Predict: The Next Generation Console Tech

Status
Not open for further replies.
I think the next xbox should launch in either the Fall of 2011 or 2012 to be honest with you.Have the hardware be very powerful but also affordable with 3 different SKU's.Each SKU should come with a hdd as standard but of course, they differ in hdd space.Also, do you any of you guys believe the next xbox would use a manycore processor?
I hope that this time around Ms will go with only one SKU. I mean that I would want the lowest sku to be fully functional, to me the arcade one is not mostly because it could not support NXE and caching.
I'm all with granmaster here, I would prefer that all system include 16/32GB (enough for to help streaming from ODD and some demo/DLC /patchs) of flash memory. Ms could sell HDD separetely or do some bundles.

Many core processor ala Larrabee? While I find the idea seducing for a while, early leak about larrabee size, perfs and power consumption made the all thing unlikely. But if you think of a HD4850 as ten arrays/cores multiprocessor then most likely yes as future part could include twice as much cores/arrays. It could be even more if ATI decides to have less wide array. I will be interesting to this how GPU past RV8xx will shape up :) But I think that CPU will still be there. The balance in computational resources between CPU and GPU will be decided with efficience in mind in regard to various workloads.
I'm willing to see how ATI and Nvidia will evolve to match Larrabee flexibility and to support software based rendering. Intel solution is not the only one solution and it looks like one that have been designed with quiet some constrains.
 
Last edited by a moderator:
My 80 GB Intel X-25 was $325 from Amazon. But it was well worth it considering the reduction in build times and solution-load times. I believe it's down to $314 now -- a great deal for software developers, but of course completely unsuitable for consoles.

Yeah, the Intel and some of the Samsung and new drives from OCZ in the Vertex series are decent. Pretty much everything else uses broken hardware. But all the good ones have a price premium vs the cheap broken stuff.
 
How can Xenon & Xenos evolve into a better architecture without just adding new cores or shaders, more embedded memory, increasing clock or combining into one silicon, all while still maintaining 100% backward compatibility? What are the possibilities? I seriously think Microsoft will go this route, but something tells me they will do something much more than what I came up with above.

Tommy McClain
 
Well, the Xcpu needs to be "fixed" first (i.e. the LHS issue), as doing so would probably result in a nice performance increase on its own. Beyond that, I think (hope) MS will go for flexibility and programability for both the next CPU and gpu. Judging from devs' excitement over the Larabee design, I would think flexibility and programability would be more desirable than really fast hardware.
 
Rethinking my earlier post

I am reconsidering my thinking that Sony would launch in 2011 if Microsoft launched in 2010 (which I believe they are likely to do). Additionally, it might also be possible that the PS4's architecture would be UMA if XDR2 can provide adequate bandwidth for the GPU. With that in mind:

2010 Playstation 4
UMA 2 GB XDR2 8x2Gb devices @ 38.4 GB/s ea. (total ~300GB/s bandwidth)
40nm 4.5Ghz Cell (4 PPEs & 28 SPEs)
40nm 1.2 Ghz Nvidia GPU w/64 MB EDRAM

Does this seem more in accordance with reality? I think it makes a lot more sense than a NUMA design with 3 GB.
 
The reality is that both Microsoft and Sony have lost so much money this generation that the chances of a 2010 launch of new hardware from either vendor are basically impossible. We have reached a point in the life-cycle where - maybe - they can start to claw back losses. Bringing another loss-leading console into the market next year is crazy talk.

Additionally, if we were indeed 15 months from "launch" there would be far more awareness of any new hardware on the horizon. Microsoft's third party team is touring developers now, but they're talking about Natal, not Xbox Next.
 
Don't mention the time, money and effort the developers must spend to utilize all of that processing power either:)
 
my take on the sony console

2012 Playstation 4
UMA 4GB GDDR5 on a 256bit bus (180GB/s or whatever)
32nm Cell ( 2 or 3 improved PPE, 11 or 14 SPE)
28nm GT300-based GPU (1.7 teraflops)

Backwards compatible, PS3 games run at 1080p/AA 4x if applicable.
120Hz output (2D or 3D)
500GB and 1.5TB SKU
 
really? that explains the lack of news beside powerXCell8i (or whatever the quirky name is).

so Cell was an one-off shot? if so there wasn't much communication on that *cough*.
let's take the Cell off.
Who provides the CPU? Is a big-ass nvidia Tegra an option :p with 1.7 teraflops of GT300 units and four custom high speed ARM cores. Can VIA license the Nano to be part of a nVidia Tegra chip? Does IBM provide something new?

The only high performance powerPC currently are Xenon and the Cell's PPE. how funny would it be to see Xbox 720 and PS4 sharing the same CPU?
 
EDRAM is unneeded and wasteful if you get such enormous bandwith from your main ram. (plus clockspeed is not how we measure GPUs)

Do you think that EDRAM would have a place for either latency reasons, or if memory bandwidth was actually only 150 GB/s? Say 4x4Gb devices? Cell2 would be very bandwidth hungry, easily using over 100 GB/s at peak.
 
IBM may have ended their program.

But have they ended their program? What is this rumour based on?
The next gen Cell wasn't due until next year and they have been talking about using it as an add-on to their bigger systems as an XML accelerator.
 
I am reconsidering my thinking that Sony would launch in 2011 if Microsoft launched in 2010 (which I believe they are likely to do). Additionally, it might also be possible that the PS4's architecture would be UMA if XDR2 can provide adequate bandwidth for the GPU. With that in mind:

2010 Playstation 4
UMA 2 GB XDR2 8x2Gb devices @ 38.4 GB/s ea. (total ~300GB/s bandwidth)
40nm 4.5Ghz Cell (4 PPEs & 28 SPEs)
40nm 1.2 Ghz Nvidia GPU w/64 MB EDRAM

Does this seem more in accordance with reality? I think it makes a lot more sense than a NUMA design with 3 GB.

PS3 is already UMA-ish and was to be fully until the final cost of XDR hit them, I haven't seen anything from XDR2 that would suggest it'll be any cheaper. My guess is they'll pursue an option that lets the Tiawanese gov subsidize they're memory needs like everyone else.

Why would they go with a 40nm Cell when they're already at 45nm? That makes no sense at all. 4 PPE's are unneccessary, 2 PPE's would be more likely with possibly a 3rd if they went with a unified architecure approach. 28 SPE's is again unneccessary unless they went with a unified architecure, 12 is more likely for a stand alone Cell CPU.

At this point, I doubt we'll ever see another Nividia gpu in a Playstation. It's not that everything is Nividia's fault, Sony made it's fair share of poor decisions too, but in the end (and that's what counts) they charged Sony a huge premium to repackage a 7800 with UMA which wasn't even used (fully) all the while using that money to finance their next gen chip which Sony couldn't even use (original G80 was as big as Cell and RSX combined with a 512 bit bus). Add to it, they launch the G80 3 days before the RSX launch which made it effectively obsolete on launch and you have Sony going wtf. To put it another way, Sony paid NV big bucks to create magic and instead got card tricks in return.
As to edram, it's basically there to stabilize your bit rate to the gpu in a uma design. Yes you can do some tricks to take advantage of the larger bandwith but ultimately you can't pass thru more than your fed. It may have made sense in 2005 for Xenos but I don't believe it does today or in the future.

Looking at whatever future Playstation Sony puts out, you should first look at the failings of the last one. In this regards, the biggest failing with the PS3 was Sony putting themselves in a position where they had to make major compromises in the grand design scheme just to get the product out the door. That was so un-Sony which I'm sure it won't happen again and will probably lead to a much more conservative approach to their next one which would also be in many respects un-Sony.
 
But have they ended their program? What is this rumour based on?
The next gen Cell wasn't due until next year and they have been talking about using it as an add-on to their bigger systems as an XML accelerator.

I certainly do not have any privileged information on this. I seem to recall that earlier in this thread someone had heard word from someone associated with Gedae. That's all I know.
 
But have they ended their program? What is this rumour based on?
The next gen Cell wasn't due until next year and they have been talking about using it as an add-on to their bigger systems as an XML accelerator.
In the linked thread, Enzyme reports he spoke with Peter Hofstee who said they currently aren't working on upgraded Cell hardware, instead focussing on software development. Now as Cell already works very well in networked systems, it wouldn't be hard (theoretically) to combine 2 or 3 on a chip as discrete dies in a network, creating perhaps a 2 PPE, 14 SPE Cell system. True Cell development is likely on hold until the software side makes it appealing and competition can be gauged. No point building a new chip if the world doesn't want to know. Intel are going to push Larabee hard, and IBM need a software platform that makes Cell easier to work with if Cell's to compete.
 
Status
Not open for further replies.
Back
Top