Predict: The Next Generation Console Tech

[MrFox]

Yield due to defects is based on area; processes trend towards comparable yields as they mature.

Okay, so considering the A2 has most of it's area covered by identical small processors and their own cache (redundancy, any one or two disabled to improve yield), does the yield becomes practically only dependent of the critical non-redundant area? If we compare the cell die versus the A2 die, there's a gigantic reduction of the percentage of non-redundant area. The A2 could have an amazing yield even shrunk to a state-of-the-art process very early, so less waiting for the production to be mature.

Current chip is 360mm2 at 45nm, 200gflops DP (do we double that for SP?) at 1.6ghz, 55 watts. It could be 180mm2 at 28nm already, and clocked at 4ghz it could do 1 Tflops SP. It's almost an off-the-shelf part by IBM, so it doesn't cost much RnD from sony.

 

[bgassassin]

Scuttlebutt has it that he may now be an _ex_ Crytek dev.

For that? I hope that's just a rumor.

 

[Rangers]

Charlie is going to have some console nonsense up in the morning. Giggity.

Two stories for tomorrow morning, servers related and consoles.

Charlie Demerjian ‏ @CDemerjian

The other is outing an upcoming console…… No hints though.
1h Charlie Demerjian Charlie Demerjian ‏ @CDemerjian

 

[3dilettante]

Okay, so considering the A2 has most of it's area covered by identical small processors and their own cache (redundancy, any one or two disabled to improve yield), does the yield becomes practically only dependent of the critical non-redundant area?

It becomes dependent on the number of defects per mm2 and the area that is composed of critical structures and the area of redundant silicon that must be without flaw.

Current chip is 360mm2 at 45nm, 200gflops DP (do we double that for SP?) at 1.6ghz, 55 watts. It could be 180mm2 at 28nm already, and clocked at 4ghz it could do 1 Tflops SP. It's almost an off-the-shelf part by IBM, so it doesn't cost much RnD from sony.

How close is the current chip to being pad limited?

The clock scaling from a 45nm part to 32nm with the same design has turned out to be in the tens of percent, not more than doubling.
If a chip is going to go for 4GHz, it is going to be designed for it and would probably get close to that speed at 45nm.

 

[steampoweredgod]

High speed wifi, high speed disc drive, and high speed solid storage would be ideal.

Those are realistic short term technologies that would DO THE MOST difference IMHO.

The memories can serve as caches for advanced streaming.(and there are pretty advanced algorithms that would benefit from this)

AND IT WOULD BE PLAYING IT SAFE BY KEEPING ALL THE OPTIONS OPEN

 

[Rangers]

Charlie's article is up and it's stupid

http://semiaccurate.com/2012/03/02/sony-playstation-4-will-be-an-x86-cpu-with-an-amd-gpu/

He says Sony is pursuing 3D stacking, in order to make even more massive chips.

Never mind that:

-Is there any chance 3D stacking is commercially ready in 2013-14?
-Why bother? The top normal GPU's are likely outside the thermal limits of a console anyway. Therefore if it's raw power you're pursuing, 3D stacking is irrelevant.
-He also says its an SOC or could be an SOC at points, again if you're pursuing max power that in itself is contradictory
-Sony's financial status probably precludes them losing a ton of money on PS4, and the only other option is price it really high and that's no option after PS3 debacle.

My point is if you're looking to outmuscle the other guy you almost surely dont need exotic 3D stacking to do it. Just get a bigger GPU than him.

I think Charlie just read this article and made up the rest with his imagination. Since he seemed really a bit crazy with some of his Kepler reporting too I'm beginning to seriously doubt his credibility.

 

[Gubbi]

Stacking allows for lots of inter-die bandwidth, I suspect this is what Sony is pursuing, not larger total die area. Stack the CPU, GPU and some DRAM for super high bandwidth, - and have some more DRAM on a "slow" XDR or GDDR5 bus.

Cheers

 

[steampoweredgod]

Honestly it's sad the guy lose his job for this he didn't revealed anything unknown.

Depending on the way you state obvious, it may offend or go beyond what is considered as appropriate protocol or etiquette. It's called tact, and even a single word in a single contex heck a letter or no letter at all even silence.

Because under some abstract interpretations in the subconscious( it may be considered an implicitly monstrous act). Such is defined as trolling, anonymous, emanon, using anonymity to abuse your position and exploit it to your advantage maybe to boost your ego or something.

Of course epic trolling in the right context looks ridiculous, case in point the NOOOOOOOOOOOOOOOOOOOO of ever increasing length that seems to be appearing in both present and new revisions of past Star wars.

It look ridiculous, but there's some very very deep meaning behind it. expect it to keep increasing

btw, I do so like the anonymous state, emanon, tuxedo mask. Ravaging.

hint: deep non sequitur

Launching roses with your hand purely ridiculous

 

[Rangers]

Stacking allows for lots of inter-die bandwidth, I suspect this is what Sony is pursuing, not larger total die area. Stack the CPU, GPU and some DRAM for super high bandwidth, - and have some more DRAM on a "slow" XDR or GDDR5 bus.

Anybody know if this would be commercially viable within 1-3 years?

 

[upnorthsox]

Charlie's article is up and it's stupid

http://semiaccurate.com/2012/03/02/sony-playstation-4-will-be-an-x86-cpu-with-an-amd-gpu/

He says Sony is pursuing 3D stacking, in order to make even more massive chips.

Never mind that:

-Is there any chance 3D stacking is commercially ready in 2013-14?

Absolutely

-Why bother? The top normal GPU's are likely outside the thermal limits of a console anyway. Therefore if it's raw power you're pursuing, 3D stacking is irrelevant.

Bandwidth, bandwidth,bandwidth

-He also says its an SOC or could be an SOC at points, again if you're pursuing max power that in itself is contradictory

A SOC doesn't have to be a single chip, but even if it is, direct attach (or via an interposer) is still possible and makes for an efficient UMA design.

-Sony's financial status probably precludes them losing a ton of money on PS4, and the only other option is price it really high and that's no option after PS3 debacle.

Stacking though is one of the sure ways to reduce costs over the gen. It's also tech that Sony currently uses and could become the leader of which may pay for itself right there. There's certainly no gaurantee that there will be any node shrinks during next gen.

My point is if you're looking to outmuscle the other guy you almost surely dont need exotic 3D stacking to do it. Just get a bigger GPU than him.

128bit vs 512bit or even 1024bit. That's alot of muscle

 

[ultragpu]

SemiAccurate’s sources won’t spill the beans on the exact generation of CPU and GPU that are in the PS4, but we expect it to be a very customized version of an existing or near future design.

Sounds like a modified 7000 series gpu to me. And the uber amount of bandwidth nature of a heavily 3d stacked chip sounds like Sony really want a 1080p standard for PS4 games, theoretically speaking.

 

[GuestLV]

May be AMD kaveri will be in the xbox next ?

 

[upnorthsox]

Stacking allows for lots of inter-die bandwidth, I suspect this is what Sony is pursuing, not larger total die area. Stack the CPU, GPU and some DRAM for super high bandwidth, - and have some more DRAM on a "slow" XDR or GDDR5 bus.

Cheers

This

And the thing with consoles is, if you don't, you've missed out for the next 7-8 yrs. You can't later on go "oh we'll just go stacked now and bump the bandwidth from 128 to 512". Look at what happened to MS with Valhalla, they went to a SOC chip but couldn't take advantage of any intercore efficiencies.

There's some risk with being on the cutting edge but this isn't something where they'll get stranded in a deadend tech, everybody is going there and there's lots potential partners out there looking to help defray costs.

I could see a hybrid UMA design where the split is by bandwidth. It makes a lot of sense and should be fairly simple to implement.

 

[Rangers]

Absolutely

Why are there no products using this from AMD, Nvidia, Intel, companies much more on the cutting edge, and who's lives depend on pushing cutting edge, than Sony?

I'm pretty skeptical right now.

Stacking allows for lots of inter-die bandwidth, I suspect this is what Sony is pursuing, not larger total die area.

But that isn't what the article says, that's your spin, which gets at Charlie's credibility which gets at this likely being false.

 

[upnorthsox]

I think Charlie just read this article and made up the rest with his imagination. Since he seemed really a bit crazy with some of his Kepler reporting too I'm beginning to seriously doubt his credibility.

Or maybe it a continuation from this:

http://semiaccurate.com/2011/10/27/amd-far-future-prototype-gpu-pictured/

I certainly wouldn't be surprised if AMD launched a stacked mem gpu at about the same time that the next gen consoles launch.

 

[Rangers]

Or maybe it a continuation from this:

http://semiaccurate.com/2011/10/27/amd-far-future-prototype-gpu-pictured/

I certainly wouldn't be surprised if AMD launched a stacked mem gpu at about the same time that the next gen consoles launch.

You've got not later than 2014 (even then would allow a year head start probably). Wont be Sea Islands, so it would have to be Sea Islands successor.

We're hearing increased rumblings of PS4 already so it may be 2013 now.

 

[upnorthsox]

Why are there no products using this from AMD, Nvidia, Intel, companies much more on the cutting edge, and who's lives depend on pushing cutting edge, than Sony?

I'm pretty skeptical right now.

So you ask about the possibility in 2013-14 and then dismiss it because of today?

AMD

Nvidia:

LeiLei Zhang, of NVIDIA, made what could become the rallying cry of the upcoming 3D decade when she said “Scaling is ending. Let’s get over it and move our resources elsewhere.” Zhang declared that for them bandwidth is the issue. She indicated that NVIDIA is likely to use a turnkey solution such as TSMC is offering with such 2.5D TSV solutions entering the NVIDIA roadmap with their TESLA and CUDA high end networking GPU product lines

Intel:

Of course, they are going with memory on an interposer with Ivy Bridge so that'll be before the end 2012.

 

[3dilettante]

A SOC doesn't have to be a single chip, but even if it is, direct attach (or via an interposer) is still possible and makes for an efficient UMA design.

SOC is by definition a single chip.
If it's a 2.5D arrangment on an interposer, is better described as a SOP.

Using an interposer seems much more feasible than stacking with the CPU and GPU in the time frames involved.

Charlie's statement about BD's resemblance to Cell doesn't look valid on any important point.
They both use transistors, which is a similarity, I suppose.

 

[DieH@rd]

So, any chance of BC?

 

[3dilettante]

Straight usage of legacy code would be right out.

With binary translation, it may be possible for games that don't push PPE and the SPEs hard (or use SPEs at all) to run on a different architecture. This depends on Sony caring enough to spend resources on a solution like that.

Anything that really depends on quirks in the PS3 or drives any system component very hard probably won't.
This may rule out anything but the earliest or least demanding games.