With PSX being a fan event, I'm not sure I see Sony revealing or even teasing new hardware there. You want press in attendance for that kind of thing and that takes up room meant for fans.
-
Ever wanted an RSS feed of all your favorite gaming news sites? Go check out our new Gaming Headlines feed! Read more about it here.
-
We have made minor adjustments to how the search bar works on ResetEra. You can read about the changes here.
PS5 and next Xbox launch speculation - Post E3 2018
- Thread starter Phoenix Splash
- Start date
With PSX being a fan event, I'm not sure I see Sony revealing or even teasing new hardware there. You want press in attendance for that kind of thing and that takes up room meant for fans.
I don't know if it's worth it on a console SoC at 7nm. I definitely think it will make sense at 3nm where you can fit 4 GPUs on one console sized SoC.
AMDs current GCN architecture seems to have hit a scalability wall at around 56 CUs, with Vega 56s clocked to Vega 64 speeds offering virtually identical performance in games.
Supposedly Navi will offer improved scalability, but we can only guess what that means for right now. I don't take AMD's statements so far as a complete 100% denial that Navi will feature an MCM design. In fact, I am suspicious that the 7nm Vega Radeon Instinct releasing later this year for the HPC and Machine Learning markets might be a dual GPU MCM design.
But Sony and Microsoft will not want a huge (over 400mm^2) SoC. I think ~56 Active CUs out of ~64 total with a high clock speed (1500+ mhz) is most likely. They could maybe cram two 32/36 CU GPUs on one SoC, but I'm pretty sure that would cost more and with lower clock speeds and potential issues/learning curve for Split Frame Rendering I am skeptical that that would lead to improved performance any time soon.
We've been listening to companies for at least a decade who've "solved" game streaming .
I think 2020 is all but confirmed after Sony announced the fiscal plans thing. I predict Xbox will be slightly stronger this time around(because it matters more to them imo), but barely... like ps4/x1 original. I am having a hard time deciding which to buy first this time around..I didn't play as many ps4 exclusives as i'd have liked.
Yep. There is nothing being said in the last few months about streaming that are actually new. That is why I no longer care about any more "developments" until they actually have something to demonstrate. All the talking and theories are over, what matters is practical applications. And so far, we have seen none.
The difference between og ps4/x1 was decent. Doubt we'll see anything like that next gen.
To some I guess. 1080p vs 900p threads were abundant indeed. I do think the difference will be smaller than that like you said.
Yeah, I mean on paper Practicality the difference was slight.
Hmm I dunno about that, shit was noticeably blurry on the Xbone, plus some games were running at higher frame rates on PS4, like Tomb Raider DE for example.
Like you guys I don't see there being a huge difference next time, unless secret/special sauce does wonders right out of the gate.
I guess we can agree that we'll see less then the ~40% difference we saw this time round and whatever difference it is will be proportionately less obvious ?
It was blurry on games like BF4 imo..where it was 900p vs 720p.. 720p was just low. but for most games I didn't notice much difference on my set. Like I said though, it will be close, and I think the xbox will be slightly stronger this time.. They proved they want to be known as the stronger one..like with ps4pro / xb1x.
I expect that with console optimizations a 20% difference won't be noticable in most games. I don't think we will see many games with a clear cut 1800p vs. 4K difference. Many games will have dynamic resolution or use temporal injection.
We'll get articles from digital foundry saying stuff like in this scene we found 3600x2160 on Scarlett and 3200x2160 on PS5, but in this next scene both were full 4k.
I don't think it will be as clear cut as 900 vs 1080. And stories that aren't simple just don't have the same resonance with most games media outlets. We're already seeing that with how poorly most jornos understand checkerboarding.
The latency solution linked a few pages back still requires a low-fi version of the game running locally, so it needs local storage.
I'd argue a HDD would be overkill. I'd guess local flash would be used as minimum hdd size/price is determined by global hdd market/manufacturing and doesn't really come down in price and 500GB is likely unnecessary.
In fact, with current consoles being able to store and run games off external hdds, I wonder if a hddless ps4/xbox1s sku could get down to ~$100.
The narrow Front-End of current GCN designs can become a bottleneck if the workload isn't big enough or the duration too short, then the dispatch rate is too low to fill a big shader array quick enough:
https://32ipi028l5q82yhj72224m8j-wp...18_sponsored_engine_optimization_hot_lap.pptx
And Vega20 is a single GPU chip, the CEO Lisa Su presented a V20 sample at the Computex stage 2018:
For me personally 1080p versus 900p was a pretty sizeable difference on my TV but something like 2160p versus 1800p is almost impossible for me to tell the difference at my viewing distance even though the difference in pixel count is much larger there. Combine that with dynamic res and reconstruction (perhaps even both at the same time) and the differences in IQ between platforms will be pretty small.
That's about 11 tflops, right? I always screw up the math.
Last edited:
No way -- I cannot see Sony teasing the PS5 at PSX in December, even if Sony does release it Holiday 2019 (which I still really doubt (but won't rule out completely). If Sony intends to launch in late 2019, then the earliest I'd expect an announcement/reveal is February 2019. If Sony intends to release it for Holiday 2020, again, I wouldn't expect an announcement until February 2020. I think Sony will try to follow the same schedule they used in 2013 with PS4. It worked extremely well for them, so why wouldn't they try to repeat it almost exactly in either 2019 or 2020? Bottom line, Sony most likely won't give us even a glimpse of PS5 until we're into the same year it comes out.
Also, we still do not know as much about PS5 in 2018 as we knew about PS4 in 2012 with the Orbis spec/architecture leak.
There were very specific pieces of information about Orbis/PS4, the CPU, GPU, memory bus details and even the name of the APU (Liverpool). So if PS5 is really coming in late 2019, we'd probably already know for sure by now.
I think they might even wait a little bit longer than February. They've said they want a shorter time between announcement and release of hardware. Obviously they can't do the tight timing of the Pro but something like an April reveal of the console and new features with an E3 focused on brand new games could work. MS tried that with the Xbox One but completely botched the messaging.
It would be fun if they teased the PS5 at PSX with something like 'fully backwards compatible with all your PS4 and PS4 Pro games. Full announcement in 2019'.
They won't announce it or hint at it during the holiday season. That could kill PS4 sales for the rest of the year.
That would make for a range of 10.03520 to 10.75200 tflops.
I'm thinking it could go as high as 1590 mhz though, which would give you 11.39712 tflops.
Yeah, fair enough. I was more getting at that they can announce and reveal all kinds of things without having to talk about specs or show the box.
You are probably right.
But did Lisa Su state that Vega 20 isn't an MCM?
I've saw that picture back during Computex. I don't think it alone confirms that Vega 20 is not an MCM. Just look at a delidded Ryzen chip with 2 CCXs. It looks pretty similar minus the HBM.
All I am saying is that I'm suspicious...it's probably just me being crazy, but just maaaybe it's not.
There is no heatspreader, you see the bare chips, one V20 GPU and four HBM2 stacks.
As for Ryzen which uses one Zeppelin die, it's just one chip which internally groups its functional blocks into two modules and connects them.
But it's not an MCM it's just one chip on a package.
Threadripper or EPYC products are using MCMs, 4 Zeppelin dies on a package:
I'm not talking about whether or not they've truly "solved" game streaming here. I'm just saying that they'll finally have a service that makes sense for the rumored hardware. They didn't have one at the time a similar device was rumored for the Xbox One
There is no heatspreader, you see the bare chips, one V20 GPU and four HBM2 stacks.
As for Ryzen which uses one Zeppelin die, it's just one chip which internally groups its functional blocks into two modules and connects them.
But it's not an MCM it's just one chip on a package.
Threadripper or EPYC products are using MCMs, 4 Zeppelin dies on a package:
Yup. and on the current Threadripper which were launched last year, 2 of the 4 Zeppelin dies are active.
That may change with the upcoming Threadripper 2000 series based on Zen+ coming soon. Upto 32 cores with all 4 Zeppelin dies active.
https://www.anandtech.com/show/12906/amd-reveals-threadripper-2-up-to-32-cores-250w-x399-refresh
There will likely be some Threadripper 2000 processors that have 3 active Zeppelin dies, for 24 core versions.
https://www.anandtech.com/show/12906/amd-reveals-threadripper-2-up-to-32-cores-250w-x399-refresh
Not to be confused with Zen 2 architecture, which will be the basis for the RyZen 3000 and Threadripper 3000 series next year.
The best is that next-gen consoles will use Zen 2 architecture, but nothing confirmed. It just makes sense given Zen 2 is 7nm (Navi GPUs will be also) and next-gen consoles will be made using 7nm
There is no heatspreader, you see the bare chips, one V20 GPU and four HBM2 stacks.
As for Ryzen which uses one Zeppelin die, it's just one chip which internally groups its functional blocks into two modules and connects them.
But it's not an MCM it's just one chip on a package.
Threadripper or EPYC products are using MCMs, 4 Zeppelin dies on a package:
Okay, you're right.
What I'm thinking of wouldn't really be a Multi-Chip Module. I'm think of more how Ryzen has two CCXs on the same die.
Two GPUs on the same die. Referring to that as an MCM was incorrect.
Reading the patent application for Split Frame Rendering "The Processor includes a central processing unit (CPR), a graphics processing unit (GPU), a CPU and GPU located on the same die, or one or more processor cores, wherein each processor core may be a CPU or a GPU."
thats disappointing, was hopping for higher clock speeds, not just extra cores, oh well.Yup. and on the current Threadripper which were launched last year, 2 of the 4 Zeppelin dies are active.
That may change with the upcoming Threadripper 2000 series based on Zen+ coming soon. Upto 32 cores with all 4 Zeppelin dies active.
https://www.anandtech.com/show/12906/amd-reveals-threadripper-2-up-to-32-cores-250w-x399-refresh
There will likely be some Threadripper 2000 processors that have 3 active Zeppelin dies, for 24 core versions.
https://www.anandtech.com/show/12906/amd-reveals-threadripper-2-up-to-32-cores-250w-x399-refresh
Not to be confused with Zen 2 architecture, which will be the basis for the RyZen 3000 and Threadripper 3000 series next year.
The best is that next-gen consoles will use Zen 2 architecture, but nothing confirmed. It just makes sense given Zen 2 is 7nm (Navi GPUs will be also) and next-gen consoles will be made using 7nm
You mean a more rigid slicing of the units and performance degradation for the sake of chip complexity?
Thats like really a cost optimization, where I am due to the parallel nature of GPUs and better scaling mechanism in comparison to CPUs wouldn't welcome such grouping.
Edit: Well it does help to just google the paper and read it. :)
http://www.freepatentsonline.com/20180211435.pdf
By skimming over it, it looks pretty much like the slicing AMD currently does but with better work distrubution.
As far as I understand it AMD currently slices up their GPUs to 1-4 pieces which AMD calls a Shader-Engine, each SE has one Geometry/Rasterizer-Engine which are connected through a crossbar and can be synchronized but work distribution seem to be rather stiff or was pretty bad in the past for AMD, where Nvidias Front-End is/was much better.
After the Fixed-Function-Frontend are the Compute Units, the ROPs and L2$ tiles connected to corresponding memory channels.
The improved approach seems to have multiple Command Processor who can feed the Shader-Engines with 3D or Compute-Queues, currently there is only one Graphics Command Processor feeding all of them with 3D or Compute-Queues, there should be another Processor which can only handle 3D-Queues.
Since GCN2 there are two compute processors which have up to 4 compute pipes and feed all SEs/CUs with Compute-Jobs.
The SEs should have better mechanism to devide and distrubute geometry/rasterization-work between each other.
Last edited:
Sony are expecting a drop in sales this fall. Which is odd considering the games being released. You would think Spider-Man and RDR2 would drive sales. Them teasing PS5 could be a reason why they expect a drop in sales or its just a natural drop as we enter the back end of the life cycle. who knows.
If you are talking about console sales, I don't think Sony expects many customers left who doesn't already own a PS4 at current price points. There are obviously people who would be interested after another price drop, but that is likely being reserved for later.
Yep, even without the external capability you may recall that the PS3 super slim was available in a diskless SKU.
to be 100% clear for that setup, you still need the assets though, its not running wireframes, but ether running a low res version at full speed, or high detaail version at atleast 1/3rd speed. ether way you would still need all the assets locally.
This patent is for the post-process scaling future when there would be no other option but to use multiple dies (stacked or not) to further improve performance. It's highly unlikely that they'll go this weird route of SFR on a single GPU since this is basically unnecessary.
I also kinda doubt that GCN's shader core scaling issues are due to global scheduling but we'll have to wait for Navi to see what they'll change there to be sure.
I am not sure your first sentence have anything to do with your second. Sony isn't going to either sit on PS5 production OR rush it out ahead of schedule. And they certainly aren't going to do it because of the sale expections.
I find it interesting that ERA likes to believe console manufacturers can release consoles at will. Remember, the last time someone did that, it was the Sega Saturn release in the US. It was released 4 months ahead of announced schedule, and the result is extremely unhappy retailers. All to try to headoff the playstation. And we know who ended up winning.
You don't do that. EVER. You can surprise your customers, but never surprise your retailers.
Ps4 and Xb1 were both revealed and launched in the same year. The announcement dates in the quote was entirely reasonable, assuming it's what Sony and MS have been planning for behind the scenes for years.
Sega was a very schizophrenic company, with various competing factions. That Saturn situation doesn't really relate here at all.
I was merely trying to say that as of now, the sales of PS4 don't mandate a 2019 release date for PS5 and since Sony is almost certainly aware of sales, it's unlikely that they're planning a 2019 launch. I believe there's the connection between my first and my second sentence.
If the planning had been going on for years, then the sales expectations for PS4 this year should have no relevance on the final release date for PS5.
I agree, for a 2020 holiday release those dates seem reasonable. I think because we don't have any definite leaks yet it seems unreasonable.
We don't know how clocks are going to scale on 7nm, and we don't know what yield and die size will look like either. Maybe they go to 300mm^2 and crank the clocks (and voltage) or go big at 400mm^2 to bring down clocks and voltage.
From Seeking Alpha article on AMD's last quarter revenue. Is this a Sony projected ramp in sales for the PS4.3 Semi-accurate is predicting?
AMD has a new accounting scheme that balances what used to be Q3 Semi-custom for holiday sales across the whole year. It's also possible that the semi-custom increase could be an Atari or Google console Purchase order.
Last financial figure I saw was PS4 sales down $700K for last year.
From the 2015 agenda @ Efficientgaming.eu:
Plans in 2015 included 2019 power caps which seems to me to indicate Sony had plans to release a console iteration in 2019. In the Semi-accurate article linked in the OP, Charlie is stating that in 2013 he predicted a PS4.1/4.2/4.3 etc. with the PS4.3 being the 2019 console. Combining this with the statement in the same eu site (2017 file since removed) that after the next or 2019 console some other way to define consoles (BEYOND 4k resolution) will be needed.
What could they use/need to determine the 2019 date from 2015?
HDMI 2.1 spec release (late 2017 so useable 2018) Sony is a Founder of the HDMI org. 2012 Sony CTO article mentioned 8K and 300FPS which is only possible with HDMI 2.1
AMD GPU iterations (Navi (2016? Sony worked with AMD on Navi features) and AMD's SME/SVE (2016)
GDDR6 memory (2018)
Node plans by TSMC and Global Foundries (2018 - 2019)
Why do you think so? That would seem logical if real world GCN performance saw linear scaling with more CUs.
But, that's really not the case. If you compare a Vega 64 to a Vega 56 at identical clocks they have virtually identical performance in dozens of different games.
It's very difficult for current software to keep all of those CUs busy with only 4 Shader Engines and limited bandwidth. I'm sure Naughty Dog or Guerrilla could figure it out, but those studios are going to produce grade A games from a technical standpoint regardless.
Fewer CUs with a higher clock speed is better for winning cross-platform comparisons PS5 vs Scarlett. And most games are cross platform anyway.
Of course, if Navi overhauls the GCN architecture who knows what will be best.
GCN doesn't scale that well with neither CU counts nor clocks, though - actual perf per watt for Vega plateaus pretty hard as you go higher in power dissipation. The design is inefficient all around. It's all gut feeling, but I think the stated goal of "better scalability" for Navi does mean an overhaul to the front end, schedulers up, for better work division efficiency.
Even the Xbox One X is only running at 1172 Mhz. Vega scales quite a lot higher than that across a range of games.
Take a look at these articles fro Gamer's Nexus.
https://www.gamersnexus.net/guides/3040-amd-vega-56-hybrid-results-1742mhz-400w-power/page-2
https://www.gamersnexus.net/guides/3053-vega-64-vs-vega-56-clock-for-clock-shader-differences
In real games they see no benefit from Vega 64s extra CUs, but overclocking the 56 (even without overclocking the HBM) can provide sizable performance gains in some games.
Yeah, this too. I think both are prepared with a date, with time to practice their launch talking points, last minute changes if possible. (PS4 ram situation) With preparation to launch sooner if need be.
Was shocking to hear MS launched the XBO sooner than MS wanted, odd since they launched before the PS3.