Cell being shrunk to 22 nm? *spawn

Not related to anything, but she also had this in her linkedin:

migration of PS3 Cell processor to 22nm
Responsible for top-level large-block synthesis of the entire SPU (Synergistic Processing Unit). Was able to achieve route-ability and timing closure through sophisticated synthesis techniques involving detailed floor-planning, custom pre-wires and latch placement, custom routing algorithms, and soft/hard placement boundaries.

Superslim was still 40nm/45nm if I'm not mistaken, so another revision coming?
 
Last edited by a moderator:
Wouldn't that be tiny and low power? Could a 22 nm Cell see a reinvention of the platform in CE devices? I expect it's still as good as anything for running upscaling algorithms and interpolation. It's main problem was size and heat - you'd need a fan in your TV. If small enough and cool enough, it'd be as capable as dedicated silicon but far more versatile. A Cell in PS4 (missed opportunity) would be an ideal audio processor, post-FX processor, and provide BC. Looking forward to future hardware, Cell would still be competitive as a post-FX and audio processor, yet more versatile and programmable than a GPU or DSP.

So, is Sony wanting this just for PS3superslim supercool, for another 10 million units, or have they an idea on some grander vision?
 
Considering the price they are still asking for PS3 it does make sense to do yet another shrink. If they're going to keep it alive for another few years surely they can't keep asking consumers to pay any more than $199 once the PS4 is out. They won't lower the price that much straight away, but it will have to get down to that cheap. Hell, the PS2 and PSone went down to $99 IIRC... that clearly won't be happening this generation.
 
I wonder how small that would be.
The power consumption, heat and noise will be much better i'd imagine.
 
Last edited by a moderator:
B3D article by Carl. 65nm Cell was 174mm^2. 45nm Cell was 115 mm^2. IO was considered a limiting factor in the shrink. 22 nm will be two whole nodes, so 1/4 of the size in ideal terms. That'd be about 30 mm^2, but that's clearly not going to happen. To move Cell out of PS3 would require dropping XDR. If it's just a shrink for PS3, I'm curious where Sony see their market. Is this for a cheapo box to sell in the RotW territories? PS2's sales plummeted after PS3, but it had sat at the much cheaper price for a long time. Maybe Sony sees 30-40 million PS3s can be sold at the much cheaper price around $100?
 
Considering the price they are still asking for PS3 it does make sense to do yet another shrink. If they're going to keep it alive for another few years surely they can't keep asking consumers to pay any more than $199 once the PS4 is out. They won't lower the price that much straight away, but it will have to get down to that cheap. Hell, the PS2 and PSone went down to $99 IIRC... that clearly won't be happening this generation.

another shrink yeah but i didn't think sony nor ms would bother past 28/32nm...didn't seem cost effective to push the bleeding edge.
 
It would appear Sony have skipped 32 nm. Perhaps that's the future, with two full nodes delivering what was one full node reduction of the past decades? We could be looking at 22nm Durango and Orbis, then no shrink until 10 nm in 2016. Which would also help explain the choice of power. Imagine launching a monster chunk of silicon in your console and being unable to cost-reduce it for 3 or 4 years!
 
A Cell in PS4 (missed opportunity) would be an ideal audio processor, post-FX processor, and provide BC.
Perhaps, but cell has the drawback in that it requires its own dedicated RAM, which carries a noticeable cost premium, particularly if XDR is required - and you would pretty much need it if PS3 compatibility is the goal, for the bandwidth... Only practical way to reach 25GB/s with DDR3 is using a 128 bit bus, and how do you find RAM chips today to reach a lowly 256MB over such a wide (comparatively speaking) bus?

If MS wanted to do 360 BC in durango by hanging a xenos off of the APU they could re-use the onboard DDR3 since 360 did not have a memory controller integrated into the CPU itself. Sony doesn't seem to have the same luxury, and in any case I doubt they want to do it anyway, since hardware BC in PS3 was both half-assed and expensive. I remember reports of both image quality and compatibility issues way back then.
 
If they have to replace the memory controller, what about a pair of wideIO chip identical to the one in the Vita? That would be 256MB@25GB/s
 
22nm CELL will enable creation of substantially smaller PS3 console, but that will not come soon. PS3 Superslim was released only ~6 months ago.
 
22nm CELL will enable creation of substantially smaller PS3 console, but that will not come soon. PS3 Superslim was released only ~6 months ago.

Yeah they won't revise the hardware in terms of appearance. They'll just sell the current super slim with the 22nm chips at a higher profit margin at first, and then have more room to drop price once PS4 is happily on the market.
 
I can't see there being another PS3 revision fullstop. The PS3 SuperSlim was only released 6 months ago and is already significantly smaller, lighter and cheaper than the original. I actually picked one up last weekend for £139 after my original PS3 60Gb Phat YLOD on me 2 weeks ago. They are also selling very well @ around 120k globally per week over the last few months....so I fully expect to only see a pricedrop and for retailers to sell the unit around £99 within the next year.
 
B3D article by Carl. 65nm Cell was 174mm^2. 45nm Cell was 115 mm^2. IO was considered a limiting factor in the shrink. 22 nm will be two whole nodes, so 1/4 of the size in ideal terms. That'd be about 30 mm^2, but that's clearly not going to happen. To move Cell out of PS3 would require dropping XDR. If it's just a shrink for PS3, I'm curious where Sony see their market. Is this for a cheapo box to sell in the RotW territories? PS2's sales plummeted after PS3, but it had sat at the much cheaper price for a long time. Maybe Sony sees 30-40 million PS3s can be sold at the much cheaper price around $100?


Yep, clearly it'd need to be a SoC and even then it might be too small or it might have to contain more than cpu+gpu. That's 2 years away minimum though.
 
Yep, clearly it'd need to be a SoC and even then it might be too small or it might have to contain more than cpu+gpu. That's 2 years away minimum though.
But this is what makes me curious, because in two years time, is there really going to be much market for a PS3? It seems a considerable investment in a product, for which there must be an expectation of considerable sales. Maybe the target is RotW with a view to a few tens of millions, but only when the box is cheap enough?
 
But this is what makes me curious, because in two years time, is there really going to be much market for a PS3? It seems a considerable investment in a product, for which there must be an expectation of considerable sales. Maybe the target is RotW with a view to a few tens of millions, but only when the box is cheap enough?

Maybe they intend to release a PS4 SKU with hardware BC a few years down the road, after they have cost reduced enough to make it viable at $399? Or a BC addon? Sony has quite a collection of PS3 content on the PSN, you would think that they would want to find a way to continue leveraging that content even after the PS3 is discontinued. I'm sure some PS3/PSN games will see ports/remakes for the PS4, but certainly not the majority.

Or perhaps they will need these for Gaikai, if they really plan on streaming PS3 games. Assuming the service takes off they will need a Cell/RSX cluster. The smaller and less power hungry the better.
 
I prefer a Cell plugin instead of a new PS4 SKU.

Cell for Gaikai server sounds interesting too if they can solve the latency issues adequately.
 
Or perhaps they will need these for Gaikai, if they really plan on streaming PS3 games. Assuming the service takes off they will need a Cell/RSX cluster. The smaller and less power hungry the better.

Uhh, That´s a bingo!

Don´t you americans say that?

...

Jew hunting aside, seriously, I think that is the most reasonable explanation to such investment. Using that on the next super-slim models doesn´t hurt either, but I guess if that was the only use sony wouldn´t have bothered with the engineering efort.
 
Maybe they intend to release a PS4 SKU with hardware BC a few years down the road, after they have cost reduced enough to make it viable at $399?
I wouldn't have thought so. BC is most valuable early on. Once a platform is established, people typically play new games over old and BC gets removed to cheapen a console.

Or perhaps they will need these for Gaikai, if they really plan on streaming PS3 games. Assuming the service takes off they will need a Cell/RSX cluster. The smaller and less power hungry the better.
Interesting idea.
 
Back
Top