Predict: Next gen console tech (9th iteration and 10th iteration edition) [2014 - 2017]

Status
Not open for further replies.
I wonder how much die area is required for each HBM memory controllers on the SoC itself. The first HBM products coming out right now have a large die, very high end stuff, and they have only 4 stacks.

The Fiji die shot indicated about 11mm^2 for each 1024-bit connection?
 
It looks like Zen will support a superset of the instructions that Jaguar supports. It supports some things like FMA that Jaguar does not and dumps a number of extensions that were introduced for Bulldozer that Jaguar did not get.
It seems like it should work terms of instructions, even if it turns out to be less optimal for the changed CPU organization.


Hmm then I would expect something like a quad core Zen with a Fury-like GPU at 14nm and 16GB of ram by the end of 2018, early 2019.
I really don't see how they will achieve the usual generational upgrade (x8-x10) unless the 10nm node rolls quickly out, which seems unlikely.
That particular machine would be sold as a PS4+ (on steroids) and everything backward compatible with the current PS4 vanilla.
 
The jury is out on whether getting more than 4x-6x is readily available before 7nm. Scaling problems see to be making 10nm more like 20nm was to 28nm, where getting a combined improvement in density (interconnect, transistor size) and structural improvement (all-around gates, new materials, etc.) will not happen for two nodes.
HBM or the equivalent energy-efficient tech might free up a chunk of power budget that could go towards clawing back some of the performance improvement the process transitions will probably not deliver.

Whether architectural or implementation changes can make up for the rest when they've not contributed that much before is uncertain.
 
So are we saying 10nm FinFET is really equivalent to 20nm? I'll admit I understand very little about process technology.

I'd agree we're not likely to see a healthy ~10x increase in performance until we have 7nm which I don't believe could be ready for console chips until late 2020 at the earliest.

I saw this article on EEtimes a few days ago:


EUV Gets $500M Center

Globalfoundries, New York partner on center

Globalfoundries and SUNY Polytechnic Institute will spend a total of $500 million over five years to create a new R&D center to accelerate the introduction of Extreme Ultraviolet (EUV) lithography into the 7nm process node and beyond. The move is the latest sign EUV will finally make its way into production fabs, albeit probably not until 2018 or later.

The Advanced Patterning and Productivity Center (APPC) will be located at the Colleges of Nanoscale Science and Engineering (CNSE) in Albany, N.Y. It will have a ASML NXE:3300 EUV scanner and a staff of about 100 researchers.

“I consider this a very positive sign,” said Gary Patton, chief technology officer and senior vice president of worldwide R&D at Globalfoundries. EUV “has gone through a lot of ups and down from unrealistic exuberance four or five years ago to pessimism two or three years ago to our current thinking it’s going to be real pretty soon,” Patton said.

EUV could be ready for use in production fabs "as early as 2018-2019," said Patton.

“EUV technology has emerged from R&D and the new center will meet the rising demand to commercialize this technology and put it in the hands of end users,” said Gishi Chung, a senior vice president at Tokyo Electron, speaking in a press statement.

The much delayed EUV effort is one of the most technically ambitious projects geared at driving progress toward making smaller chips. The work has required billions of investments from big chip makers including Intel and TSMC.
 
So are we saying 10nm FinFET is really equivalent to 20nm? I'll admit I understand very little about process technology.
No, but that its positioning relative to its predecessor in terms of improvement and desirability may be comparable to what happened with 20nm, which was for various reasons skipped for various processor and GPU lines.
 
Since Phil Spencer suggested that instead of a whole new console we might see "upgrades" to Xbox One. I was thinking what are the chances when Microsoft releases a slim model Xbox One they start including a thunderbolt 3 port on it making it able to work with an some sort of eGPU down the road.
 
He clarified that consoles need to remain "like appliances". They need to get bought and then just to work as intended without any complications. That is the right approach.
 
I'm ready to revise there is no new console generation. Consoles from sony and microsoft will be following similar pattern to mobile phones albeit in 3-4 year cycles or whatever timing provides sufficient jump in computing technology without raising the cost too much.

There will be 2-3 consoles from single manufacturer at the market at any given time. One higher end 399$ model and cheaper 199$ model. And possibly something cheaper or more expensive. Consoles from same manufacturer will be ~compatible with each other and most games will keep running on the latest 2-3 iterations of hardware. If economy gets much better it might be 499, 299 and 149 price points.

Both sony and microsoft will do mid generation refresh to hardware and that will kick start the "new" console cycle. For sony I would guess 12-16GB HBM2 memory, better amd gpu and cpu and cost is around 399$. I would hope for some ssd like storage but that might be optimistic. I wouldn't be surprised if the new architecture would be used for cost reduced "ps4" that is inside even smaller casing.

Microsoft might go full pc with xbox and the hw cycle is completely broken.
 
Last edited:
He clarified that consoles need to remain "like appliances". They need to get bought and then just to work as intended without any complications. That is the right approach.

you can still do that and have an upgradable console.

I can buy a tv and then add any number of components to it and the tv will continue to work. I can buy a blender and add any number of attachments to it.

I bought a mixer for baking and then bought an attachment for ice cream...

So yea. I wouldn't rule out an upgradable console. I just think it will be the next one and not the xbox one
 
you can still do that and have an upgradable console.

I can buy a tv and then add any number of components to it and the tv will continue to work. I can buy a blender and add any number of attachments to it.

I bought a mixer for baking and then bought an attachment for ice cream...


All you are talking is about accessories. You cannot upgrade your TV with better screen, put stronger motor in blender and mixer.

Console do have support for accessories. They are custom controllers, arcade sticks, wheels, cameras, motion controllers and VR headsets...
 
All you are talking is about accessories. You cannot upgrade your TV with better screen, put stronger motor in blender and mixer.

Console do have support for accessories. They are custom controllers, arcade sticks, wheels, cameras, motion controllers and VR headsets...
The TV analogy is a curious one though. It's backwards and forward compatible with the content you view on it.
 
Expect an incremental gain to 7-8 TFlops in 2020.

Why? That seems like an overly pessimistic outlook...

If Sony is intent on shortening the duration of their console generations in order to take advantage of maturing production technologies then the gains over a traditional 6-8 years generational period (i.e. across a couple of iterations) should still be much greater than a measly 4 x TFLOP GPU performance.

7-8 TFLOPs for a console GPU in 2020 (four years from now) will be embarrassingly bad. Like Nintendo levels of performance gain.

If Sony/MS want people to upgrade their box more frequently then they'll need to offer a damn sight more performance than a mere 2 x jump per hardware update.

If PS5 in 2020 has a GPU with lower than 15-16 TFLOPs then I simply wouldn't be interested in upgrading, when the perceivable performance gains aren't visible in the games. PS5 needs to offer much better than simply Uncharted 4 graphics at 4K res...
 
Why? That seems like an overly pessimistic outlook...

If Sony is intent on shortening the duration of their console generations in order to take advantage of maturing production technologies then the gains over a traditional 6-8 years generational period (i.e. across a couple of iterations) should still be much greater than a measly 4 x TFLOP GPU performance.

7-8 TFLOPs for a console GPU in 2020 (four years from now) will be embarrassingly bad. Like Nintendo levels of performance gain.

If Sony/MS want people to upgrade their box more frequently then they'll need to offer a damn sight more performance than a mere 2 x jump per hardware update.

If PS5 in 2020 has a GPU with lower than 15-16 TFLOPs then I simply wouldn't be interested in upgrading, when the perceivable performance gains aren't visible in the games. PS5 needs to offer much better than simply Uncharted 4 graphics at 4K res...

You are not going see a 15-16 Tflops console in 2020.
 
Status
Not open for further replies.
Back
Top