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

Status
Not open for further replies.
Don't confuse developers and publishers. Developers probably want the latest and greatest hardware at all times where as publishers want a just want a massive single spec install base that makes it economically feasible to develop for.

You must give customers what they want, this is true whether you are a dev or a publisher. (simplified version)
Cutomers want better games (moar pixels, moar frame rate, moar whatever) and it's not like publisher can say "No sorry we can't right now, please be kind and wait 5 years".

If fixed hardware every 5years is not the best way to give the customers/players the quality they deiste then it will be abandoned.


P.S.
Providing customers a sustainable "service/product" is not the goal, the goal is to never make a product/service be good for too long!
 
Last edited:
I think there's an argument against that in that if you progress too fast, you then crash - 'That which burns brightest burns shortest.' That's really a socioeconomic discussion though. Ultimately at this point in time there's a cost to creating games that increases with game quality, and more power means more expense and a harder target to make money from. I wouldn't be surprised if devs/pubs shy away from 'early' hardware. For that reason PS4N likely won't be properly utilised, nor high-end PCs, and more so a new generation of XBox requiring far more investment to really make use of it.
 
I think there's an argument against that in that if you progress too fast, you then crash - 'That which burns brightest burns shortest.' That's really a socioeconomic discussion though. Ultimately at this point in time there's a cost to creating games that increases with game quality, and more power means more expense and a harder target to make money from. I wouldn't be surprised if devs/pubs shy away from 'early' hardware. For that reason PS4N likely won't be properly utilised, nor high-end PCs, and more so a new generation of XBox requiring far more investment to really make use of it.
well, better hardware also means less optimizations required. So you can just use a little bit more of everything and still everything is fine, without any more optimization. Sure, it won't run as effective as fully optimized, but it is an easy win for developers. As they must also support PS4 the game will be optimized for that platform (well I hope that it still get's optimized for the "old" PS4) and details get increased on the new platform.
 
As much as it probably wouldn't happen I'd like to see a console released with same power of new high-end video card. Build a console around AMD Vega and release around the same time as the card is released so it is cutting edge in terms of graphics.. Price the console $50 less than the actual card ie $599.

It would pull in a lot of people I think with that value. Even get some pc gamers back into consoles.

Also to put it in perspective PS3 released in 2006 at $499/$599. With inflation $599 today would be the equivalent of $498.40 in 2006. In 2017 probably even less.

So a $599 console could be possible if it provides enough value to consumers.
 
It would pull in a lot of people I think with that value. Even get some pc gamers back into consoles.
Doubtful. Any PC gamer would buy the graphics card and play their existing library cranked up to 11 and really appreciate the improved performance. If they bought a console, they'd have a couple of games they've bought for it for notably more than the PC versions.

Such a monster console wouldn't convert PC gamers, but would stop potential console defectors envious of PC's performance who'd have the option to go monster console.
 
IMO, a 10TFlops next gen console is not unlikely, it is more than likely

The XBox one soc is 363mm², the 64 CU Fury X chip is a little less than 600mm² on 28nm process tech. Samsung and Global foundries 14nm FinFet process has twice the density and significant power consumption advantages compared to 28 nm; 40% faster clock at same power consumption or 70% less power spent at same speed.

I can imagine a soc with 64 CUs, 8 Zen cores, cache, uncore etc in less than 400mm² and with <200W power consumption. 8/16 GB HBM2 for memory.

I'm completely baffled by the PS4K rumours. What do Sony hope to achieve? They are already winning BIG this gen, all they'll achieve is to fragment their own userbase. Unless, of course, they are planning their own 10TFlops console.

Cheers
 
IMO, a 10TFlops next gen console is not unlikely, it is more than likely

The XBox one soc is 363mm², the 64 CU Fury X chip is a little less than 600mm² on 28nm process tech. Samsung and Global foundries 14nm FinFet process has twice the density and significant power consumption advantages compared to 28 nm; 40% faster clock at same power consumption or 70% less power spent at same speed.

I can imagine a soc with 64 CUs, 8 Zen cores, cache, uncore etc in less than 400mm² and with <200W power consumption. 8/16 GB HBM2 for memory.

I'm completely baffled by the PS4K rumours. What do Sony hope to achieve? They are already winning BIG this gen, all they'll achieve is to fragment their own userbase. Unless, of course, they are planning their own 10TFlops console.

Cheers

Of course a 10TFlops console is likely...I just don't know if it would be likely by fall 2017 which I suspect is the latest Microsoft would release something new. It would probably be possible to release a near 10Tflops console next year. However it would be cutting edge and wouldn't cost less than $600.
 
However it would be cutting edge and wouldn't cost less than $600

Why wouldn't it cost less than $600 ?

The main soc would be around the size of the XB1 soc. MS will pay a premium for HBM2, but won't have to bundle a Kinect with each unit sold. I'd expect the BOM to be <$500.

Cheers
 
Why wouldn't it cost less than $600 ?

The main soc would be around the size of the XB1 soc. MS will pay a premium for HBM2, but won't have to bundle a Kinect with each unit sold. I'd expect the BOM to be <$500.

Cheers
You think they could offer a console with near 10 Tflops and 16GB HBM2 for under $500 by next year?

AMD Vega GPU equivalent will probably be around $650 itself.
 
You think they could offer a console with near 10 Tflops and 16GB HBM2 for under $500 by next year?

AMD Vega GPU equivalent will probably be around $650 itself.

Don't know how likely it is, but a mixture of HBM1/2 and DDR4 is possible...
 
Don't know how likely it is, but a mixture of HBM1/2 and DDR4 is possible...
Only way I could see it is something like 4GB of DDR4 for OS and 8GB of HBM1/2 for GPU...
That would keep cost down.

8GB doesn't seem like a lot for such massive GPU but when you think about it Fury X is 8Tflops and only has 4GB of HBM1...
 
AMD Vega GPU equivalent will probably be around $650 itself.

Cost != price

AMD has a profit margin on the soc, the board vendor has a profit margin on the board, the retailer has a profit on the final sale. Being a highend product, the profit margins are comparatively fat vs a loss leading (or close to break even) console.

Cheers
 
Cost != price

AMD has a profit margin on the soc, the board vendor has a profit margin on the board, the retailer has a profit on the final sale. Being a highend product, the profit margins are comparatively fat vs a loss leading (or close to break even) console.

Cheers

Yeah but Microsoft isn't just selling a gpu...they have an entire system that has to go with it..

With all the talk of Microsoft releasing a PC like device for Xbox I just wish they would just release an additional model where the hard drive wasn't included but with an easy to access internal bay like PS4. I'd love the option to buy a system where I could choose my own type of storage without incurring the additional cost of a pre-added drive that I would probably never end up using anyway...

Also they could probably drop the disc drive from that model as well to shed more cost.
 
but a mixture of HBM1/2 and DDR4 is possible...

IMHO, unlikely. You'll need PHYS on the soc for the DDR4, you'll need traces on the PCB and you'll need a lot of devices (widest DDR4-device is x16, so you'll need 8 of them for a 128bit bus). On top of that the energy spent per bit/s transferred is significantly worse than that of HBM2, - important in what is likely to be a power constrained system.

IMO, It's more likely they'll opt for 3x4GB HBM2 than 2x4GB HBM2+8GB DDR4, - and a really fast NVME Flash storage solution for super fast asset streaming/app swapping.

Cheers
 
I'd love the option to buy a system where I could choose my own type of storage without incurring the additional cost of a pre-added drive that I would probably never end up using anyway...

Exactly, just let the base SKU be as barebones as possible (Like the 360 Arcarde). I would expect next gen to ship only with Flash storage and options for expanding with additional drives, either internal bays, or external USBc.

Cheers
 
Only way I could see it is something like 4GB of DDR4 for OS and 8GB of HBM1/2 for GPU...
That would keep cost down.
When you pay for the substrate I don't know how convenient is to add a different controller, the logic to manage the different pools, the traces on the board and something else that I don't even imagine, instead of just use another bus for hbm
 
IMHO, unlikely. You'll need PHYS on the soc for the DDR4, you'll need traces on the PCB and you'll need a lot of devices (widest DDR4-device is x16, so you'll need 8 of them for a 128bit bus). On top of that the energy spent per bit/s transferred is significantly worse than that of HBM2, - important in what is likely to be a power constrained system.

IMO, It's more likely they'll opt for 3x4GB HBM2 than 2x4GB HBM2+8GB DDR4, - and a really fast NVME Flash storage solution for super fast asset streaming/app swapping.

Cheers
If Microsoft wanted to build a really well balanced small efficient console they would include a flash storage solution right on the board. However I think it means you'd cut back on GPU and memory.

4 core zen
GPU 4-6Tflops
8GB HBM2
500GB flash storage

I'd buy that.
 
IMHO, unlikely. You'll need PHYS on the soc for the DDR4, you'll need traces on the PCB and you'll need a lot of devices (widest DDR4-device is x16, so you'll need 8 of them for a 128bit bus). On top of that the energy spent per bit/s transferred is significantly worse than that of HBM2, - important in what is likely to be a power constrained system.

I was thinking along the lines of the (rumoured) APUs that are supposed to use one or two (possibly HBM1) stacks on a socket that also supports DDR4. If AMD already had the designs in place, it might save a lot of R&D time. I'd certainly rather have an all HBM2 arrangement though, so long as the OS reserve was small.

IMO, It's more likely they'll opt for 3x4GB HBM2 than 2x4GB HBM2+8GB DDR4, - and a really fast NVME Flash storage solution for super fast asset streaming/app swapping.

A small amount of super fast flash would be my preference too, as it would set the base level for streaming. If games were required to cache to it, that would also open up the opportunity for any external device to be fast enough. If they supported network storage too, then that would be perfect ...
 
If Microsoft wanted to build a really well balanced small efficient console they would include a flash storage solution right on the board. However I think it means you'd cut back on GPU and memory.

Current gen games are supposed to be using anywhere between 1 ~ 3 GB simply for caching assets not needed immediately because of the laptop HDDs they use. When you figure that's out of a pool of only 5GB available for games, I think there's plenty of room for flash to allow for far more impressive games while still having only 8GB of system memory.
 
Status
Not open for further replies.
Back
Top