There's no way Sony launches PlayStation 5 in late 2019 with only a 9 Teraflop GPU. Regardless of how much a boost a new CPU will bring. It would just look bad from a marketing perspective, more so than a technical perspective (and technically, a 9TF GPU in PS5 won't be impressive). Waiting 2 years (after XB1X) to get 50% more power, when Microsoft did the same to PS4 Pro in just 1 year. Games running at 4K or anywhere near it, is still gonna chew up a large percentage of a 9 TFLOP GPU (it uses about 75% of a 6 TF GPU doing current-gen class graphics) so there wouldn't be enough left over for a generational leap beyond OG PS4, unless developers stuck with 1080p and no checker boarding to 4K or near it. PS5 GPU needs to be 12 TFLOP, minimum. Also, keep in mind, Cerny's comment about needing at least 8 TF GPU to render at native 4K without compromise, was referring to PS4 games, not next gen games in the early 2020s / PS5.
-
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.
Pachter: PS5 to be a half step, release in 2019 with PS4 BC
- Thread starter bitbydeath
- Start date
There's no way Sony launches PlayStation 5 in late 2019 with only a 9 Teraflop GPU. Regardless of how much a boost a new CPU will bring. It would just look bad from a marketing perspective, more so than a technical perspective (and technically, a 9TF GPU in PS5 won't be impressive). Waiting 2 years (after XB1X) to get 50% more power, when Microsoft did the same to PS4 Pro in just 1 year. Games running at 4K or anywhere near it, is still gonna chew up a large percentage of a 9 TFLOP GPU (it uses about 75% of a 6 TF GPU doing current-gen class graphics) so there wouldn't be enough left over for a generational leap beyond OG PS4, unless developers stuck with 1080p and no checker boarding to 4K or near it. PS5 GPU needs to be 12 TFLOP, minimum. Also, keep in mind, Cerny's comment about needing at least 8 TF GPU to render at native 4K without compromise, was referring to PS4 games, not next gen games in the early 2020s / PS5.
It really seems to get a proper 10 times power generational leap you are going to have to wait until 2020 based on what a lot of people have been posting. I still think 7 years is too long, but releasing a half step doesn't accomplish much either. Gonna be a long generation it seems.
Honestly from my perspective (large 4k Tv owner) the pro really did keep me console gaming longer than I normally do. By now I'm almost entirely pc gaming as the console hardware ages. The pro cleaned up the image for me a lot.
Shin
Banned
Most important thing to takeaway is that HD 7850 was launched in 2012 (PS4 - 2013), RX480 launched in 2016 same with PS4 Pro.
Using that logic, time frame or whatever you want to call it, it will be a GPU from 2019 if they are launching 2020.
nVidia will launch Volta in some months that's 15TF, AMD will need to keep up, it's where their refresh of Vega comes in and rebrand.
Navi should be out by 2019 according to that path otherwise they'll have to rebrand Vega 7nm again in 2019 which doesn't seem right.
As things stand GTX 1080 (8.9TF?) beats the shit out of Vega 64 (13.1TF) card, AMD can have all the flops they want but it ain't the same.
So even if PS5 has 15TF or whatever the hell it's going to be slightly better than a GTX 1080 but below a GTX 1080Ti.
Three years from now a GTX 1080 class GPU will be more along the lines of a mainstream card, at least if we look at progression.
Agreed.
Yeah, and that's why I think a ~15+ TFLOP GPU for PS5, largely based on Navi architecture, and perhaps some IP / blocks / features from the next GPU architecture (AMD "Next Gen") is a reasonable expectation.
Sulik is right. Sony needs a roughly 10 fold leap overall with PS5 over the original PS4. A 15+ TF GPU with the most modern architecture, not to mention something like a Zen2 Mobile class CPU, plus significantly more and faster RAM, could get Sony there, but not for release in 2019, but rather 2020.
I still am not expecting much over 12tflops unless Navi has some major performance per watt gains. I expect techniques like adaptive resolution, checkerboard, and temporal resolution to remain common on the PS5. I do agree that there will probably be some features of the post-Navi architecture though. Sony has had some key tech from upcoming AMD designs for both the PS4 and Pro feature sets.
Shin
Banned
It seems that is something they are focusing on and doubling down on the GPU after Navi, but it's yet to be seen.
Mr_Antimatter
Member
I'm really looking forward to seeing where Sony and AMD (if they stick together) go with the ps5.
By this time, their R&D team must have some sort of clear initial vision for what they want, and I'd suspect they have had initial talks with key devs/publishers about what feedback they have on the design and the ps4/pro.
By this time, their R&D team must have some sort of clear initial vision for what they want, and I'd suspect they have had initial talks with key devs/publishers about what feedback they have on the design and the ps4/pro.
12 TFlops would be enough, the minimum, but enough, especially if AMD's improves on its own efficiency and assuming Navi and semi-custom additions / tweaks provide enough additional capability over previous generation GPUs.
PS5 games built from the ground up to take advantage of twice the raw GPU power of XBX1, in addition to a whole new ocean of freedom in a new CPU. Could be pretty great, and even make it out in late 2019. I still think 2020 is the safest bet, but we'll see.
Shin
Banned
It's a tricky position, do you ride it out and sell as much PS4 as you can and risk Xbox launching before you.
Are the parts available that you need in 2019 and will it be a significant jump or do you wait till 2020 and hope MS launches the same time or a year later.
Given that Scorpio launches closer to 2018 I'd say 2020/2021 for their next console, Sony depends on the industry and advancement there.
Everything has to be right before they release a new console, if BC is a thing then it comes down to parts and node availability, they can leave PS4 on the shelves and sell it for $199.
Are the parts available that you need in 2019 and will it be a significant jump or do you wait till 2020 and hope MS launches the same time or a year later.
Given that Scorpio launches closer to 2018 I'd say 2020/2021 for their next console, Sony depends on the industry and advancement there.
Everything has to be right before they release a new console, if BC is a thing then it comes down to parts and node availability, they can leave PS4 on the shelves and sell it for $199.
Imo, the key here is how PS4 is doing. Sure, sooner or later Sony does HAVE to pick a date for releasing PS5, but as the market stands right now, I just don't see them rushing it. Sure, the X1x is more powerfull than the pro, but everything points toward that it's going to be a niche product so it's unlikely that a more powerful Xbox will suddenly turn tides and the Switch is aimed at an entirely different audience, so judging by how Sony stands right now in the competition, making the PS4 run as far as they can is the most likely scenario.
Hell, the very existence of X1x makes me believe Sony won't come with PS5 because they know MS won't be so stupid to come out with a new xbox in at least 2 (3 is more likely though) years.
This leads me to believe that the soones we hear from PS5 is early 2019, but a 2020 release is more likely imo. (also, 7nm will be more mature by then, and god knows even AMD might finds itself again by that time)
EDIT: a little addition:
X1x is releasing in 2018 right? So the earliest we see a new Xbox releasing is probably around holiday 2021 (but even that's too soon imo), so if Sony dates PS5 to either around holiday 2020 or early 2021, they'll have a couple months of advantage. Enough to be cheaper, but not significantly less powerful.
Shin
Banned
In 4 months (November), close to 2018 but still in 2017 and to take advantage of holidays.
On paper or presentation or whichever they could word it as "we released a console in 2017", making 2020 possible (3 years).
I'd say either early or late 2020 is when it will launch, you'll have maybe TLoU2, DS, Horizon2, GTA 6, FF7:R to go with the launch (be it DE versions or what's not).
That's still a better launch than PS4 had IMO, all that 4K (native or otherwise) goodie...man do want!
Emulated Hype
Member
I'm still curious how they'll handle cross-gen games if out-of-the-box backwards compatibility ends up being a factor.
Pubs made a lot of money on remasters and ports this gen.
Pubs made a lot of money on remasters and ports this gen.
That's exactly one of my fears. I can see them ditcing bc for this very reason.
But perhaps they'll do a 1:1 bc and if you want enhancements to your old game, you'll just have to purchase a patch or something.
kraspkibble
Permabanned.
I disagree. 30fps is OK but 60fps should be standard. A lot of games are let down if they aren't at 60fps. For example Destiny 2...a fps and it's only going to be 30fps. That's ridiculous. If you were to play both versions you wouldn't be saying it's not hindered by 30fps. Then again I guess if you play on consoles you don't really care about performance in the first place.
Oh by the way just ignore me because I'm just a ragin fanboy /s
imo people will still pay to have "current gen" versions of their favorite games regardless of BC. most of these will be GOTY editions with all DLC and such
Negotiator
Member
Zombie James
Banned
Remember when people said Sony would follow MS with their pre-launch DRM policies because that's what publishers want and would force them into it? Yeah, didn't happen.
Even with full backwards compatibility, nothing's stopping publishers from re-releasing last-gen games. They'd just have to put effort into it (oh no😱
THE:MILKMAN
Member
The 7 and 5nm design costs are scary compared to 28nm. Assuming the figures are close to true (or the relative difference) then what effect would this have on price? I mean I can't see a near half a billion investment on the APU alone would allow another $399 without a loss?
On the other hand a $499/599 price will mean slow/low sales again.
TheThreadsThatBindUs said:
Sorry, I wasn't trying to catch you out! From his posts I don't think he has said anything explicit, but reading between the lines there is at least some level of info he has been made aware of?
Hopefully it isn't too long before we get new rumours...
TheThreadsThatBindUs
Member
Oh no need to apologise. I realise that you weren't trying to catch me out. I just thought I'd put a caveat in there, since my memory of vague rumours derived from various Gaf insiders is hazy at best, lol.
TheThreadsThatBindUs
Member
That's the design cost. It shouldn't have any direct impact on console price, since AMD would mark it down as a sunk cost for R&D on their balance sheet.
It's essentially the investment AMD puts in for their 7nm/5nm node designs. The amount is amortised over the life of all products produced on that node, so desktop chips, console designs, mobile chips etc..
It's the reason MS and Sony go with IHVs like AMD in the first place, so that they don't have to bear the full brunt of these costs.
THE:MILKMAN
Member
Honestly I don't know how any of that works. I think very simply and thought Sony ask AMD to design the chip and AMD say yes sure...$400 million (or whatever) please!
TheThreadsThatBindUs
Member
Destiny 1 played perfectly well at 30fps... so not D2 isn't hindered by being 30 fps.
Sure it would play better at 60fps, but with development on a closed box like a console, with a fixed platform hardware target, 60fps doesn't just exist in a vacuum. Targeting 60fps invariably means compromises elsewhere, so a 60fps Destiny 2 on PS4 wouldn't be the same game as 30fps D2.
So no,... 60fps shouldn't be "standard" on consoles. If you desperately want 60fps so bad, buy a PC... nobody is stopping you.
Never ever happening.
I was thinking the same thing. All it takes is one publisher releasing a "next-gen upgrade expansion pack" DLC that becomes a big success to get the ball rolling.
That said, cross-gen games have coexisted with BC before. It's technically happening now on Xbox, since there are 360 games that are BC with Xbox One but also have Xbox One versions. The same thing happened when the PS2 firs came out, there were cross-gen PS1 and PS2 games (think Tony Hawk and a lot of sports games) even though PS2 was BC. Same thing technically happened when all PS3s were still BC with PS2 discs and there were cross-gen PS2/PS3 games.
The difference today is the absence of a major architecture shift between consoles and this whole "upgrade" path, where now the console market has opened up the idea of one console game SKU running at different parameters on on different hardware models. If something changes in the relationship between BC and cross-gen, I feel it will be Microsoft setting the trend with its own first party games. I could definitely see Halo 6 or something getting a free update for the next Xbox model after Xbox One X that improves its graphics at no extra charge. I could even see Microsoft going back and releasing those kinds of patches for all its first party Xbox One games.
I think Sony just waits. They have nothing to prove. The Pro does everything that most gamers outside high-end PCs care for.... markets these days react to threats to stay successful and they predict what threats might be around the corner to stymie their growth. Right now Sony could not give a shit about Xbox One X because the baseline is no competition.
While they will lose slight ground to third parties they won't lose that much, and what they do lose will easily be made up of superstar 10M possible sellers like Spiderman and marketing deals. There is simply no way to stop the PS4 train at this point. And if Spiderman succeeds--I think it will--the advent of a 200 dollar PS4 soon after could seriously destroy everyone around them.
Why go half way and piss off your Pro customers and confuse everyone else with some shitty upgrade?
I'm thinking 2020 here unless they really feel the need to get something out there.
While they will lose slight ground to third parties they won't lose that much, and what they do lose will easily be made up of superstar 10M possible sellers like Spiderman and marketing deals. There is simply no way to stop the PS4 train at this point. And if Spiderman succeeds--I think it will--the advent of a 200 dollar PS4 soon after could seriously destroy everyone around them.
Why go half way and piss off your Pro customers and confuse everyone else with some shitty upgrade?
I'm thinking 2020 here unless they really feel the need to get something out there.
geordiemp
Member
I am sure Sony dont want a consumer to start their collection again, as a Ps4 player could easily defect. Pubs made money on remasters, BUT allot of gamers swiched sides.
The sure way to keep players is by keeping the gamers collection. Would Sony risk loosing share ?
Last gen I went 360 to Ps4. If Xb1 was not so fucked up (water cooler Kinect and 720p) and had played 360 games, I would be a xb1 owner and not ps4.
Sony cant be that stupid, can they ?
If Sony dont do BC, I will go back to MS as they sure will.
I agree. Sony needs to monitor pc migration more than Xbox competition at this point. MS needs blockbuster games to draw people from PlayStation. Sony is already on record they noticed last gen that gamers left to PC as the generation aged. I think 2020 or xmas 2019 is what we can expect.
THE:MILKMAN
Member
I think 2019 seems to be the date being aimed for. Whether the 7nm node will be ready in time is my biggest question. Didn't Matt say "broken clocks and all that" meaning Pachter is right (about the date) but not because he necessarily knows anything?
The fact this thread exists suggests PS5 is aimed to come sooner rather than later IMO. 2020/21/22 just seems far too far away for me.
The Greek Freak
Member
Yup, I have a 4k tv and would take 1080/60 over 4k/30 in any game. Fuck it give me 720/60 over 4k/30
60 fps games both look and control better by a wide margin.
Shin
Banned
Vega will be announced tomorrow at Siggraph so we should have an idea about the price, performance has already been leaked so meh.
When they switch to 7nm next year the TDP should be down to 120-150w well within range of what can be used in an APU (as per RX480).
Settling for 10TF console in 2020 is kinda silly, Vega is 13.1TF that's around a GTX 1080 or below in some cases just goes to show how different power measurements are.
We would need something like a 14-16TF AMD GPU in a console to truly push native 4K with some room to spare, this ain't a nVidia card, not performance nor TDP wise.
We already got the sub-4K or fur-K or whatever you call it this generation, the next generation should push that boundry even further while being balanced CPU/GPU.
10TF would be what GTX 1070, a 4 year old card by 2020, that doesn't sound impressive or accurate at all and a let down at that in my honest opinion.
When they switch to 7nm next year the TDP should be down to 120-150w well within range of what can be used in an APU (as per RX480).
Settling for 10TF console in 2020 is kinda silly, Vega is 13.1TF that's around a GTX 1080 or below in some cases just goes to show how different power measurements are.
We would need something like a 14-16TF AMD GPU in a console to truly push native 4K with some room to spare, this ain't a nVidia card, not performance nor TDP wise.
We already got the sub-4K or fur-K or whatever you call it this generation, the next generation should push that boundry even further while being balanced CPU/GPU.
10TF would be what GTX 1070, a 4 year old card by 2020, that doesn't sound impressive or accurate at all and a let down at that in my honest opinion.
THE:MILKMAN
Member
I'm sure I've mentioned this before Shin but 120-150W is completely unrealistic for the PS5 APU to have. The whole launch PS4 and Pro use ~150W. The APU will also have at least 4 Ryzen cores.
Also if big Vega is ~300W TDP/TBP then a 7nm shrink won't enable a reduction to 120-150W surely? The slim PS4 and Xbox One S both saw in the range of a ~35% power reduction with the 16nm shrink and would expect similar for 7nm shrink of Vega i.e. from ~300W to ~195W for a straight node shrink.
The claimed power reductions of the likes of TSMC seem "optimistic" when you compare real world results....though I'm sure they could come up with a specific example.
Also if big Vega is ~300W TDP/TBP then a 7nm shrink won't enable a reduction to 120-150W surely? The slim PS4 and Xbox One S both saw in the range of a ~35% power reduction with the 16nm shrink and would expect similar for 7nm shrink of Vega i.e. from ~300W to ~195W for a straight node shrink.
The claimed power reductions of the likes of TSMC seem "optimistic" when you compare real world results....though I'm sure they could come up with a specific example.
Shin
Banned
Doesn't sound unrealistic to me, reminds me of the $399 or 8GB GDDR being impossible/unrealistic back then (in other words everything is possible).
50 or even 100w more is a reasonable expectation that would come with more power, 40-60% power reduction, I'd say TSMC/GloFlo knows more than us.
PS4 Pro power consumption matches the OG PS4 and in some cases lower also in other cases it's just above it, 1.84TF vs 4.2TF so....uhh?
As history would have it power consumption has gone up with every generation (generally speaking too lazy to research), so why would the next gen be any different?
THE:MILKMAN
Member
It is far too late and I've drunk far too much to look it up, Shin!
The main reason why neither Sony or Microsoft want to go back to or above what the 360/PS3 did is 1, RROD costing MS over a billion and 2, Sony losing ~$300 on each PS3 at launch because of it being over engineered to cool and power the beast within.
I'm interested if Xbox One X uses more or less power than the OG 360.
It wasn't really the 8GB that amazed people, it was the fact that it was GDDR5. Xbox One from the outset was 8GB but because it wasn't the trendy sort that even high-end GPUs only had 2/3GB of....
I still think a lot of the fan amazement over the RAM was jumped on by Sony PR. Not to say the 8GB wasn't a relatively late decision, just I think it was still months before the announcement of PS4 at least and dependent on the 4Gbit chips coming out.
The main reason why neither Sony or Microsoft want to go back to or above what the 360/PS3 did is 1, RROD costing MS over a billion and 2, Sony losing ~$300 on each PS3 at launch because of it being over engineered to cool and power the beast within.
I'm interested if Xbox One X uses more or less power than the OG 360.
It wasn't really the 8GB that amazed people, it was the fact that it was GDDR5. Xbox One from the outset was 8GB but because it wasn't the trendy sort that even high-end GPUs only had 2/3GB of....
I still think a lot of the fan amazement over the RAM was jumped on by Sony PR. Not to say the 8GB wasn't a relatively late decision, just I think it was still months before the announcement of PS4 at least and dependent on the 4Gbit chips coming out.
Trogdor1123
Member
I'm thinking I will wait on ps5 this time and wait for the pro version. I was a launch ps4 and pro buyer but I think I will wait this time. That is unless ps5 comes day and date with a demon souls remaster lol
AMD will do the heavy lifting on APU design because it'll sell outside the console space. Sony's input will be - as previously evidenced - in bringing custom features that aren't necessarily available on that standard design.
Which is the point in buying off-the-shelf components rather than going into chip design again.
THE:MILKMAN
Member
Cheers. Makes sense. I do wonder how much Sony have to pay AMD versus doing their own chip?
I'm not sure Vega is what PS5 would be based around. OG PS4 had Polaris features or features that went into Polaris at least? And Pro has Vega features, so PS5 should/could be Navi with "Next Gen" features?
StrongBlackVine
Banned
Navi considering we are talking 2019.
This.
Shin
Banned
So a year ago AMD launched RX 480 for around $299 or close to that at 5.1TF, a year later they launch Vega 56 11.3TF for $399 (which will soon drop in price knowing AMD).
That's a 6.2TF jump for about $100-150 more in just a year (leaving Vega TDP aside) and you guys want to tell me that 3 years from now we might not get a 14-16TF console?
After factoring the new architecture, 7nm / 7nm EUV, assuming it launches 2019 as scheduled by 2020, Sony wouldn't even have to customize it and it would reach 14-16TF.
You can be sure that it's TDP will be near Polaris with the die shrink and them designing everything from the ground up again, memory is less of an issue boost that GPU/CPU!
Also Vega uses HBM2 which definitely drove the price up for the card, I'm sure they had their reasons for using HBM to begin with, so it's really like a $50 difference in the end.
Let's see next year when they move Vega to 7nm and clock it higher how many teraflops it is then and it's power consumption, I'm more optimistic now damn debbie downers.
That's a 6.2TF jump for about $100-150 more in just a year (leaving Vega TDP aside) and you guys want to tell me that 3 years from now we might not get a 14-16TF console?
After factoring the new architecture, 7nm / 7nm EUV, assuming it launches 2019 as scheduled by 2020, Sony wouldn't even have to customize it and it would reach 14-16TF.
You can be sure that it's TDP will be near Polaris with the die shrink and them designing everything from the ground up again, memory is less of an issue boost that GPU/CPU!
Also Vega uses HBM2 which definitely drove the price up for the card, I'm sure they had their reasons for using HBM to begin with, so it's really like a $50 difference in the end.
Let's see next year when they move Vega to 7nm and clock it higher how many teraflops it is then and it's power consumption, I'm more optimistic now damn debbie downers.
Yup, this. I get why they had to dump PS3 (although I suspect they likely could easily add PS1 and 2 BC support via emulation ala Xbox, and don't because $$$, which is slightly harder to forgive...), and of course Microsoft didn't offer any BC with X1 initially either, but things will likely be different next time around. If Microsoft supports launches Xbox 2 with BC for X1, 360, and OG Xbox all at launch, and PS5 just has a typically pitiful line-up of launch PS5 games... I think that would be a horrific start to next-gen for Sony, and a huge advantage for MS, especially since with the advent of games as a Service, online features, etc. people would likely still be interested in playing X1 games at the very least. PS4 BC has to happen, IMO.
dano1
A Sheep
And that's fine YOU want to wait. But I'm ready to upgrade every 3 years.
I'm sure you didn't buy the PS4 Pro but I'm loving mine!
Correction. It was $199.99 for 4GB, and $229.99 for the 8GB.
I'm hoping the PS5 comes out in 2019. I don't want to wait 3-4 years for next gen. :/
edit not 34 years.
Shin
Banned
I saw that on Videocardz, thought they adjusted it to current price (in Netherlands it was around 300+- according to a price watch).
Funny enough that's exactly $30 difference for 4GB, the same $30 AMD said they paid for 4GB to Eurogamer (coinsidence?).
If that's the case they could have sold Vega 56 cheaper than $399 if they went with GDDR5x instead of HBM2, it's all good though.
Hopefully the CU count on Navi will go up because even at 64CU you'd be around 10.2TF I don't expect higher memory clock speeds than 1.3GHz max in a console.
I could see 72-80 CUs for a Navi GPU.
BTW what is the tflop/CU equation again? I always blank on it.
I guess it would depend on what volume deal Sony has in place. Being able to say with some degree of confidence you're ordering 60-100mn units of something over a 6-7 year period would give you some decent negotiating leverage!
It's also why looking at retail prices for graphics cards & memory is pointless. You need to look at OEM catalogues and trade pricing, as I imagine gfx cards especially have a huge markup at retail.
Miniature Kaiju
Member
80 CUs, 8 disabled for redundancy and yields = 72 active CUs. With memory clocked from 1300 to 1500MHz, that would give us 12 to 13,8TFlops. That's feasible, I feel.