Next Generation Hardware Speculation with a Technical Spin [pre E3 2019]

[mpg1]

40% power gain is what we should expect on a node transition on the same product.
Navi will be a different product, so those gains are not the only ones to take into account.
Looking at the performance per watt charts from AMD, they claimed polaris offered a 2.5x increase in performance/watt over 28 nm products. That’s a 150% increase!
That translates into a 136 pixels height difference!
If the scale is right, Navi should offer about a 122% increase on performance/watt over Vega. So that should not be the whole picture.

I hope this is sarcasm...

 

[Xbat]

I don't understand the total disregard for the architecture in the this thread? Yes there's physical limitations with moving to a smaller node but that's not the whole story. If Nvidia stuck with Fermi do you think it would have the same performance as Turing at 12nm? Not a bloody chance.

 

[Metal_Spirit]

I hope this is sarcasm...

What do you mean?
Notes from AMD on a recent Linux driver claim that future ASICS could use a new sw SMU framework superceding the current PowerPlay driver. This might enable dynamic per-core voltage and frequency adjustment resulting in less power-hungry GPUs and more performance per Watt.
The node reduction gains are not the only thing to take into account. So with possible architectural changes those 40 or 60% are just part of the equation.
It is a fact some of the Ryzen engineers were moved to the Navi development team. If they manage to implement some form of Ryzen's Synergetic Power Rail Sharing circuitry they could offer fine-grained power and voltage regulation alongside the possibility of using low-dropout (LDO) regulators that allow per-core voltage control.
AMD claims this reduces maximum current requirements by 36% on Ryzen.

 

[anexanhume]

This is really interesting. An LED display with embedded e-ink substrate that can detect IR and laser light.

http://www.freepatentsonline.com/20190064953.pdf

 

[Shifty Geezer]

... what's the relevance to a next-gen console?

Oh, it's a Sony Interactive Entertainment patent. That was probably worth you mentioning. So potential for a display in the controller with an e-Ink background to show buttons.

Although it really reads like a complete muddle of ideas, and better suited to mobile devices than a controller. Document scanning and TOF distance measuring in a display?

 

[anexanhume]

... what's the relevance to a next-gen console?

Oh, it's a Sony Interactive Entertainment patent. That was probably worth you mentioning. So potential for a display in the controller with an e-Ink background to show buttons.

Although it really reads like a complete muddle of ideas, and better suited to mobile devices than a controller. Document scanning and TOF distance measuring in a display?

TOF is all over their HMD patents, so it makes sense. They have another holographic display patent which re-uses a diagram from this one.

edit: here's the holo display patent: http://www.freepatentsonline.com/20190064734.pdf

 

[Shortbread]

This is really interesting. An LED display with embedded e-ink substrate that can detect IR and laser light.

http://www.freepatentsonline.com/20190064953.pdf

I don't know about that particular patent (for some odd reason your link isn't working on my tablet), however, there is a possibility that DS5 may have a display screen.

 

[anexanhume]

I don't know about that particular patent (for some odd reason your link isn't working on my tablet), however, there is a possibility that DS5 may have a display screen.

Yes, and this could be related to that, and is incredibly generic. Frankly, the amount of tech they talk about putting in it makes it sound cost impractical for a controller.

They even mention powerpoint presentations with a laser pointer as an example.

 

[Shifty Geezer]

That's just to cover other fields they aren't interested in but someone else might, enabling a license fee. So, the engineers at SIE were toying with controllers with screens in, and someone had an idea for a sensor or something, and they started talking about that idea and where it'd be useful. "Imagine a tablet you could scan documents with. Or point a laser at to control from a distance." And then that all goes into the patent so if someone somewhere decided to make a tablet with a light sensor sat behind the screen for laser-pointer control, SIE step up and ask for money or block them.

To understand the relevance to PS5, we need to pare back the patent to the points that'd actually be relevant to a controller, or something else relevant. What good is a display with light sensors built in? With no focussing elements, it's just measuring ambient light the moment you get any distance from the sensor so I've no idea what it could bring to a controller. For touch, existing tech is functional and cheap. In a headset, could it be used for pupil tracking thanks to the lens working both ways?

 

[JoeJ]

With no focussing elements, it's just measuring ambient light the moment you get any distance from the sensor so I've no idea what it could bring to a controller. For touch, existing tech is functional and cheap.

The friction on phone touch screens is very unpleasant. Could it allow precise but relaxed hovering control? (Edit: even 3D?) Could it bring FPS genre to consoles? Hell, even i could become console player this way.

 

[mpg1]

What do you mean?
Notes from AMD on a recent Linux driver claim that future ASICS could use a new sw SMU framework superceding the current PowerPlay driver. This might enable dynamic per-core voltage and frequency adjustment resulting in less power-hungry GPUs and more performance per Watt.
The node reduction gains are not the only thing to take into account. So with possible architectural changes those 40 or 60% are just part of the equation.
It is a fact some of the Ryzen engineers were moved to the Navi development team. If they manage to implement some form of Ryzen's Synergetic Power Rail Sharing circuitry they could offer fine-grained power and voltage regulation alongside the possibility of using low-dropout (LDO) regulators that allow per-core voltage control.
AMD claims this reduces maximum current requirements by 36% on Ryzen.

I'm talking about analysing the pixels of a pretty basic graph on a presentation slide to determine perf/watt

It's clearly for illustration purposes. Not to be taken exact.

Not to mention that chart is several years old. So it's stupid on multiple levels..

 

[temesgen]

I really hope Sony don't commit a disproportionate amount of the PS5 expense into a controller. The current DS is very good as is.

 

[MrFox]

I really hope Sony don't commit a disproportionate amount of the PS5 expense into a controller. The current DS is very good as is.

Considering ps1, ps2, ps3, ps4, and ps4 pro, were bundled with reasonably inexpensive controllers, the chances of sony doing a 180 and spending a lot on the controller is practically zero.

There's a questionable rumor of evolving the ds4 with a small display, and/or a small camera, but neither are expensive additions nowadays. It won't magically rise to a $100 BOM as happened with kinect and wiipad.

 

[Metal_Spirit]

I'm talking about analysing the pixels of a pretty basic graph on a presentation slide to determine perf/watt

It's clearly for illustration purposes. Not to be taken exact.

Not to mention that chart is several years old. So it's stupid on multiple levels..

Quoting from the message:

“IF the scale is right...”

When there is an IF there is a doubt, so, no one is claiming that is correct. The point was just to show that AMD expects big gains, and an illustration picture, if that is trully what she is (and we only have your oppinion on that, although I could very well agree) could very well point that, even if not accurate.

But personally, I would not state that the image is a mere illustration without any intent of precision (even if only a prediction at the time). Because I do know that for a fact.

I would also not say that the chart beeing old makes it stupid. If she is misleading in what AMD expects from Navi (or at least expected), then she would be stupid. But the mere fact of beeing old... just makes it old.

But the point of the message was another one. That power gains on a new architecture cannot be measured by lithography gains only, and apparently, even if on a mere old illustration, AMD passed (on a official presentation) an idea that it expects a gain with about the same proportion as the one she got from the polaris to Vega transition.

The pixel count... well, that was the version of the image I happened to have at hand.

 

[lefantome]

Reading the leaks about Ryzen 3000 series, they look very good.

Will next gen use Ryzen 2 or the following architecture?

 

[vjPiedPiper]

Reading the leaks about Ryzen 3000 series, they look very good.

Will next gen use Ryzen 2 or the following architecture?

Probably depends a lot on if the Next-Gen consoles are full APU's like we have now, or something closer to a chiplet configuration.
If it's GPU + mem controller in 1 die, and a chiplet CPU, I would think it's highly likely that the console has the latest possible version of the Zen Arch.
But if it's a single die for GPU+ CPU, then i rekcon we will be stuck with a slightly older CPU design eg. zen or zen+.

Personally I think it's going to be a Custom GPU/memcontroller/IO Die, with a CPU chiplet on a second die.
With the chiplet being exactly the same chiplet that is used for commercial CPUs.

 

[Deleted member 13524]

Why would it be stuck to Zen 1/1.5 if it's a monolithic chip?
The PC Zen 2 APUs are predicted to come in Q4 2019 and those should bring Zen 2 and Navi, just like next-gen consoles.

 

[Urian]

Is this possible for the Next Gen?

Thanks for your time.

 

[Deleted member 13524]

That just seems to be Vega GFX9, with the ROPs connecting back to the CUs through the L1, for the primitive shaders to work.

 

[vjPiedPiper]

Why would it be stuck to Zen 1/1.5 if it's a monolithic chip?
The PC Zen 2 APUs are predicted to come in Q4 2019 and those should bring Zen 2 and Navi, just like next-gen consoles.

Because of the time involved in creating and verifying a custom part.
If they use an existing Zen2 chiplet, they know the chiplet is good because it has already been verified for use in Ryzen and Epyc systems.
If they do a single APU consisting of CPU + GPU, there is a considerable cost in integrating the CPU parts into the rest of the system.

If it's an APU design they probably needed to start the design, what 12-24 months ago?
How far along was the Zen2 arch at that time? were they confident enough in the core arch to start integrating it into a monolothic APU?
maybe?

Perhaps i should have said i think it's a higher possibility of proper Zen2 if the console uses a chiplet design.
I didn't mean a 1:1, chiplet design = Zen2, pure APU = Zen1/1.5,
just i think, and imho only, a chiplet design is more likely to have the most recent CPU cores as part of the system.

Also there are the fab benefits to a chiplet design using 2 smaller dies, and even the ability to shrink 1 part independently of the other. 7nm+? anyone.
At the cost of overall system complexity of course...

Hell, we might end up with both options?
2 different consoles with very close performance, but quite different internal structure...
Sony with chiplet APU design, 8c/16th Zen2 + 10TF GPU
MS with a slightly "custom Zen2 like CPU", in an APU design but still 8c/16th + 10TF GPU.
(if your gonna go for a custom APU, then it's possible to do CPU customizations in the process, hence the "custom CPU")
** TF numbers pulled from thin air!

as always - speculation