Middle Generation Console Upgrade Discussion [Scorpio, 4Pro]

Status
Not open for further replies.
I think it does. It draws 2x the pixels or 2x the framerate.


Not true directly from a someone working for a first party Sony developer talking with me. You can do betterusing fp16 when it is possible(postprocessing for example), dual fp16 if possible need to packed similar fp 16 instruction by group when possible (a GDC conference by Timothy Lottes will give advice to improve performance on Vega architecture and he will talk about the subject). Use the ID buffer for other things than checkerboard rendering and so on...

And the first party will do it later... Now the optimization for PS4 Pro is minimal... And it will probably be the same on Scorpio developer will need some time to improve their game...
 
Last edited:
there are 2 games that stick out, mantis burn racing that manages native 4k 60fps by utilizing the fp16 functionality, looks identical to the PS4 but 4k. Also Tombraider has a 1080p pro mode when they push framerate towards 60 but do not make it locked, or they push for 4k 30fps with all the effects at max including cut scene renderers and hair effects used in-game.

I am sure we will see some better efforts on the pro going forward as people get used to it and its a consideration from earlier in the games development.
 
sorry, I didn't help.
I should've reworded my response as i think it's relevant to Scorpio, as in how much difference can you expect to see, and compared to if game was made from ground up, etc
the specs is what determines it
 
sorry, I didn't help.
I should've reworded my response as i think it's relevant to Scorpio, as in how much difference can you expect to see, and compared to if game was made from ground up, etc
the specs is what determines it

For your defense the Scorpio GPU will probably use Vega architecture too and developer doing extra effort will extract more from Scorpio than Xbox One/PS4 better resolution game...
 
To Al's point:
There is large quantity of strategy talk here, mostly unnecessary, about the exchange between Sony and MS on a lot of what ifs. This discussion would be much more accurate and precise if we knew whether Scorpio was Vega or Polaris, at least that would funnel the strategy discussion into something with more signal-to-noise ratio.

please correct any inconsistencies, I've not been keeping up to date:
I don't think the fab technology between 2016 and 2017 is changing, so if Sony could have fabbed 4Pro with 6TF on Polaris at TDP they would have.
We're basically comparing a cut down Vega (14nm) vs a full 480X (14nm)

That pretty much means to me it wouldn't be possible in 2017 either since nothing is changing.
So that is my intuition on it having to be Vega.
 
Amd will be releasing raven ridge apu with 4 core 8 threads zen cpu + 1024/16cu vega gpu ( measuring around 210 mm2) tdp of 95watt in 2h of 2017 . Can scorpio use this configuration with much more compute units and higher clocks to reach 6tfs ?
 
if both sony and ms release similar 7nm console the relative performance difference from old sony platform to new platform is bigger. This would give bigger incentive for ps4/ps4 pro owner to upgrade than people owning scorpio. Sony is better aligned to have more balanced roadmap than microsoft. Microsoft also might have pricing difficulties as scorpio is likely to be either expensive or sold at loss. If ms&sony intend to bring in new 399$ 7nm console what is the pricepoint scorpio needs to be at or would it be discontinued?

People don't buy new consoles (or any other product) because of their relative performance increase compared to their older counterparts. If XB5 utilize similar tech at similar price relative to PS5 then clearly it will have higher value for those who were Microsoft consumers before and those people who tend to buy games digitaly. Forward/Backward software compatibility, constant library of game/services/applications/features, compatible hardware accessories, cross-play with Windows 10 & older Xbox consoles and increasing number of Xbox players on Xbox live is something that will make XB5 much more attractive.

َAlso Scorpio at 7nm won't have any pricing issue as that wasn't a problem for XB1S/PS4 Slim, Xbox 360 S/PS3 Slim, etc.
 
It may be enough to go beyond the jump Scorpio has over PS4 Pro.

True, but for the vast majority of users PS4Pro isn't going to be on their radar (Scorpio might be either, for that matter).

In terms of CPU, it'll depend on what Scorpio has and in terms of memory Sony would need to be looking at above 16GB of RAM.

http://www.eetimes.com/document.asp?doc_id=1330940

"TSMC described a 256-Mbit SRAM test chip with the cell density of 0.027 mm2 with full read/write capabilities down to 0.5 V. The node should provide up to a 40% speed gain, a 65% power reduction, and a 3.3x routed gate density increase compared to TSMC’s 16FF+ process now in volume production, said Michael Shien-Yang Wu, a senior director of N7 development at TSMC."

Well, if TSMC's and GF's comments both bear out then TSMCs 7nm should have vastly greater density than Samsung and GF's. But ... I'm a little cautious about "up tos". Idealised figures from SRAM test chips are unlikely to accurately represent what happens when you shunt entire, complex processors over.

And GF may be the fab of choice for both Scorpio and PS5, given that it's currently the natural home of Vega and Zen.

PS4 Pro does not have any exclusive games, its 2x tflop advantage does not get utilized to its maximum. Sony has always maintained that the Pro is still just a PS4.

Indeed, and that's always likely to be the case. MS's plan - if it can be believed - is to leverage higher quality assets, shaders, shiney stuff, from PC land and so Scorpio should go beyond just getting tweaked X1 games. Well, in theory ...
 
Digital sales are only going to continue to rise unless something catastrophic happens, like the internet falling apart. Especially as console prices drop and more of your average consumer who prefers digital purchases enter the ecosystem.
So...Do you back the idea or no? :D

Because CPU-dependant things are far less scalable than GPU-dependant things, for example, everything(AFAIK) still runs on a desktop version of X1's gpu, but a lot of games simply don't run on even better cpus, than jaguar. In numbers, GPU scale - ~15 times, CPU - ~5, at best, jaguar is at absolute bottom right now, PS5 CPU could be 6 times more powerful, with 4 times, imo, being given.
Scorpio comes out before ps5 and by how many years we don't know.
2-3 years, 2, most likely, since, 7nm Zen(~= full blown 14nm desktop Zen that Scorpio won't have) - check, Na'vi - check, HBM - check, (somewhat)mature techprocess - check, it might be not worth waiting another year(for what?) from a timing perspective, especially if targeting $499, ~190W from the get go.
There could be tens of millions of scorpios sold with just slower zens and gpus attached.
"Could be" is the key word, i don't see why there would be. "Just slower Zens", but still not up to par(Switch also got Maxwell(or Pascal), but it won't make it run games that run on Titans), that alone screams about how unlikely it is, even if it is possible(think not, i haven't seen anything regarding low-power Zen), post from June that might help to illustrate my logic:
Look, here's the thing, even if Sony releases the most fantastic, theoretically impossible, PS5 next year, total reset won't happen anyway until the next decade starts(and by that time shining 7nm(it's should look way better than 14nm hardware does now, especially for -) consoles become a reality, rendering 14nm PS5(Scorpio?) redundant and even harmful financially), and that's even with Sony actively pushing it
Zen requires way more investment, yet doesn't warrant any return at such stage.

It will be much easier to design for scorpio and port down to xboxone/ps4 and up to ps5 .
I don't see how, in order to make sure it even works(especially since your version of Scorpio has Zen in it), devs would still have to focus on a PS4/X1 versions first, and why would anyone, other than few select devs even design for Scorpio first anyway? It's infinitely unlikely to ever get larger active paying audience than PS4(P)/X1.
Developers also wouldn't want to give up any installed base to design games exclusively for a new system that sells 0 games. Look at how poor the switches line up is at launch.
There is already millions upon millions in "Zen-based tier", PC gamers, plus i'm not saying everyone would just abandon everything the minute PS5 comes out, i'm saying no one would stop half-way for the reasons already mentioned.


P.S. I'm not interested in arguing, merely expressing my opinion, this situation is all sorts of mind-boggling to me, and while i do go into detail where one can be easily wrong, to illustrate at the core this topic is pretty fundamental and it doesn't change overnight.
 
There is large quantity of strategy talk here, mostly unnecessary, about the exchange between Sony and MS on a lot of what ifs. This discussion would be much more accurate and precise if we knew whether Scorpio was Vega or Polaris, at least that would funnel the strategy discussion into something with more signal-to-noise ratio.

How is it not very safe to assume it's at least Vega if not even Vega+?
Xbone and PS4 got GCN2 features (8 ACEs, TruAudio on PS4) + custom features about 6 months before the first desktop GCN2 GPU release and the PS4 Pro got Vega features also ~6 months before the first Vega card will come to the market.
Why assume Scorpio got an older architecture?

Unless AMD decided to give Microsoft the middle finger regarding which architecture they made available for Scorpio (which I find very hard to believe), you can count on the new console bringing the latest and greatest from AMD.
In fact, a much more plausible discussion is if Scorpio is bringing some Navi features.

EDIT: a letter
 
Last edited by a moderator:
Well, if TSMC's and GF's comments both bear out then TSMCs 7nm should have vastly greater density than Samsung and GF's. But ... I'm a little cautious about "up tos". Idealised figures from SRAM test chips are unlikely to accurately represent what happens when you shunt entire, complex processors over.

And GF may be the fab of choice for both Scorpio and PS5, given that it's currently the natural home of Vega and Zen.

Global Foundries 7nm will be developed with IBM (their 14nm process was with Samsung) while Samsung will most likely arrive late with their 7nm node since it uses EUV. They have already stated they will extend the lifespan of their 10nm process. Based on everything i have read so far TSMCs 7nm will be as dense as Intels 10nm wich is fairly impressive so i dont think its idealised, it will be a fairly decent jump from their 16nm process

I dont think AMD gets to decide foundry for a custom APU funded by Sony or Microsoft. I would bet dollar to donuts that Scorpio will be on a TSMC process and not GF, just like PS4 Pro being TSMC while RX 480 was GF.


Indeed, and that's always likely to be the case. MS's plan - if it can be believed - is to leverage higher quality assets, shaders, shiney stuff, from PC land and so Scorpio should go beyond just getting tweaked X1 games. Well, in theory ...

While i think the jump from XB1 to Scorpio will be larger than PS4>Pro, i believe sebbbi posted on another thread that 12 GB DDR5 is still nowhere near enough for 4K assets. So even Scorpios 4K experience will be a compromised one
 
Global Foundries 7nm will be developed with IBM (their 14nm process was with Samsung) while Samsung will most likely arrive late with their 7nm node since it uses EUV. They have already stated they will extend the lifespan of their 10nm process. Based on everything i have read so far TSMCs 7nm will be as dense as Intels 10nm wich is fairly impressive so i dont think its idealised, it will be a fairly decent jump from their 16nm process

Ah! That makes more sense then. Have to say I'm hoping GF don't end up so far behind TSMC though.

I dont think AMD gets to decide foundry for a custom APU funded by Sony or Microsoft. I would bet dollar to donuts that Scorpio will be on a TSMC process and not GF, just like PS4 Pro being TSMC while RX 480 was GF

I was thinking more along the lines of Sony and MS going where they can get the best deal / bang for buck rather than AMD making them go somewhere. If designs are tuned for a certain fab, it may work out best to stay there. There's also the potential issue of whether AMD have the engineering capacity to move designs across. I'm 50/50 on where Scorpio is going to be made. Neither would surprise me, or seem like a bad choice. Vega seems to be shaping up okay at GF if the latest leaked slides really do indicate ~1.46 ghz for the 64 CU Vega.

While i think the jump from XB1 to Scorpio will be larger than PS4>Pro, i believe sebbbi posted on another thread that 12 GB DDR5 is still nowhere near enough for 4K assets. So even Scorpios 4K experience will be a compromised one

Even a 16 GB console probably wouldn't allow a 2x2 scale up of all your assets. However as with the PC, where cards typically double up memory between models, you can still selectively pick improved assets to get you the biggest improvement for the footprint.

PS4Pro, Scorpio, PS5 ... the 4K experience for all of them will be compromised to some extent. But you can still make improvements where there's headroom.
 
While i think the jump from XB1 to Scorpio will be larger than PS4>Pro, i believe sebbbi posted on another thread that 12 GB DDR5 is still nowhere near enough for 4K assets. So even Scorpios 4K experience will be a compromised one

I thought that is exactly the problem tiled resources aims to solve?

GOW4 on PC uses upto 5.2 GB of vram at 4K. That will make it tight within what we must assume Scorpio games will get. Does GOW ship with true 4K textures?
 
Status
Not open for further replies.
Back
Top