we were in a voice call together at some point?

Hm, might have me confused with someone else. Or maybe just confusing a voice call with messages (we've probably sent a couple messages back & forth here in the past).
 
Well if they're going to make Pro-level refreshes then yes I'd agree. However IMO they aren't going to do that this generation.
But if they're not launching Pro-level refreshes then they won't launch new hardware with a different architecture and small performance gains of 25-30%.

Launching a new console with different hardware means developing for one more hardware target, which takes more development time and makes the game more expensive to make. Many developers are already unhappy with Microsoft launching two consoles at the same time.

I concede they may not launch Pro-level refreshes this generation (though I think it's a bit unlikely considering Sony already has a bunch of patents for MCM solutions out there), but if they're not launching new consoles with >=2x the performance then all they'll do is launch slim versions with process shrinks. At best, you'd see updated video codecs like AV1 decoding or an incremental update to HDMI 2.1, but nothing that would require developers to work on yet another hardware target.
 

PeteBull

Member
Consoles need more gpu power, much more , at least 2x current and it will come as soon as sony and ms can make sure those midgen upgrades wont cost 999$ like they would coz of current crypto boom(which somehow lesseneed thankfully but prices didnt come back to normal yet, there are tens if not hundreds of milions of gpu's mining shit currently btw, just so u can imagine the scale of this shit).

What more gpu power can be used for- obvious choice will be rt/more rt but imagine on ur ps5pro u got not only resolution and fps mode but u got special ps5pro mode that is same settings/res as base ps5 resmode but still runs in 60fps- would be pretty pro, no? ;)

About possible cpu upgrade- hard to say- but by the time new consoles launch in 2024/25 zen2 arch will be quite outdated, this year so in 2022 there is coming ryzen 6000 series which is zen3+ architecture so on same tdp/cores u will easily get 30-40% improvement even if consoles gonna use just that(so 2022 tech as a 2024/25 launch).

There will be market for it, even if those consoles cost 799$ for the simple reason- even if/when ps5 price goes down to 400$ for base model and 300$ for digital one, logically thinking wont u pay 2x price for 30%stronger cpu, 100% stronger gpu and probably even more memory too? i know i will and many other enthusiasts too, such midgen upgrade gonna get 20-30% of sales of base ps5 easily even at 800$, if price is lower then obviously even more ppl will notice its great deal and want to buy it.

Dont think relatively high price will stop enthusiasts, the hunger for more performance is real, ppl already tasted the beauty of being pr0 user and not pleb user few years back when ps4pro and xbox one x came out :)
 
Last edited:
The PS4 Pro and especially the One X didn’t even sell that well to justify Pro revisions for the PS5 and XSX. Most people were happy with their original consoles. Not sure if announcing new, better versions of the hard to find consoles is a good idea when the original versions will probably be very difficult to acquire for a few years. The Slim versions? Maybe, but I definitely think several years from now.
 

asdasdasdbb

Neo Member
IMO, no price cuts for the current consoles and no slim models are coming.

I do think there will be pro models but at higher prices, but the timing is proving difficult because TSMC is proving to be so popular.

And no... neither will do a minor update. It'd have to be 2.5x+ at least. I am not sure if N4X would be enough of a shrink for that.
 

PeteBull

Member
Xbox one x sold terribly coz whole xbox one sold terribly(50-55m total?), but since ps4 base+pro is now a at 116-117mln then u can bet pro sales are somewhere between 15-20mln too.
U can bet as soon as the tech to make slim version is there(so 3nm) there will definitely be presented bigger pro brother :)
Edit to not double post https://www.ign.com/articles/2017/06/06/ps4-pro-represents-nearly-one-in-every-five-ps4s-sold
When talking to Glixel, the CEO of Sony Entertainment Europe Jim Ryan, stated that the success of the PS4 Pro exceeded the company's expectations, with the PS4 Pro accounting for nearly one of every five PS4s sold.

"We were confident about the product but have really been taken aback by how well it's done," Ryan said. "Almost one in five PlayStations sold since that launch in November has been a Pro. That's significantly ahead of our expectations. We're feeling pretty good about that."
 
Last edited:
Architectural updates won't net you any meaningful performances advantages for a mid-gen refresh. I think you're forgetting how woefully neglected the specific hardware features on the PS4 Pro were. The mid-gen consoles are designed to play the same console games as the base consoles exclusively. The idea is that devs build the same game binary and it runs on both boxes, meaning any new microarchitectural updates on the mid-gen box GPU will be invisible to the game by definition; since the games will run exclusively in effectively BC mode.

In order to take advantage of any new microarchitectural feature in hypothetical mid-gen boxes, devs will have to code for them specifically, which defeats the entire design objective of the mid-gen console.

Wait, how true is this necessarily? Because there are definitely PS4 games for example that still keep base PS4 perf profile even on the PS4 Pro because they never got updated by the developer to utilize the extra horsepower. Street Fighter V is a good example of that; altho it already ran at 60 on the base system, they could've tried upping the resolution for a Pro-enhanced mode but never did.

As well, there were some 1P games from certain Sony studios that took advantage of RPM IIRC, which was specific to the Pro and those would've required versions coded to Pro's specifications and using its features explicitly. I get that these don't represent the use-case for the majority of PS4 games running on the Pro, that's true. But it still had some titles which leveraged specific hardware features and to do that they needed to explicitly target those right?

When the cache gets too small the hit rate reduces and it starts to get less effective than straight up better memory bandwidth. It's visible in most 6000 gpu benchmarks vs 3000 gpus as the resolution increases. The simple view on this is look at 6600 vs 2060 at 1080p and 1440p, the 32mb IC does ok at 1080p but 1440p it isn't enough and the 2060's increased real bandwidth pulls it ahead. Now if we think about these systems coming out in 2-3 years and games get better visually as expected that small IC cache looks pointless unless everything is 1080p I guess.

Good point. But, can't there be a caveat here? If we're talking about refreshes with better dedicated silicon for image upscaling comparable to DLSS or XeSS, then we're talking about games going with natively lower internal resolutions anyway. So a 32 MB IC pool for example wouldn't be as much of a detriment in the design, because it could be enough for a 1080p framebuffer. If the internal resolution can be set even lower, and upscale to something like 2K and guarantee 60 - 120 FPS (depending on mode), it sounds like it could be worth it.

I get what you're saying about cache size and the hit rate, but in this example what about making the cache slightly larger (48 MB for instance, or 64 MB as another example)? You still have the benefit of lower energy usage vs. going across the memory bus to access the data in RAM, and it's a better solution for data locality as well. That said I can see equally where a design just going with a larger bandwidth main RAM pool in theory can work just as well (tho I do prefer something focusing on a larger last-level cache, personally).

RT is also heavy on bandwidth and it's probably why AMD sponsored titles only go for shadows because they are more coherent and probably have a better cache hit rate than other effects, while nvidia chose to go with pushing gddr6 to gddr6x for more straight up bandwidth. We will get the answers to that this year though.

Here's what Infinity cache does when you take the clock speed advantage away from the 6700xt and bench it against a 5700xt without any. I honestly think there are better uses for the die space in your scenario.


I'm somewhat curious if some of this could be improved if they leveraged cache scrubbers the way the PS5 does (which I'm assuming use patrol-based scrubbing). In some ways the IC design, at least as it exists currently, seems like more a "brute force" means of adding cache and maybe not the most efficient cache design that could be had.

What I was meaning is when would the tech be ready to have that sort of power in a console form. Like, if they use 5nm fab, could they double the GPU size and still keep a Chip at the same size and heat as the current one in the PS5?

Not with 5nm I think; there were some guessimate TDPs for top-end RDNA3 GPUs (15,360 shader cores) on 5nm and I think it's something around 275 - 300+ watts. Tho that's also a potential 75 TF card; I guess you could divide that by 3 and you get a GPU of something around 100 watts at 25 TF (it wouldn't actually work this way, just doing it as an example).

Although a PS5 Pro has to double the CU count, and still run at at least 2230 MHz; the 6900 XT is rated at 300 watts and that's a GPU on 7nm; I think you'd need something like 4nm (if that brings a reduction on power consumption) to make that work for a PS5 Pro, and even then it'd have to keep literally the same design as the PS5 architecturally just doubled.

But if they're not launching Pro-level refreshes then they won't launch new hardware with a different architecture and small performance gains of 25-30%.

If they can squeeze out that extra perf. without increasing power consumption over the base systems, or increasing BOM over the base systems, then it would still be worth doing IMO, especially if such a model could also phase out the base systems. It's just that the extra 25-30% perf would have to come from raw performance software or devs themselves needn't explicitly target.

Launching a new console with different hardware means developing for one more hardware target, which takes more development time and makes the game more expensive to make. Many developers are already unhappy with Microsoft launching two consoles at the same time.

True, but if RDNA 3 or 4 are essentially similar to RDNA 2, just with extra features baked in, and the microarchitecture remains virtually similar, is it really the same as being two wholly different architectures? I think the issue with the Series S is that it's already fairly constrained in ways for what the baseline of performance for 9th-gen AAA games should be (and certainly will be within a year or so). The PS5 is in many ways the top performer for 9th-gen AAA content at the moment, outside of some exceptions, and fairly well designed to not be a constraint on the dev process as the generation goes onward.

So if on the Sony side that's the baseline, I don't think as many developers would be upset if they needed/wanted to make games for the regular PS5 and then build a higher-performance version for a PS5 Pro.

I concede they may not launch Pro-level refreshes this generation (though I think it's a bit unlikely considering Sony already has a bunch of patents for MCM solutions out there), but if they're not launching new consoles with >=2x the performance then all they'll do is launch slim versions with process shrinks. At best, you'd see updated video codecs like AV1 decoding or an incremental update to HDMI 2.1, but nothing that would require developers to work on yet another hardware target.

Have the MCM designs been solidly connected to a potential Pro? I thought they were hinting more towards a server-side cluster type of thing.

Can definitely see slim models (in fact that's basically what I'm proposing, just potentially slimmer models supporting RDNA 3 or 4, which should have full BC with RDNA 2) and updates for video codecs (they can also add these to the base systems with firmware updates).
 
Wait, how true is this necessarily? Because there are definitely PS4 games for example that still keep base PS4 perf profile even on the PS4 Pro because they never got updated by the developer to utilize the extra horsepower. Street Fighter V is a good example of that; altho it already ran at 60 on the base system, they could've tried upping the resolution for a Pro-enhanced mode but never did.

I'm a little confused why you cite this as a counter-example when it decidedly proves the point I made. Yes. PS4 Pro played unpatched PS4 games with exactly the same performance profile as the base console.

As well, there were some 1P games from certain Sony studios that took advantage of RPM IIRC, which was specific to the Pro and those would've required versions coded to Pro's specifications and using its features explicitly. I get that these don't represent the use-case for the majority of PS4 games running on the Pro, that's true. But it still had some titles which leveraged specific hardware features and to do that they needed to explicitly target those right?

Yes. That's exactly what happened for those games, i.e. the games were specifically coded for the PS4 Pro, which is why you only saw generally first parties doing this because not having to deploy on both Xboxes and PC in addition, meant they had the bandwidth to be able to exploit PS4 Pro a bit more fully. The vast majority of devs didn't even bother, however.
 

hybrid_birth

Gold Member
LOL I don't think anyone really understands the gravity of the current market climate to understand that.

How many next gen exclusives do we even have on the market right now? Have developers even really taken advantage of the next gen feature set of these consoles? But sure these companies are going to spend more $$$ on Pro revisions when they can't even ship enough consoles to meet current demand 😂

May as well start the PS6 and Xbox Series XX threads.
Aw shit. I better cancel my plans to make a PS6 thread. I had a page all written up on information.
Destroyer Rage GIF by A1 eSports
 

LordBlodgett

Gold Member
I just want to say that this is a really well thought out OP. I can see you put a lot of time into it, and I appreciate that.
Morgan Freeman Applause GIF by The Academy Awards


I think that if we do see any mid-gen refresh this time it is going to happen later than last gen just because this chip shortage is not getting better. I also agree that slim models will happen as soon as the supply chain opens up a little bit, especially if they can move to a smaller node. Both the PS5 and Series X APU chips are pretty huge, which means prices will be hard to get down without reducing their sizes. Sony's Spartacus is too much of an unknown for me, best case it is like Gamepass with day one releases, but I kind of bet it won't be. I do see both Sony and Microsoft getting rid of PS Plus/Gold Live and folding them into their other subscription plans (Spartacus/Gamepass). I am not sure who will blink first though. I also don't see this happening until there are more than enough consoles being produced to meet demand, because why change when you are raking it in with multiple services and demand outstrips supply.

In terms of some of your other predictions, I can't see Sony launching a mobile type device anytime soon. With the focus they are currently putting on VR I cannot see them stretching themselves too thin by having three unique devices to develop for. I could see Microsoft introducing some type of Roku/Apple TV like device if they don't get enough buy-in from TV manufacturers on built-in Gamepass apps, however I don't think they are going to have any issues there. I expect most TV manufacturers are going to be happy to have Gamepass built-in and Microsoft will actually be able to be picky about minimum processing power/RAM/etc for manufacturers.

Anyways, it is so early to be making these kinds of predictions that everything is kind of a crapshoot, still very interesting to think about though!
 
The PS4 Pro and especially the One X didn’t even sell that well to justify Pro revisions for the PS5 and XSX. Most people were happy with their original consoles. Not sure if announcing new, better versions of the hard to find consoles is a good idea when the original versions will probably be very difficult to acquire for a few years. The Slim versions? Maybe, but I definitely think several years from now.
Didn’t Sony not disclose how many pros sold and just lumped in all ps4 models together for sales records?
 

Keihart

Member
if the chip shortage continues for a couple of years i don't the last gen getting updates unless it's extra long this time...wich could be the case with all the PC porting and shit happening.
 
The pro models isn’t happening because it would make zero sense when you consider that most people can’t even get the regular models and we are still getting so many cross gen games

The justification for them just isn’t there like last gen.
 
Last edited:

PeteBull

Member
The pro models isn’t happening because it would make zero sense when you consider that most people can’t even get the regular models and we are still getting so many cross gen games

The justification for them just isn’t there like last gen.
U talking today, give it time to 2024/2025 and situation will be vastly different, by then both sony and ms will have access to tech that will allow them for 600$ or so 2x power vs current gen, lots of ppl will want it.
 
U talking today, give it time to 2024/2025 and situation will be vastly different, by then both sony and ms will have access to tech that will allow them for 600$ or so 2x power vs current gen, lots of ppl will want it.
No one is going to spend $600 on a console just to get a slight bump in performance. Sony tried that before with the PS3 and it screwed them big time.

We have not even had enough native next gen games to gauge what these new consoles are even capable of doing and people want more? It’s not like 8K will be a thing anytime soon, so what value would the pro models really bring?
 

PeteBull

Member
No one is going to spend $600 on a console just to get a slight bump in performance. Sony tried that before with the PS3 and it screwed them big time.

We have not even had enough native next gen games to gauge what these new consoles are even capable of doing and people want more? It’s not like 8K will be a thing anytime soon, so what value would the pro models really bring?
600$ now or especially in 2024/2025 will be equivalent to what 300$ was back at og playstation launch, u really understimate how much enthusiasts are willing to spend on their hobby, we talking about ppl buying 20+ games/year. Im saying it as a 3080ti owner that i had to buy for 2,2k euro, i was angry but i did it, coz im enthusiast.
 
Sam Lowry (Jonathan Pryce) is a harried technocrat in a futuristic society that is needlessly convoluted and inefficient. He dreams of a life where he can fly away from technology and overpowering bureaucracy, and spend eternity with the woman of his dreams. While trying to rectify the wrongful arrest of one Harry Buttle (Brian Miller), Lowry meets the woman he is always chasing in his dreams, Jill Layton (Kim Greist). Meanwhile, the bureaucracy has fingered him responsible for a rash of terrorist bombings, and Sam and Jill's lives are put in danger.

That movie is so trippy
 
600$ now or especially in 2024/2025 will be equivalent to what 300$ was back at og playstation launch, u really understimate how much enthusiasts are willing to spend on their hobby, we talking about ppl buying 20+ games/year. Im saying it as a 3080ti owner that i had to buy for 2,2k euro, i was angry but i did it, coz im enthusiast.

People that care that much about slight upgrades don't game on consoles. That is more of a PC thing, last gen refreshed happened because of the rise in 4k and the weak state of the base consoles (PS4/Xbox one). That really isn't the issue now with the PS5/Series X. Right now the only two things holding them back is supply issues and cross gen games.
 

PeteBull

Member
People that care that much about slight upgrades don't game on consoles. That is more of a PC thing, last gen refreshed happened because of the rise in 4k and the weak state of the base consoles (PS4/Xbox one). That really isn't the issue now with the PS5/Series X. Right now the only two things holding them back is supply issues and cross gen games.
Crossgen game doesnt mean it somehow runs 100% better once its not crossgen anymore, visual masterpiece like newest horizon wouldnt look/run better on ps5 if it wasnt avaiable on ps4/ps4pro, it will look and run better on ps5pro if it runs at 60fps at same settings it currently runs on ps5 in 30fps mode.
U even got examples of demons souls remaster, which is ps5 exclusive and not a crossgen game, u still have to chose, 4k30 or 1440p60 with it, not to mention ratchet, again beautiful game and ps5 exclusive, not a crossgen, still would benefit tons if ps5pro was avaiable and 2x more powerful.
 
Great speculation. Interesting read. Well thought out and detailed.

My fear is that with chip shortages, supply issues, and uncertainty on the world stage pertaining to war... As well as the fact you still can't find a Series X or PS5 easily at retail over a year out...

If any of those plans existed, I feel they're going to change or be postponed unfortunately. I think the newest revisions will maybe be the same form factors with more SSD storage.
 
I'm a little confused why you cite this as a counter-example when it decidedly proves the point I made. Yes. PS4 Pro played unpatched PS4 games with exactly the same performance profile as the base console.

That's right, but PS4 Pro also has a different GPU architecture, Polaris, compared to base PS4. Maybe "architecture" is not the best term; IIRC it's basically a more advanced version of GCN, 4.0 or something like that. But it's still not 1:1 same as the base PS4's GCN; it just features that design spec at the core with other features built into it.

Yes. That's exactly what happened for those games, i.e. the games were specifically coded for the PS4 Pro, which is why you only saw generally first parties doing this because not having to deploy on both Xboxes and PC in addition, meant they had the bandwidth to be able to exploit PS4 Pro a bit more fully. The vast majority of devs didn't even bother, however.

That's true, most didn't bother. But, even those games got some slight benefits from running on PS4 Pro. I wish more games did develop upgraded performance modes explicitly targeting PS4 Pro (and same with 3P games on the One X).

Aw shit. I better cancel my plans to make a PS6 thread. I had a page all written up on information.
Destroyer Rage GIF by A1 eSports

Nah go for it; the more the merrier!

I just want to say that this is a really well thought out OP. I can see you put a lot of time into it, and I appreciate that.
Morgan Freeman Applause GIF by The Academy Awards


I think that if we do see any mid-gen refresh this time it is going to happen later than last gen just because this chip shortage is not getting better. I also agree that slim models will happen as soon as the supply chain opens up a little bit, especially if they can move to a smaller node. Both the PS5 and Series X APU chips are pretty huge, which means prices will be hard to get down without reducing their sizes. Sony's Spartacus is too much of an unknown for me, best case it is like Gamepass with day one releases, but I kind of bet it won't be. I do see both Sony and Microsoft getting rid of PS Plus/Gold Live and folding them into their other subscription plans (Spartacus/Gamepass). I am not sure who will blink first though. I also don't see this happening until there are more than enough consoles being produced to meet demand, because why change when you are raking it in with multiple services and demand outstrips supply.

Considering Microsoft were eager early last year to try shifting people to GamePass by almost raising XBL Gold by 100%, I have a feeling it'll be them...just hopefully without pulling that stunt again 😉. Also agreed with this stuff mostly being held off for a couple years, I'd like to think by 2024 supply finally catches up to and begins outstripping demand, if we want a semi-normal console generation length-wise.

In terms of some of your other predictions, I can't see Sony launching a mobile type device anytime soon. With the focus they are currently putting on VR I cannot see them stretching themselves too thin by having three unique devices to develop for. I could see Microsoft introducing some type of Roku/Apple TV like device if they don't get enough buy-in from TV manufacturers on built-in Gamepass apps, however I don't think they are going to have any issues there. I expect most TV manufacturers are going to be happy to have Gamepass built-in and Microsoft will actually be able to be picky about minimum processing power/RAM/etc for manufacturers.

Anyways, it is so early to be making these kinds of predictions that everything is kind of a crapshoot, still very interesting to think about though!

I think the only potential problem with pushing the super-cheap GamePass option as a bundled app on TVs is that, if the minimum specs required will necessitate more people to buy newer TVs, that actually raises the entry price for GamePass (and just for the streaming component at that) to a few hundreds of dollars. Microsoft would just have to go with the absolute bare minimum of min specs, but leaving it in the hands of TV manufacturers could create a quality control issue when it comes to consistent baseline performance.

If they can just make a small Apple TV-style streaming box themselves, they eliminate that potential issue. Also in mass bulk MS can get those components cheaper per-device than a TV manufacturer could, and given it's a streaming box dependent on a subscription plan, those upfront costs for the device can be passed onto the customer i.e they can sell that type of streaming box for very cheap or maybe even as a loss-leader model ;)

People that care that much about slight upgrades don't game on consoles. That is more of a PC thing, last gen refreshed happened because of the rise in 4k and the weak state of the base consoles (PS4/Xbox one). That really isn't the issue now with the PS5/Series X. Right now the only two things holding them back is supply issues and cross gen games.

If upper-end PC gaming keeps trending to more power-hungry GPUs and increasing costs for components, not to mention continued GPU shortages, it could actually present a unique proposition for upgraded 9th-gen systems.

I agree though that for the current iterations, it's mainly down to supply and many teams still prioritizing crossgen development.

Great speculation. Interesting read. Well thought out and detailed.

My fear is that with chip shortages, supply issues, and uncertainty on the world stage pertaining to war... As well as the fact you still can't find a Series X or PS5 easily at retail over a year out...

If any of those plans existed, I feel they're going to change or be postponed unfortunately. I think the newest revisions will maybe be the same form factors with more SSD storage.

That's probably the safest bet to make at this point in all honesty. A bit unfortunate considering the possibilities, but as you mentioned there are the shortages and now the very unfortunate Russia/Ukraine conflict that (hopefully) is resolved swiftly and absolutely with all parties generally satisfied (and Ukraine holding onto their independence, of course).
 
Last edited:
That's right, but PS4 Pro also has a different GPU architecture, Polaris, compared to base PS4. Maybe "architecture" is not the best term; IIRC it's basically a more advanced version of GCN, 4.0 or something like that. But it's still not 1:1 same as the base PS4's GCN; it just features that design spec at the core with other features built into it.

No. Polaris is just a name for the same underlying GCN microarchitecture that underpins both the base PS4/XB1 and the Pro. They are totally 1:1. since the iteration of the microarchitecture is designed to be 100% backwards compatible.

When you talk about exploiting new architectural features, you have to be specific. I'm talking about things like with how RDNA supports wave32. A game built for the PS4 but played on the PS5 in BC mode wouldn't be able to take wave32 wavefronts into consideration, because doing so would make the game unable to run on the base console.

The above example is pretty analogous to how mid-gen consoles work.

That's true, most didn't bother. But, even those games got some slight benefits from running on PS4 Pro. I wish more games did develop upgraded performance modes explicitly targeting PS4 Pro (and same with 3P games on the One X).

Games coded with unlocked framerate and soft resolution caps can take advantage of higher clockspeeds but that's about it. They can't even address a wider GPU vector array, which is why the PS4 Pro disable shader arrays to run PS4 games vanilla in BC mode.
 

Not necessarily related to the thread, but just wanted to put it as a proof of concept, that a new PlayStation portable with native PS4-level performance would be possible by 2024/2025 as a first-party device. Steam Deck is already encroaching near that and is available now (or very close to now), although it's not an exact PS4 spec in a portable form factor.
 
Soooo, seems like there might be a handheld in the works after all...


I think a lot of this is definitely happening. However I don't think the handheld's going to be PS4 Pro level of performance, and I don't think Sony are going to split around their teams to do exclusive 1P games for it, either, since they'll probably already be busy with PS5 & PSVR2 games content.

Personally, that handheld will probably be closer to base PS4 in performance, and like I've been speculating for a while now, being ore a PS4 portable system and PS5 Remote Play streaming device. Any 'exclusive' software would come from smaller teams still targeting the PS4 market, and any new games targeting the portable specifically would also be playable on PS4, PS4 Pro and PS5 consoles natively and with no extra work outside of maybe some controller options (if the handheld has some features that aren't replicated on a console/controller setup mainly).

It'd also allow them to have good profit margins on the hardware from Day 1, which is what Nintendo manages with their hardware; it could even function as a means of enabling wireless use of PSVR2 in some way. I'm less certain if it would be something that could play PS5 games scaled down; though that'd be great, it'd probably need at least 2.57 TF of performance and support the RDNA2 featureset explicitly.

That would use more power, though, and probably require it to be docked with a design mechanism for expanded cooling so the clocks could be increased. There's the more pressing concern of if 1P teams would choose to prioritize PS5 and scale down to the handheld by changing resolution settings and asset sizes, or just how much game logic, physics, AI systems and the such (anything aside from textures, graphics assets and resolution) specifically designed for a PS5 performance profile, could scale down to a portable with a much weaker performance profile.

Maybe in that case games which need it could use local processing and streaming (Remote Play or cloud online) to handle the experience on the portable. But, at least it seems like things are likely going forward with a portable from them, and I still think they'll do a PS5 Digital refresh making the SSD capacity much smaller but adding in microSD storage, to bring BOM prices down so they can manufacture more Digital systems at real volumes and not losing money on units sold. Also I know guys like RGT and MLID are kicking around a more powerful PS5 Pro in 2023/2024, but IMO I don't see it happening, mostly for logistics and supply reasons.
 

ethomaz

Banned
I hope not.

We need more focus in optimizations and less modes in actual generation hardware.

Now a handheld is something Sony needs to try again.
That time with a better ecosystem integration.

The handheld should be a part of PS5... both needs to work together.

PS5 + PSVR2 + PSP3 = Perfect world.
 
Last edited:
Top Bottom