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

[Nesh]

I don't even think a 20-40% TF advantage would make much difference in the grand scheme of things personally. Not for Sony anyway.

Raw TF, very simular arch, wont be differentiaters.
Maybe hybrid streaming may make the cost to performance big enough for low end.

But it would be nice to see MS try to take a risk, AI accelerater, Denoiser, something I've been mulling over to raise in the RT thread is, general PT/RT that would work for things like RT, Object based audio, AI, or if the CU's could be customised enough to allow for it etc.

The chips are going to be used in their Racks also for both game and non gaming workloads so those types of customisations could make sense.

MS needs to produce some jaw dropping games even if they are shallow, 3P won't show enough of a difference.
At the moment the expectation is that all the consoles are going to be pretty simular though. MS will be ok, just competing on breadth of market more so than depth of console market.
As a tech junkie I would just like to see different things from both Sony and MS.

The way I see it is that MS wants to cover all customer profiles. Those willing to pay less and those willing to pay more.
The hyprid cheaper console might be able to provide the same experience as the expensive one for those that have the internet that can support it.
On the other hand I wonder if the hyprid will provide an experience equal to PS5 while the more expensive will be assisted also by hyprid streaming immediately or later to outperform the PS5.
The first sounds like a more achievable and less complicated solution.

edit: I dont know how Sony will provide a similar offer I dont see them having the infrastructure (and if they rent it will be more expensive for them). MS is literally thinking out of the box or at least they have the means to expand faster than Sony does. Ken Kutaragi once envisioned the Playstation becoming a service that can be provided by other media devices. But MS seems more capable and managed to even expand on competitive platforms which is a very powerful strategy. I dont know how can Sony compete the PC space and the console space simultaneously. Game pass and streaming won't just be on consoles. It will become a PC, TV and mobile service.
PS Now has been a mediocre experience and market performer. Maybe they will revamp it during the PS5 announcement. But if they continue like this it will be evaporated l

 

[Shifty Geezer]

The hybrid solution would be one heck of a gamble at this point though without the market being tested first. The tech hasn't even been rolled out for field evaluation! Worst case, there's an expensive console very few buy because it costs so much, and a cheap console no-one buys because the hybrid streaming doesn't work as well as hoped and they want the responsive experience of a proper console. If I were in charge, I'd want a 'normal' option, a normally priced console to serve the normal market. It wouldn't have to win as that's a tall ask, but it'd be wanted to shore up XB fans for the next five years of transition to Live as a gaming platform on any device, by which point MS would want tens of millions of PC owners locked in. That's worth way more for them and where I expect their console hardware choices to be determined, in serving that goal.

 

[seahorsesaw]

Both will be more powerful than each other at the same time

Yes, it's called the Schrodinger (X)Box. It's both a Playstation and Xbox depending on when and how it is observed.

 

[iroboto]

It would appear as though Thurrott and myself are on similar wavelengths, but coming at the same solution from different angles. He has more connections than I do here, so his take is likely the correct one here.
https://www.thurrott.com/xbox/200140/microsofts-evolving-gaming-strategy-takes-a-giant-step-forward#

What it looks like Microsoft is doing, instead of porting each Xbox feature back to the PC one by one, they are simply dumping the entire Xbox one installation/servicing plumbing and making it the primary installation for Windows.

When you extract the installer from the Store after downloading, you will discover that it is using the .xvc file format. While that may not sound familiar, Microsoft introduced this format around 2013 specifically for Xbox One games. Further, you can now install this file format using PowerShell in 19H1; I never thought I would write the words “PowerShell can now install Xbox One games” but here we are.

On the surface, what it looks like Microsoft is doing is collapsing any differences between PC and Xbox gaming to make it the same experience on both devices. The company is in the process of making the delivery mechanisms based on the Xbox infrastructure, appears to be making it possible to run Xbox games on the PC, and replacing the existing Store PC games infrastructure on that of what Xbox has built.

An interesting thing happening here because, Xbox binaries come with compressed audio and PC binaries usually come with uncompressed audio, thus the large install file sizes.

 

[Jay]

There's a few things I wouldn't mind hearing about in build 2019 regarding OS side of things :
One core
C shell
Game OS
UWP progress and road map
Intelligent delivery coming to pc
Fast start coming to pc
Win store

XO is a custom win 10 system.
Next gen and Windows will not be like that, will be modular based, and Xbox modules will be shared with win 10.

 

[Jay]

It would appear as though Thurrott and myself are on similar wavelengths, but coming at the same solution from different angles. He has more connections than I do here, so his take is likely the correct one here.
https://www.thurrott.com/xbox/200140/microsofts-evolving-gaming-strategy-takes-a-giant-step-forward#





An interesting thing happening here because, Xbox binaries come with compressed audio and PC binaries usually come with uncompressed audio, thus the large install file sizes.

Just listened to brads video.
He's really hyped about it like it's going to expand the Xbox ecosystem and make it easier for devs thereby explosion of games and market share.

The problem is that isn't the reason people didn't play games from windows store.
Build should be interesting none the less

Edit :
Xbox gamepass comes to pc, play XO versions of your games on pc

 

[turkey]

.

Going with a 4k / 1080 target for both systems (or a 4x difference) you'd be looking at a 12 TF and a 3TF console if everything scaled perfectly with resolution. The rumours state a 4TF low end system. That's actually quite a lot to cover any overhead from none-resolution-scaling elements of the game.

A friend of a friend's dog told me that was because they are time slicing the GPU again for Kinect 3

 

[chris1515]

https://www.resetera.com/threads/ne...ces-spicing-2019.91830/page-213#post-18292022

https://www.resetera.com/threads/ne...ces-spicing-2019.91830/page-213#post-18292757

https://www.resetera.com/threads/ne...ces-spicing-2019.91830/page-213#post-18293886

https://www.resetera.com/threads/ne...ces-spicing-2019.91830/page-213#post-18294050


On resetera, Albert Penello who retired from MS a few months ago, gives some very interesting information about console launch and he gives some hints about Sony and MS choosing AMD and the approximative size of console APU 350 mm2 and en eventual 1080p console next gen

EDIT: This is the second time he explains the problem of HDD 2,5 and BR drive with less economy of scale because consoles are nearly the only place where they are used and probably the reason of the SSD choice and a sku without BR.

 

[iroboto]

https://www.resetera.com/threads/ne...ces-spicing-2019.91830/page-213#post-18292022

https://www.resetera.com/threads/ne...ces-spicing-2019.91830/page-213#post-18292757

https://www.resetera.com/threads/ne...ces-spicing-2019.91830/page-213#post-18293886

https://www.resetera.com/threads/ne...ces-spicing-2019.91830/page-213#post-18294050


On resetera, Albert Penello who quit MS a few months ago, gives some very interesting information about console launch and he gives some hints about Sony and MS choosing AMD and the approximative size of console APU 350 mm2 and en eventual 1080p console next gen

I can’t beleive there is some guy still chasing him about Scorpio being a response to 4Pro and having 18 months lead time.

Goodness. Yea, let’s re-vector the whole thing based upon a random unverified pastebin.

Albert provides a lot of good information and should keep most people more grounded about what’s coming.

There was a question that if everything is locked up so early how do you ensure you can beat your competitor in power?

Lol. You don’t. There are no guarantees. This is business.

And at the very least, with MS specs being leaked out everywhere, how come Sony can’t respond to these “credible” leaks and build a stronger console for next gen? For some reason only MS is allowed this line of reasoning.

Anyway, I digress.

Imo With limited silicon you need hardware accelerators, not more flexible compute. Team hardware RT.

 

[Cyan]

Quite good video, needed reminder. There is even screenshot of B3D. Rich tells interentz to hold the horses, there will be fakes, lekas and dissapointmens, there are less apples to apples HW comparisons and in the end we should be pleased.

JeuxVideo details the possible specs of the next generation of Xbox consoles.

They would be shown at E3 2019.

 

[dobwal]

Personally I don’t see MS investing in 2 different Tflop level skus.

One performant sku. One streaming sku.

A high priced performant sku that will serve the gaming market and anyone in the need of Azure cloud GPU features. Since the hardware will be shared across businesses and be part of multiple revenue streams outside of gaming, it makes it easier to drive up unit volumes. Thus, allowing way more users (Anaconda users, Lockhart users and Azure users) to help drive down BOM.

One streaming sku that’s probably ARM based with a tad bit of memory to handle mobile type media apps like Netflix or Hulu. I see MS selling it for $150-$180 dollars with a year’s worth of Live and Game Pass included.

Streaming can act as an extender of performant sku with the ability streaming locally or stream from the cloud if main console is in use.

It’s ultimately one device powering the graphics of the next gen Xbox market. It’s more about segmenting just the userbase versus segmenting the userbase as well as the hardware base.

 

[ultragpu]

I can’t beleive there is some guy still chasing him about Scorpio being a response to 4Pro and having 18 months lead time.

Goodness. Yea, let’s re-vector the whole thing based upon a random unverified pastebin.

Albert provides a lot of good information and should keep most people more grounded about what’s coming.

There was a question that if everything is locked up so early how do you ensure you can beat your competitor in power?

Lol. You don’t. There are no guarantees. This is business.

And at the very least, with MS specs being leaked out everywhere, how come Sony can’t respond to these “credible” leaks and build a stronger console for next gen? For some reason only MS is allowed this line of reasoning.

Anyway, I digress.

Imo With limited silicon you need hardware accelerators, not more flexible compute. Team hardware RT.

He did mention that supposed one year delay of PS5 launch could lead to spec changes, maybe in the realm of a slight clock bump. But the main thing is we don't know the price and performance target Sony has set out at the beginning, common logic tells me Sony wanted a $399 console and be content with 10 TF but the hopeful side tells me they wanted at least twice Scorpio tflops and be more bold with a $449-$499 launch.

 

[Metal_Spirit]

He did mention that supposed one year delay of PS5 launch could lead to spec changes, maybe in the realm of a slight clock bump. But the main thing is we don't know the price and performance target Sony has set out at the beginning, common logic tells me Sony wanted a $399 console and be content with 10 TF but the hopeful side tells me they wanted at least twice Scorpio tflops and be more bold with a $449-$499 launch.

I would also prefer the Bold option. 10 Tflops seems like a very small generational jump given the specs of the One X. In terms of Marketing, something bellow 12 Tflops can just seem lie very little. Regardless of the results produced, 10 Tflops would seem like a 66% jump from previous generation, and if Microsoft manages 100%, they would shure promote that. And I bet that would be a very efecctive marketing weapon.

 

[fehu]

What about "20TF*" ?

*half precision

 

[AlNom]

Thanks for the feedback.

I plan to update the diagram correcting it and changing with the new info.

*nudge*

It'd be nice to have a decent tech discussion instead of wishlists over and over again like riding the village flopcycle.

The patent has an embodiment where the ALU block is more like one Core ALU that can pair with a Side ALU, and another Core ALU. The patent doesn't limit itself to any one specific combination, although in that embodiment a Core ALU has a multiplier that makes it more suited for the primary set of instructions, while the Side ALU only has an adder and ancillary hardware for assisting a Core ALU on an complex op. A Core+Side is needed to get the full range of operations of a traditional GCN unit.
As far as the diagram goes, the two Core ALUs have straight links to the operand network, while the Side ALU hangs off their ports. The destination cache also feeds into the register file or operand network, rather than the scheduler. As far as an L0 cache goes, the related patent on the register file has the output flops of the register file labelled as an L0.
Part of the motivation for the destination cache is that it can feed back into the ALU operand ports, since the register file itself has lower peak bandwidth than the ALU blocks require.

So are we looking at:

Navi ALU (as in the patent) = 2 MUL + ADD (3 flops per ALU)
GCN ALU = MADD (2 flops)

I believe the patent also goes into the number of tex units per unit as well, though my brain isn't quite fully operational buttAl station this morning.


ergo, a compute unit

GCN CU = 4 x (16 x MADD)
NaviCU = 4 x 16 x (2 MUL + 1 ADD)

 

[Heinrich4]

What about "20TF*" ?

*half precision

Ahamm ( 10TFlops FP32 and 40 Tflops ML ...)

 

[DrJay24]

I would also prefer the Bold option. 10 Tflops seems like a very small generational jump given the specs of the One X. In terms of Marketing, something bellow 12 Tflops can just seem lie very little. Regardless of the results produced, 10 Tflops would seem like a 66% jump from previous generation, and if Microsoft manages 100%, they would shure promote that. And I bet that would be a very efecctive marketing weapon.

You think a TF number is a marketing weapon? I think the games will sell the system, not technical specs. Most people don't know what a TF is.

 

[fehu]

And many people don't know what ram is, or what frequency means.
A bigger number is a bigger number.

Cue moderator's hammer.

 

[milk]

And many people don't know what ram is, or what frequency means.
A bigger number is a bigger number.

Cue moderator's hammer.

And that's why next gen play station will be PS5. 5>4. That's all they need. Notice One < 360, thus the low sales. Big mistake.

 

[Kaotik]

And that's why next gen play station will be PS5. 5>4. That's all they need. Notice One < 360, thus the low sales. Bug mistake.

on the other hand if you sort something alphabetically the ones starting with number come first, and since the numbers are ordered from smallest to biggest, we can assume same goes for letters, which means One > 360