Next Generation Hardware Speculation with a Technical Spin [2018]

[temesgen]

The only justification for Microsoft going with NVIDIA the next generation would be to leapfrog Sony on the GPU front. As NVIDIA has better performance per watt and better performance per mm, as well the faster architectures on PC. But that doesn't necessarily translate to a better custom GPU design for consoles. It might not matter at all.

I'd expect MS to have more powerful hardware this time but I doubt it will have anything to do with Nvidia because 1) One X would provide really good experience for pushing the engineering in a consumer device which will help a lot. 2) MS hasn't put as much into first party titles as Sony so there's a stronger need to provide third party experience that's best as well as online services which are more refined. 3) MS has more money and can afford push more in this regard. 4) MS vision: BC, online services, development tools are somewhat ready now. MS can simply wait for hardware that is viable and release next console and fill in any software gaps with BC till third party or first party offerings are ready.

Sony will likely be comfortable releasing a machine that is strong graphically but relying more on first party to drive interest. Further I think Sony is going to have a real issue if PS5 is not BC, I think some of their budget is going to have to go towards developing that capability. And I'm not sure Sony has tools as far along as MS which would potentially eat even more into a budget. And keep in mind Sony will have multiple teams working in first party which will also compete for cash.

In any case I don't think MS needs to anything radical like partner with Nvidia to gain a hardware advantage. They can simply build a larger APU and devote dollars to keeping it cool and quiet.

 

[chris1515]

That maybe true, but how was this even possible in the first place without a hardware design that accommodates these shaders? You can think of Task and Mesh shaders as a hardware accelerated compute solution for Geometry. Without them, running Tessellation on regular compute units would a performance suicide.

https://twitter.com/i/web/status/1042282150312194048

And I heard non happy developer with hardware tesselator not only @sebbbi

Geometry shader and hardware tesselator are not favorite functionnality of rendering programmer non flexible enough or non performant enough for geometry shader out of Intel implementation...

 

[chris1515]

https://twitter.com/i/web/status/976929292193026049

For example...

EDIT:

Another example:

http://aras-p.info/blog/2018/03/21/Random-Thoughts-on-Raytracing/

On the other hand, as Intern Department quipped, DirectX has a long history of “revolutionary” features that turned out to be duds too. DX7 retained mode, DX8 Matrox tessellation, DX9 ATI tessellation, DX10 geometry shaders & removal of FP16, DX11 shader interfaces, deferred contexts etc.

 

[Shifty Geezer]

We have games with planar, cubed, SSAO etc type reflections and we've seen a game with voxel cone tracing, I'm not sure what you're trying to imply here. Do they work?
Sure.
But we're talking about getting to next generation graphics not just more of the same right?

No, we're talking about what hardware needs to be in the next consoles. RT effects will be better quality. But if they're also slower and lower resolution? Reflections in racing game rear-view mirrors were 'perfect' and didn't need ray-tracing, because a hack could be used. As more power becomes available, better hacks become available.

When you compare the reflections in this demo with RTX on vs when RTX is off.

That's a demo designed to showcase raytracing. Find a best-in-class rasterised demo using a monster GPU that's happy to render 1080p30 to dedicate more time to reflections, instead of wanting as quick as possible screen-space solutions to keep the framerate up, and you'll have a comparable test. RT will be better, of course, but we don't know how close rasterisation will take us.

We're so used to baked lighting and non GI solutions, that we probably think it looks correct.

Definitely not!

Swapping sides for a moment though, I think the argument in favour of RT and ML should look at the wider picture. What if ML can be used in game for AI and things? If ML can be justified for its inclusion, then the package deal of compute+ML+memory search structures would enable denoised RT for those games that want it, but other options for other games. Is the full car-driving ML really needed for that, or could we get away with the equivalent of a few tensor cores for denoising? If ray acceleration is 10% of silicon and ML is 10%, they'd we very worth including to give devs options. Although there's the old economic issue of games not targeting this hardware if it's not everywhere, and better AI is a tough sell. Better physics is an easy sell.

 

[chris1515]

Geometry shader are not SIMD friendly... Crazy on a massively parrallel architecture like a GPU...

 

[troyan]

That's a demo designed to showcase raytracing. Find a best-in-class rasterised demo using a monster GPU that's happy to render 1080p30 to dedicate more time to reflections, instead of wanting as quick as possible screen-space solutions to keep the framerate up, and you'll have a comparable test. RT will be better, of course, but we don't know how close rasterisation will take us.

This is not a demo. It is real time in a real game with gaming assets. nVidias "SOL" demo is a demo...

 

[Xbat]

MS has more money and can afford push more in this regard.

Yes MS has more money but they have other more important divisions to there business. PlayStation has become one of if not the most important division for Sony so I wouldn't bet on it that Microsoft will pour more into the next console than Sony.

 

[temesgen]

Yes MS has more money but they have other more important divisions to there business. PlayStation has become one of if not the most important division for Sony so I wouldn't bet on it that Microsoft will pour more into the next console than Sony.

Again MS can get more powerful hardware without spending more if we look at the total cost of launching next generation consoles. MS won't spend as much on software or BC which leaves more dollars in the budget for hardware. I think they won't have as much to spend on tools either and the spend on tools is spread across PC too.

 

[Xbat]

Again MS can get more powerful hardware without spending more if we look at the total cost of launching next generation consoles. MS won't spend as much on software or BC which leaves more dollars in the budget for hardware. I think they won't have as much to spend on tools either and the spend on tools is spread across PC too.

My point being that Sony might spend more than Microsoft in there development of a next gen console, PlayStation division being so important to Sony. I also feel you making to much of the software costs.

 

[vipa899]

My point being that Sony might spend more than Microsoft in there development of a next gen console

Like with ps3?

 

[Xbat]

Like with ps3?

Yup exactly.

 

[Deleted member 13524]

This is not a demo. It is real time in a real game with gaming assets.

There are RTRT games for RTX cards out there, as of 22/10/2018?
Where??

 

[OCASM]

I'm not even talking about exotic renderers, but the fact that the top AAA engines often go through massive changes between the start and end of a console generation. They can do that due to the flexible nature of modern GPUs and in the PS3's case, the cell SPUs made up for the massive shortcomings of the GPU.

The next console generation, if one console goes with fixed function hardware that is faster at RT at the start of the generation versus a more general purpose GPU that is slower at RT at the start of the generation, it wouldn't surprise me at all if the second one ended up faster at RT by the end of the generation than the first one.

5-8 years is a long time to be stuck with stagnant (fixed function) RT hardware if it ends up being non-optimal for the most efficient RT algorithms.

Right now, no one knows what will end up being the most optimal way to do RT in games. Betting on fixed function hardware is like flipping a coin. If you win, good for you. If you lose, you just consigned yourself to the bottom of the barrel, because there isn't anything you can do about it.

Regards,
SB

RTX allows for a huge variety of algorithms because the fixed function hardware only accelerates the generic parts of RT. What developers do with the rays they trace is still entirely up to them.

 

[iroboto]

No, we're talking about what hardware needs to be in the next consoles. RT effects will be better quality. But if they're also slower and lower resolution? Reflections in racing game rear-view mirrors were 'perfect' and didn't need ray-tracing, because a hack could be used. As more power becomes available, better hacks become available.

I doubt it's linear though. It's highly unlikely that you can double up the CU and call it done. When you're reflecting from multiple angles from multiple areas, rasterization still falls apart. The thing with all algorithms is that in order for them to be optimal, they have an optimal use case that works with it. As soon as you surpass the limits of that use case, that algorithm gets entirely inefficient or unusable or no longer reaching the quality that people want. And if the goal is to continually scale higher with better quality, more resolution, and more instances happening on screen, there will be very very few studios that will be able to successfully curate their game so that all their algorithms are hitting their optimal point and no more than exactly that. We're back to the cost argument, where it's easier to rely on an accurate method of generating everything you need, and relying on the hardware to be able to produce the performance you want.

That's a demo designed to showcase raytracing. Find a best-in-class rasterised demo using a monster GPU that's happy to render 1080p30 to dedicate more time to reflections, instead of wanting as quick as possible screen-space solutions to keep the framerate up, and you'll have a comparable test. RT will be better, of course, but we don't know how close rasterisation will take us.

This is a demo, but it's arriving to BFV and we will be seeing the game on release in March or after release with all these features in place. Once again, without DirectML.

Definitely not!

Swapping sides for a moment though, I think the argument in favour of RT and ML should look at the wider picture. What if ML can be used in game for AI and things? If ML can be justified for its inclusion, then the package deal of compute+ML+memory search structures would enable denoised RT for those games that want it, but other options for other games. Is the full car-driving ML really needed for that, or could we get away with the equivalent of a few tensor cores for denoising? If ray acceleration is 10% of silicon and ML is 10%, they'd we very worth including to give devs options. Although there's the old economic issue of games not targeting this hardware if it's not everywhere, and better AI is a tough sell. Better physics is an easy sell.

ML can be used for a lot of things, I purposely stayed away from this area of discussion because I felt that it would just derail the graphics piece of it. And then we'd open up a blue sky of ideas which is more the realm of misterxmedia. When game companies start talking about using ML for those items, I'm on board to discuss it though.

 

[iroboto]

There are RTRT games for RTX cards out there, as of 22/10/2018?
Where??

Windows just released again recently, so it should be coming soon. DXR was just enabled on windows.
Metro Exodus and BFV will ship with RT.
Tomb Raider will have its patch out eventually

 

[DavidGraham]

Windows just released again recently, so it should be coming soon. DXR was just enabled on windows.
Metro Exodus and BFV will ship with RT.
Tomb Raider will have its patch out eventually

And all of these games were playable with RTX at the NVIDIA presentaion, the press tried all of them. BFV had an Alpha with RTX as well.

 

[iroboto]

And all of these games were playable with RTX at the NVIDIA presentaion, the press tried all of them. BFV had an Alpha with RTX as well.

Yea, but that's so alpha like
I'm expecting so much more performance in this area once we get into late 2019/2020.
But I agree it works. Now we need to wait for that AI Denoising to happen.

 

[anexanhume]

Sounds like Navi may be up and running in the lab.

https://www.fudzilla.com/news/graphics/47435-amd-has-7nm-navi-gpu-up-and-running-in-its-lab

 

[Shortbread]

Next generation Playstation game controller with handheld/streaming device capabilities?

Images

A controller for interfacing wirelessly with a computing device is provided, including the following: a housing defined by a main body, a first extension extending from a first end of the main body, and a second extension extending from a second end of the main body, the first extension and the second extension for holding by a first hand and a second hand of a user, respectively; a touchscreen defined along the top surface of the main body between the first extension and the second extension; a first set of buttons disposed on the top surface of the main body proximate to the first extension and on a first side of the touchscreen; and a second set of buttons disposed on the top surface of the main body proximate to the second extension and on a second side of the touchscreen.

2. The controller of claim 1, wherein the touchscreen is configured to display images controlled by an interactive application that is executed by the computing device.

18. The controller of claim 12, further comprising a speaker positioned along the top surface of the main body, wherein the speaker is configured to present a first audio portion from the game console, the first audio portion being separate from a second audio portion that is transmitted by the game console to a display device.

In one embodiment, the touch panel 26 can be coupled with a display screen to provide a touchscreen interface as part of the controller. The touchscreen interface can be controlled by an interactive application to display various images in accordance with the interactive application. For example, the touchscreen might display an image which delineates regions of the touchscreen that correspond to various functionalities for the interactive application. As another example, the touchscreen might display a line which indicates a gesture that the user may perform. The user might trace the line shown on the touchscreen so as to perform the indicated gesture. As yet another example, the touchscreen might be configured to display a gesture or gestures of the user, by providing a visual trace line of where the user touched and swiped/moved their finger(s) along the touchscreen. In one embodiment, the touchscreen is configured to display the most recent gesture or gestures which have been performed on the touchscreen. By way of example, older gestures may be eliminated from display when a new gesture is detected, or trace lines may be configured to fade from display over time or be eliminated from display after a preset time has elapsed.

Sounds like the DualShock 5, with touchscreen capabilities. See images.

Has my dream come true? Let it be so...

 

[DieH@rd]

I still think that only change on DS5 will be inclusion of small eInk display on the touchpad surface. Nice new functionality without large battery drain.

As for full screen on gamepad, I would rather they release a standalone tablet with full DS5 button layout on its sides, and then make sure their Remote Play feature is rock-solid [with a dedicated WiFi RP modules on both console and tablet].