At least for Microsoft, is there a chance we could see a reappearance of some form of Skybridge? Four Zen cores and four ARM cores combined with a Vega GPU. Save some space and cost. They seem to be doing well getting Windows code working on ARM. Also I've been wondering how much help a HPU processor would be in collision detection and physics on a console. Especially if there is a possibility of building Havoc into the hardware like they did with DX12 on One X. If it would be useful would it be cost effective? In 2020.
Predict: Next gen console tech (9th iteration and 10th iteration edition) [2014 - 2017]
- Thread starter Shifty Geezer
- Start date
- Status
Skybridge wasn't x86 & ARM on one APU/CPU, it was just socket/pin-compatible x86 & ARM APUs
Now that I've been reading more on that Navi thread...
This is where I predicting now:
PS5 no BC -> 1 APU, Zen CPU, GPU good specs, could release as early as 2020
XB2 native BC -> 1 APU, Zen CPU, Multi GPU, big/little setup, Xbox 1 family GPU + 1 Bigger one (1 year after PS5)
The challenges faced by BC I think speak for themselves when we look at how both 4Pro and 1X both do hardware downsizing to accommodate older builds, 4Pro and 1X both drop to 1/2 GPU usage to accommodate older native titles, both are respectively of nearly the same GCN family (maybe someone can clarify if 4Pro and 1X are considered GCN 2 (with customizations) or they moved directly to GCN 4)
This is where I predicting now:
PS5 no BC -> 1 APU, Zen CPU, GPU good specs, could release as early as 2020
XB2 native BC -> 1 APU, Zen CPU, Multi GPU, big/little setup, Xbox 1 family GPU + 1 Bigger one (1 year after PS5)
The challenges faced by BC I think speak for themselves when we look at how both 4Pro and 1X both do hardware downsizing to accommodate older builds, 4Pro and 1X both drop to 1/2 GPU usage to accommodate older native titles, both are respectively of nearly the same GCN family (maybe someone can clarify if 4Pro and 1X are considered GCN 2 (with customizations) or they moved directly to GCN 4)
Thanks for the response.
phoenix_chipset
Regular
Knowing Sony I think it's highly unlikely they'll offer full native BC for the ps4, though the chances are higher than they were for Ps3 BC for sure. They won't want to do what's necessary for that ; keeping the old cpu architecture around and incorporating the old gpu into the new one i.e. Wii U.
Not to mention they now have 2 different pieces of hardware to mirror thanks to the PS4 pro. They're probably not going to come up with something as clever as the Xbox one's BC method for 360 either.
MS will probably offer some sort of BC for their next machine though not native.
Not to mention they now have 2 different pieces of hardware to mirror thanks to the PS4 pro. They're probably not going to come up with something as clever as the Xbox one's BC method for 360 either.
MS will probably offer some sort of BC for their next machine though not native.
Last edited:
might be costs, which is why they continue to stick with AMD, and since at the console level, a lot of the advantages that nvidia has over AMD (particularly driver support for APIs), that goes away, and looking purely at feature set, nvidia offers only a handful more features.
phoenix_chipset
Regular
Plus with Zen they've bridged the gap in IPC so there really wouldn't be any reason to go intel. More money for same performance and zen is better in power consumption as well. AMD are superior in low power offerings ; the consoles won't likely get full fat zen cores, they'll get zen lite. Which will still be a huge improvement over jaguar of course.
One tangible benefit they'd get with Nvidia is lower power consumption but they probably still can't match AMD in price.
D
Deleted member 13524
Guest
Why would a x86 APU from AMD not be backwards compatible with the PS4 or even the PS4Pro?
Instruction set on the new APU would most probably (99% sure IMO) be a superset of the old. For both GPU and CPU..
Not sure if it's abstract enough. I think the CPU should be okay for BC. But looking at how much they had to concede to get the GPU side of things to work, for 4Pro, and for 1X I don't think it's as simple as "software emulation" When they move to Navi.
It might be worth discussing again perhaps in another thread if it get technical enough, but for 4Pro they effectively doubled the existing PS4 and could on/off for native non pro titles. And boost mode as I understand it, it just leveraging the full GPU and CPU clockspeeds, it doesn't unlock the other 1/2.
For 1X for certain SDKs they can only enable 1/2 for the game. For other SDKs they can enable the full amount. And this is when 1X is still effectively using the exact same tech.
So there seems to be some challenges at least. The move to navi could make BC harder I think. Perhaps not so hard, but not so straight forward either.
Last edited:
I really hope Sony wait it out for Ryzen in 7nm and Navi, that'd put the Ps5 somewhere in late 2019 assuming GloFo hit their target for late 2018 release of 7nm products. Would be awesome to see 8 cores/16 threads Ryzen 2 at 3.0-3.6 GHz (7nm desktop Ryzen should be close to 5 GHz so 3.0-3.6 is a safe assumption for laptop/console CPUs for maximum efficiency).
Nvidia has them beat badly in power consumption right now. Unless something changes in terms of process improvements, I'm starting to wonder if one of them makes the switch. I guess the difference is AMD can supply an x86 APU where Nvidia cannot. AMD needs to make some huge improvements in Navi for power consumption, either through architecture or process, otherwise it starts to look tough to put anything significant in a console with a power budget.
I would think that Sony would rather reap the benefits of the PS4 specific IP in their APUs and the software infrastructure that relies upon it rather than worrying as much about power consumption which is important of course. Reworking similar things into Nvidia would cost time and money which could be spent on other things that distinguish them from the PC/XBX and engineering around the power inefficiencies that plague AMD at this point.
Still getting back tens of watts to play with would be pretty sweet.
Still getting back tens of watts to play with would be pretty sweet.
I can't think of much that's really specific Sony IP in the APU, and if there is any I don't think that it's worth prioritizing over anything else. Other than the id buffer, and maybe some small bus tweaks, what else is there?
To me, the real issue with going Nvidia is you have to go intel on the cpu, if you want to stay x86. Seems unlikely that works out to be favourable cost-wise.
Prophecy2k
Veteran
When we talk about power inefficiencies plaguing AMD, aren't we just talking about Vega (which seemed to prioritize higher clocks to achieve improved performance)? Prior to Vega, hasn't AMD's GCN-based GPU compared favorably with NV in terms of perf./W?
PS5 will most likely be based on Navi, which from the sounds of things should be a step change in architecture from Vega. So are we really expecting Navi to be plagued with the same issues as Vega?
PS5 will most likely be based on Navi, which from the sounds of things should be a step change in architecture from Vega. So are we really expecting Navi to be plagued with the same issues as Vega?
I may be remembering incorrectly, but I think GTX1060 is much better in terms of power efficiency than RX480/RX580, and they're roughly competitors.
Or ARM.
Definitely one of the largest factors is that power efficiency.
This is where I feel MS has slightly more breathing room if they want to switch things up if things are not working out for next gen.
I don't know how far DX12 is abstracted the move to nvidia would pose other challenges on the BC front. This would definitely make it harder considering that a lot of those console gains have been code optimized for GCN.
Then again, if DX12 is proper, it's entirely possible that with some work, it might be able to run the DX12 based games.
The other factor is that DX12 comes with explicit async multi-adaptor. They've had this feature for a while and as sebbbi writes, the move towards reconstruction techniques means that AFR is essentially dead on multi-GPU setups. So the idea here is that, if they need to, MS can push for an older setup, mixed in a with a new one, and the developers would have to develop with explicit async multi-gpu. Overall it would probably end up pretty close to the power of a single GPU setup, but with the possibilities of some BC there.
So imagine, Ryzen + Scorpio GPU (ideal, since it's profiled for XB1 games) as the new APU, with an off chip nvidia GPU for instance as a possible configuration. That's a legitimate setup that we could have on PCs in a few years time, and really its just up to developers to make support of being able to mix the two together. Move towards a sufficient amount of shared video memory, and plop in a weaker 6-9 TF nvidia GPU, and your totals are just north of 12. As opposed to a singular 12-15TF GPU.
It doesn't necessarily need to be as cut and dry as it once was, at least not with this move to shifting the whole industry towards DX12. If MS is successful in getting everyone onto DX12, it opens up their hardware selection choices.
Last edited:
@iroboto But ARM cpus run x86 code in some kind of emulation, no? I guess they could run new games in native ARM and use emulation for backwards compatibility.
I'm not sure having an APU + discrete GPU really makes sense from a cost perspective. That's what PS3 tried to do at launch, and it was cut out. I had one of the original fats with the PS2 hardware BC. It would make more sense to me if they went some kind of Nvidia SOC with ARM and just try to make the backwards compatibility work the way Microsoft did it for 360 with Xbox One. Would need to be a pretty significant leap in processing power to make that happen, but with Nvidia they might be able to fit a bigger gpu under the thermal, power constraints of a console.
I'm not sure having an APU + discrete GPU really makes sense from a cost perspective. That's what PS3 tried to do at launch, and it was cut out. I had one of the original fats with the PS2 hardware BC. It would make more sense to me if they went some kind of Nvidia SOC with ARM and just try to make the backwards compatibility work the way Microsoft did it for 360 with Xbox One. Would need to be a pretty significant leap in processing power to make that happen, but with Nvidia they might be able to fit a bigger gpu under the thermal, power constraints of a console.
Wasn't it a cost issue because the PS2 hardware was only there to support PS2 games?
EMA enables the mixture of 2 completely separate GPUs to work together for SFR.
IIRC from being at the presentation, Max said that the GPUs have fences/triggers, that can pass work to the other GPU without having to go through the CPU. So even in a big little setup, the big GPU can do the render work and trigger the integrated GPU to do post processing without the CPU required for intervening.
edit: found it:
https://software.intel.com/en-us/articles/multi-adapter-support-in-directx-12
Code:
The above diagram shows the use of three queues to facilitate copying into cross-adapter resources. This is the technique used in this sample and showcases the following steps:
[LIST=1]
[*]Queue 1 on GPU A and Queue 1 on GPU B render portions of a 3D scene in parallel.
[*]When rendering is complete, Queue 1 signals, allowing Queue 2 to begin copying.
[*]Queue 2 copies the rendered scene into a cross-adapter resource and signals.
[*]Queue 1 on GPU B waits for Queue 2 on GPU A to signal and combines both rendered scenes into the final output.
[/LIST]http://www.anandtech.com/show/9740/directx-12-geforce-plus-radeon-mgpu-preview
https://developer.nvidia.com/explicit-multi-gpu-programming-directx-12-part-2
Yes
I don't quite understand the UWP aspect of things fully, I may have made some poor assumptions on compilation.
- Status