Sony selects Synopsys' phase shift mask software for 65nm node

[one]

I assume "Cell = 65nm" is a conjecture by fabtech.org. It makes more sense that they use 65nm for PSP that is more sensitive to power consumption.

 

[Carl B]

I think though that power concerns for PSP aside, the savings on cost by going to 65nm with Cell outweigh any benefit from marketing they could do with improved battery life for PSP. If 65nm indeed is ready in time, I'd expect it for Cell before PSP - and beyond that Sony themselves have been targeting Cell at 65nm with the Nagasaki expansion.

 

[one]

I think though that power concerns for PSP aside, the savings on cost by going to 65nm with Cell outweigh any benefit from marketing they could do with improved battery life for PSP. If 65nm indeed is ready in time, I'd expect it for Cell before PSP - and beyond that Sony themselves have been targeting Cell at 65nm with the Nagasaki expansion.

It depends on how good the yield of 65nm Cell is compared with that of 90nm Cell as they have to manufacture as many Cell as possible before the launch. For cost savings, the PSP chip and EE+GS will benefit from 65nm too and they are apparently smaller than Cell so yields in 65nm will be better than Cell.

 

[Carl B]

It depends on how good the yield of 65nm Cell is compared with that of 90nm Cell as they have to manufacture as many Cell as possible before the launch. For cost savings, the PSP chip and EE+GS will benefit from 65nm too and they are apparently smaller than Cell so yields in 65nm will be better than Cell.

Well I mean EE+GS and the Dragon chip are defintely smaller then Cell, and thus yeah would probably have better yields, but I think that the benefits of the move for Cell are just potentially much greater. Nagasaki's extension is ostensibly for Cell; there are numerous articles I could pull up referencing Sony's drive for 65nm for Cell.

Now I don't think that 65nm Cells in launch PS3's are by any means a sure thing, but eventual half-price Cells save more money than half-price PSP chips do.

Has there been any indication that Sony will move to SOI for chips other than Cell?

 

[one]

Well I mean EE+GS and the Dragon chip are defintely smaller then Cell, and thus yeah would probably have better yields, but I think that the benefits of the move for Cell are just potentially much greater. Nagasaki's extension is ostensibly for Cell; there are numerous articles I could pull up referencing Sony's drive for 65nm for Cell.

Now I don't think that 65nm Cells in launch PS3's are by any means a sure thing, but eventual half-price Cells save more money than half-price PSP chips do.

Has there been any indication that Sony will move to SOI for chips other than Cell?

SCEI Fab2 (Nagasaki) has SOI lines, yes. But OTSS (the Toshiba joint fab at Oita) has 65nm DRAM CMOS technology lines so this will be the line for PSP along with 90nm lines for RSX.
http://www.sony.net/SonyInfo/News/Press_Archive/200402/04-0202E/
I guess it's just about which, Nagasaki 65nm or Oita 65nm, comes online first.

EDIT: As for Synopsys' PSM, Toshiba licensed it in 2004 for their 65nm CMOS5 process
http://www.synopsys.com/news/announce/press2004/psm_pr.html

 

[overclocked]

If they could launch i would be very suprised, but who knows?
360 getting their 65nm cpu about six months after or something 07/Q2 or there about.
So it seems a bit to early atleast to me.

But i wonder how much the redundancy could be of a plus to make it possible and economically feasible?

There was some speculation about the different revisions of Cell so maybe they havent started production yet and could go 65nm, as nothing seems to be set in stone.

Could daydream a little i guess.

 

[Carl B]

Focusing on Cell alone and taking OTSS out of it (since they're not wrapped up in an IBM technology agreement), I wonder if and by how much Chartered trails IBM on the 65nm process, since Chartered is using IBM's process tech on that node? Also I wonder if Nagasaki in fact enjoys any sort of lead on East Fishkill since they began construction first, or if they are in lock-step with IBM, since again Nagasaki is in part reliant on IBM process technology.

I have a hard time understanding IBM's process agreements in general honestly; with co-development pacts with Sony/Toshiba, AMD, and Chartered - all seperately - what exactly comes out of this amalgamation, and how much do developments within one group effect developments in the others?

 

[PSman]

I hope that we'll get the 65nm Cell at launch

 

[Carl B]

I hope that we'll get the 65nm Cell at launch

What are the chance of us getting a 65mm CELL at launch?

You know I hear this a lot in various places - and it's not like I'm not hoping for 65nm at launch - but I do have to ask... why do you want 65nm at launch PSman? What do you think the ramifications will be for you the consumer? Now I have a list myself, but it does not include upgraded specs, which is where I feel a lot of people let themselves be led when they think of process shrinks.

PS - I'm not trying to get down on you PSman, I just want to sort of establish that 65nm Cells at launch wouldn't equate to faster Cells, or more RAM due to lower costs - which for some reason is where a lot of people seem to take it.

I've said it before but really if Sony did manage 65nm at launch, the big benefits would be theoretically quieter cooling, a smaller/cheaper/weaker power supply, and a case size and aesthetic that reflects the smaller packaging and more importantly the better thermals. Now since the case design is going to stick with the console for ~five years or so, I think that a launch at 65nm vs 90nm would have a material long-term effect in that sense. Price I don't really see as an issue, save for Sony's own costs. If they don't launch at 65nm, 65nm will be coming 'around the corner' enough such that I think they'll be willing to price the console as if it already had 65nm tech inside.

 

[Tap In]

You know I hear this a lot in various places - and it's not like I'm not hoping for 65nm at launch - but I do have to ask... why do you want 65nm at launch ....

speaking for myself I would guess that it is awfully hard to stuff all that hardware into a small package and to keep it cool (plus a big power supply) just ask MS.

So for me, I think it would mean less chance of over heating issues (as MS has had a few) and perhaps even a better launch price?

so less potential for issues and possibly Sony able to afford to take less of a loss and offer a better price point would be my reasons. :smile:

I'm really hoping for $399 although I'm thinking $499. I'll probably avoid the mass hysteria that launch brings and won't see a PS3 until Feb/March of next year either way but $399 may be more feasible at 65nm, unless I'm up in the night?


EDIT:" xb SAID:
"If they don't launch at 65nm, 65nm will be coming 'around the corner' enough such that I think they'll be willing to price the console as if it already had 65nm tech inside."

ahhh, that's a possibility too I suppose.

 

[overclocked]

speaking for myself I would guess that it is awfully hard to stuff all that hardware into a small package and to keep it cool (plus a big power supply) just ask MS.

So for me, I think it would mean less chance of over heating issues (as MS has had a few) and perhaps even a better launch price?

so less potential for issues and possibly Sony able to afford to take less of a loss and offer a better price point would be my reasons. :smile:

I'm really hoping for $399 although I'm thinking $499. I'll probably avoid the mass hysteria that launch brings and won't see a PS3 until Feb/March of next year either way but $399 may be more feasible at 65nm, unless I'm up in the night?


EDIT:" xb SAID:
"If they don't launch at 65nm, 65nm will be coming 'around the corner' enough such that I think they'll be willing to price the console as if it already had 65nm tech inside."

ahhh, that's a possibility too I suppose.

Its cooler to have the original with the big chips, not "tiny" wannabe shrinks!

 

[borntosoul]

on 65nm sony will be losing less money on each consol, i think thats the underlying reason. not to mention heat issues. i hope they do launch at 65nm. but like someone else said that even if they launch at 90nm, 65nm is just around the corner anyway. i doubt sony would wonna wait any longer and miss this christmas season. when did intel first produce retail cpu on 65nm?

 

[randycat99]

who left this bucket of gasoline sitting here?

My left-field theory- the possibility of smaller, cheaper chips could respawn the emergence of the "dual-Cell" PS3 (not unlike the 5-assed monkey). The twist I would add, is it wouldn't be a "symmetric" dual Cell. It would have the already expected 1 Cell with the customary 7 spe complement. Then there would be the 2nd Cell, with only a 5 spe complement. Why "5"?...because that would be a nifty way to use up that pile of Cell chips where 2-3 spe's are out due to defects. It's a "free" chip for the taking, so to speak! It was headed for the junkpile w/o 7 or 8 running spe's, otherwise.

Hell, why not go really perverse with the concept and allow for 3-chained Cell chips?! A 7 spe + 5 spe + 4 spe...the yields just get better and better! (a 7 + 6 + 3, 6 + 6 + 4, 6 + 5 + 5 combo shouldn't be discounted, as well- just ship whatever you find in the pile, eh?) On a napkin calculation level that would yield performance something near the server-esque 8x2 spe configuration. .5 TFLOPs in a PS3 isn't sounding so implausible, afterall. Tell me that sort of hijinks wouldn't be worth the motherboard design to accomodate it?!

I'm getting myself hot, thinking about the possibilities!

 

[akira888]

At this point what's done is done - the final PS3 we see will be virtually identical to what the current spec sheet indicates. There's not going to be a two PPE Cell, or anything else other than perhaps some clock frequency tweaks.

 

[randycat99]

The question is- have we really seen the final PS3? If Sony was to keep a secret weapon underguard, I would think this would be a goody of a secret to keep...but alas, I already admit I was thinking of the fanciful. Damn, if they didn't think of the "7/5/4" as a possibility when they were deciding what a PS3 should be, though... It would have been the uber yield power move, imo.

 

[Shifty Geezer]

Yields on those low-SPE components are likely to be very low. If a 7 working SPE cell is produced at 75% yield, you're only getting 25% yield on 6 SPEs and less. If we assume all those are 6 working SPEs, to produce PS3 with a 7 SPE Cell + 6 SPE Cell, Sony would have to make three 7 SPE Cells for each 6 SPE Cell. They will be production limited to 6 SPE Cells. Or, they have a dual-redundancy so they can use half their 7 SPEers as 6 SPEers. That'd put in PS3 a Cell with 8 SPE's with 1 disabled (deliberately or due to flaws) and another Cell with 8 SPEs and 2 disabled. That is a little under 2x the power, at probably greater than 2x the complexity and cost. Sony would lose any cost advantages of a process shrink, meaning high losses for some 2 years maybe, and any cooling advantages of the process shrink, meaning hotter, noisier PS3s for years.

Sony have announced the Specs. 1 Cell, 1 PPE, 7 SPEs. They're not going to change the CPU at the last minute with a totaly redesinged mobo and a need for devs to not only work on multiple SPe cores, but also multiple PPEs, all sharing that same RAM bandwidth that was looking so good for 1 Cell. Honestly, there's enough processing capacity to use up the rest of PS3's resources. Adding another Cell is just adding dead weight.

 

[randycat99]

That would be counterintuitive to the whole scaleability paradigm. Cell is entirely designed on the notion of scaleability and granularity. I can understand how local memory bandwidth contention can become an ultimate bottleneck, but I don't believe additional Cells can accurately be deemed "dead weight", nor do I believe the effect of contention will be an abrupt brick wall. Additional Cells will likely produce additional levels of performance, albeit with a gradual increase in diminushing returns with increasing numbers of Cells. "Complexity" need no more to be an issue than the LAN Cell scenario. It is inherent to the overall design. Additionally, the design for dual Cell motherboards is already in existence. So I don't see the notion of "complexity" as any great issue.

On the count of what Sony has spec'd, it is not necessarily a guarantee of what will be literally built (though, I grant it likely is what will be built). It can just as easily be they are simply stating what you can expect as a minimum. If you are designing a game/software to a target, design you can count on "this configuration" at the very least.

I am surprised that yields are 75%, however. I think that is tremendous if that is the case. Will it be the same on the 6nm process? Perhaps, not. Hence the opportunity to make some serious lemonade out of lemons, I think. I don't believe "cost" is a great issue, either. It will be secondary to "recoverable yields". "Recoverable yields" is essentially a hedge fund for primary yields.

I just wonder if KK was thinking about this on some deep down level. Then his colleagues said you are seriously nutzo to base your strategy on something like that. I think all of us here can agree that KK is nutzo enough to conjure something like this, not necessarily execute on something like this, right?

 

[Megadrive1988]

prediction: PS4 CPU, 2nd generation CELL architecture.

configuration:

10 upgraded PPEs, 128K cache each (only 8 are active for redundancy / yeilds)
68 upgraded SPEs 1MB local storage each ( 64 SPEs are active, 2 reserved for OS)

(w00t 72 active processors)

clockspeed is between 5 and 9 GHz (since clockspeed increases are not as large anymore)

whatever the single-precision floating point performance is, the doube-precision performance will only go down to 1/3 or 1/4, instead of doing down to ~1/10 as it is with CELL now (right?)

or, instead of so many PPEs (8/10) and SPEs (64/68) a 3rd type of processor is added to the 2nd generation CELL. the number of these will be fewer than the SPEs, but they will specialize in assisting with graphics more than the SPEs do, and will interact the most with the next-next-gen NV6x or NV7x GPU


oh yeah, 32nm process, maybe 22nm, if Sony gives PS3 a 7 year life.



ahh i am having WAY too much fun.

 

[Shifty Geezer]

ahh i am having WAY too much fun.

With way too little regard for plausibility!

 

[PSman]

You know I hear this a lot in various places - and it's not like I'm not hoping for 65nm at launch - but I do have to ask... why do you want 65nm at launch PSman? What do you think the ramifications will be for you the consumer? Now I have a list myself, but it does not include upgraded specs, which is where I feel a lot of people let themselves be led when they think of process shrinks.

PS - I'm not trying to get down on you PSman, I just want to sort of establish that 65nm Cells at launch wouldn't equate to faster Cells, or more RAM due to lower costs - which for some reason is where a lot of people seem to take it.

I've said it before but really if Sony did manage 65nm at launch, the big benefits would be theoretically quieter cooling, a smaller/cheaper/weaker power supply, and a case size and aesthetic that reflects the smaller packaging and more importantly the better thermals. Now since the case design is going to stick with the console for ~five years or so, I think that a launch at 65nm vs 90nm would have a material long-term effect in that sense. Price I don't really see as an issue, save for Sony's own costs. If they don't launch at 65nm, 65nm will be coming 'around the corner' enough such that I think they'll be willing to price the console as if it already had 65nm tech inside.

I always thought that 65mm CELL mean high specs and more RAM , guess i was wrong. Thanks for clearing that up:smile: