Will PS2 Emulation on PS3 be PS2+?

[PC-Engine]

I'm pretty sure EE+GS@90nm right now costs around $40. I doubt it would be $1 when PS3 launches. There's no chip with 50 million transistors that cost $1. It costs money just for FAB space utilization and raw materials.

 

[jvd]

I personaly don't know how much the dragon chip costs. But i don't really think its 10$ . I doubt its 40$ either . Mabye somewhere in the middle .

But even if its down to 10$ in 2006 ... Thats still 10 million per million consoles sold . Figure what 10 million first year ... thats alot of money .... Man sony should jsut give me that money ....

 

[randycat99]

Is there some cosmic rule now that EE+GS are now inseparable??? Clearly, unloading the GS would save a good amount of transistors. A lot of things are possible, at this juncture. Making wide sweeping generalizations to preface what will or will not happen are largely irrelevant as far as us "outsiders" are concerned.

 

[jvd]

Is there some cosmic rule now that EE+GS are now inseparable??? Clearly, unloading the GS would save a good amount of transistors. A lot of things are possible, at this juncture. Making wide sweeping generalizations to preface what will or will not happen are largely irrelevant as far as us "outsiders" are concerned.

Who is making any generalizations . We are only talking about possible reasons for going emulation or going hardware .

Yes they can be seperated but wouldn't that require alot of work taking the gs say and moving it from its last process to 90nm or 65nm . I think the work on the shrinks alone cost alot while the ee+gs is already shrunk .

 

[randycat99]

Yes, cuz when they finally got EE+GS on paper, they naturally threw-out out all other design work prior to that milestone (so nothing new can be created except with EE+GS as a starting point)... Making something as bold as a Cell just happens so as to power a console that will most assuredly fail (according to some), but slapping together various EE variants together is just "unthinkable".

Maybe they wouldn't have to separate anything? Maybe there has been a 90nm EE on paper for some time now (before there was even an EE+GS on paper), perhaps? No way!!! ...because no one at Sony ever even considered the idea of backwards compatibility solutions for PS3, until now... Surely, they couldn't have 7 different solutions (and then some) to address this for quite a while ago. Billion dollar technology businesses surely aren't managed like that... Kutaragi just pondered one day, "Aw crap! How we gonna handle this PS2 compatibility issue in PS3?! What was I thinking? All we got on this floppy disk is a plan for EE+GS on 90nm. Aiyeeee, I wish I was that snappy, clever Bill Gates. He would never end up in a predicament like this..."

 

[jvd]

Yes, cuz when they finally got EE+GS on paper, they naturally threw-out out all other design work prior to that milestone (so nothing new can be created except with EE+GS as a starting point)..

Who said that ? But did they ever lay out the ee on 90nm by itself ? I don't know what work it would take to cut out the gs now that they are on the same die. It will need a new mask and most likely tap outs , i don't think it will be cheap .

Then again why put only half of it in ? Why do half emulation and half hardware ?

I think your being way to defensive of sony and its stoping you from having an actual conversation about this with out resorting to sarcasm .

I gave my opinions , which its apparent by your reactions and posts are for some reason not allowed . So i'm done here .

 

[randycat99]

Naturally, you feel anybody is defensive about Sony for one reason or other...but then you are fighting in some "war of the consoles" sort of ideology, so that would seem to be natural for someone like you. It couldn't just possibly be that some of the trains of thought here are fairly well deserving of sarcasm to shine some perspective on the issue.

I wonder how they did it in the PS2??? Hmmm...

 

[PC-Engine]

Nobody is saying it's impossible to do. Is it likely or practical? Who knows?

 

[Npl]

Then again why put only half of it in ? Why do half emulation and half hardware ?

EE could be used as IO-Processor (including Sound-DSP) similar to the PS1-CPU in the PS2.
GS is a wasted effort, for what do you think it could be used, else than Backwards-Compatibility? Additionally its for the most Part a block of static Ram, with very simplistic Rendering Functions - shouldnt be hard to emulate (if PS3`s GPU will have the 48GB Bandwith in some way).

 

[Panajev2001a]

I'm pretty sure EE+GS@90nm right now costs around $40. I doubt it would be $1 when PS3 launches. There's no chip with 50 million transistors that cost $1. It costs money just for FAB space utilization and raw materials.

So... Intel could manufacture the 400+ mm^2 Itanium 2 at about $140 and SCE cannot manufacture the 86 mm^2 EE+GS@90 nm chip at less than $40 ?

 

[Fafalada]

Cost and the guess that it shouldnt be hard to emulate( now not talking about pixelperfect acuracy). It should be done someday anyway( emulation in Software ), or do you expect the PS8 to include 7 dated Playstation-Versions in Hardware

I would argue that just having GS for backwards compatibility and emulating everything else might be cheaper then other hardware alternatives.

If you want EE in hardware, you automatically have to include IOP and SPU2 + 4MB of DRAM (let's assume for this case that we can make EE memory controller work with XDR in some simple way).
Now that's quite a few extra physical packages to add, and while it may be possible to integrate it all on one die (not sure if there are licensing issues though), that'd just be another chip with 4MB embeded (albeit much slower).

You're of course right though, that emulating GS in software shouldn't be too difficult to get working reliably, while the other components would be a bit more work to get quite right.
Still, the asociated costs of embeding almost entire PS2 board in there may be prohibitive, not to mention I am quite worried that some "clever" SCEI engineer would figure out a way to convince the higher ups that SPU(the PS1/PS2 sound core, nothing to do with Cell)*8(or some other number) would be a sufficient sound solution for the nexgen :?, and thus "justifying" sticking a bunch of legacy hardware in there.

 

[Fafalada]

My guess is that dropping in an EE sans GS somewhere in the PS3 is a nobrainer.

While GS you don't need, EE without the other components won't really be of much use, so if you're trying to cut costs, going full software emu would be the way.

EE could be used as IO-Processor (including Sound-DSP) similar to the PS1-CPU in the PS2.

You really do wish evil on the poor sound software engineers in nextgen games don't you...

GS is a wasted effort, for what do you think it could be used, else than Backwards-Compatibility? Additionally its for the most Part a block of static Ram, with very simplistic Rendering Functions - shouldnt be hard to emulate (if PS3`s GPU will have the 48GB Bandwith in some way).

I agree, and clearly Sony has been aware of that fact when they made PS1 compatibility in PS2 already - they didn't try including the outdated GPU in there either (though GS does have one or two hardware features helping the emulation go smoother).

 

[Panajev2001a]

Cost and the guess that it shouldnt be hard to emulate( now not talking about pixelperfect acuracy). It should be done someday anyway( emulation in Software ), or do you expect the PS8 to include 7 dated Playstation-Versions in Hardware

I would argue that just having GS for backwards compatibility and emulating everything else might be cheaper then other hardware alternatives.

If you want EE in hardware, you automatically have to include IOP and SPU2 + 4MB of DRAM (let's assume for this case that we can make EE memory controller work with XDR in some simple way).
Now that's quite a few extra physical packages to add, and while it may be possible to integrate it all on one die (not sure if there are licensing issues though), that'd just be another chip with 4MB embeded (albeit much slower).

You're of course right though, that emulating GS in software shouldn't be too difficult to get working reliably, while the other components would be a bit more work to get quite right.
Still, the asociated costs of embeding almost entire PS2 board in there may be prohibitive, not to mention I am quite worried that some "clever" SCEI engineer would figure out a way to convince the higher ups that SPU(the PS1/PS2 sound core, nothing to do with Cell)*8(or some other number) would be a sufficient sound solution for the nexgen :?, and thus "justifying" sticking a bunch of legacy hardware in there.

Oh come on, do not tell me that the I/O CPU could not be emulated in software by the CELL CPU .

PSOne emulation should be doable in a reliable way by the CELL CPU without too much trouble.

The GS could not function as an I/O CPU in PlayStation 3 and this way you would have to have to complicate the output to video section and not only that, but you would lose any possible enhancement to PlayStation 2 graphics.

 

[Fafalada]

Oh come on, do not tell me that the I/O CPU could not be emulated in software by the CELL CPU.

You have a 2-way asymetric multiprocessor system, with the two processors running different code and constantly communicating with one another via hw interrupts and DMA transfers into each other's memory.

There's no purely software solution that will make this work - you need to replicate the communication between the two CPUs on some level in hardware, it's not like just sticking EE and connecting it to XDR would make this possible.

And isn't this a selfdefeating proposal in the first place? You first insist on sticking EE in there for "hardware level compatibility", and then you go off and emulate the other half of CPU resources in software.
Is it really that much of a stretch to go from emulating legacy DMA, R3000 and the Sound DSPs in software... to including EE emulation as well?

The GS could not function as an I/O CPU in PlayStation 3 and this way you would have to have to complicate the output to video section

Actually no - GS has its own CRT controller onboard, all you have to do is connect it to the video out. At least that should be simpler then connecting CPU and XDR to EE.

 

[Panajev2001a]

Oh come on, do not tell me that the I/O CPU could not be emulated in software by the CELL CPU.

You have a 2-way asymetric multiprocessor system, with the two processors running different code and constantly communicating with one another via hw interrupts and DMA transfers into each other's memory.

There's no purely software solution that will make this work - you need to replicate the communication between the two CPUs on some level in hardware, it's not like just sticking EE and connecting it to XDR would make this possible.

And isn't this a selfdefeating proposal in the first place? You first insist on sticking EE in there for "hardware level compatibility", and then you go off and emulate the other half of CPU resources in software.
Is it really that much of a stretch to go from emulating legacy DMA, R3000 and the Sound DSPs in software... to including EE emulation as well?

It would definately be a chalenge as the EE is not so simple to emulate.

It is also true that in lots of CELL presentation slides they show the I/O controller as part of the CPU... the CPU acting also as an I/O controller basically so full software backward-compatibility might be an idea to follow.

If you accept to remove the GS from the EE+GS chip you free quite a bit of space and you could fit the R3000A+GTE+MDEC+DMAC+SPU2: I would leave the 2 MB+2 MB of DRAM for the I/O CPU and the SPU2 off-chip: you could even have a common 4 MB block and re-work some things in the PlayStation 2 I/O CPU and the SPU2 to make this possible.

This new I/O chip would likely take about 44 mm^2 or even a bit less and having 4 MB of off-chip DRAM would not be a problem as the trace count for that block would be likely quite low.

The GS could not function as an I/O CPU in PlayStation 3 and this way you would have to have to complicate the output to video section

Actually no - GS has its own CRT controller onboard, all you have to do is connect it to the video out. At least that should be simpler then connecting CPU and XDR to EE.[/quote]

I think the memory controller in the EE can be replaced by an XDR memory controller: it can obviously operate in such a way to pull only 3.2 GB/s of bandwidth and have latencies that are at worst equal to Direct RAMBUS DRAM latencies if not decisively better in average.

 

[PC-Engine]

I'm pretty sure EE+GS@90nm right now costs around $40. I doubt it would be $1 when PS3 launches. There's no chip with 50 million transistors that cost $1. It costs money just for FAB space utilization and raw materials.

So... Intel could manufacture the 400+ mm^2 Itanium 2 at about $140 and SCE cannot manufacture the 86 mm^2 EE+GS@90 nm chip at less than $40 ?

$40 is my estimate, what's yours? Oh btw $39 is less than $40, so is $38, $37, $36, $35..

Intel's chip is 5 times bigger. A quick calculation 140/5 = $30.

 

[Fafalada]

It would definately be a chalenge as the EE is not so simple to emulate.

True, but they could have been working on it for the past 5 years, and unlike most emulators they'd have access to complete hardware details all along. Well, it's just one possibility I guess.

If you accept to remove the GS from the EE+GS chip you free quite a bit of space and you could fit the R3000A+GTE+MDEC+DMAC+SPU2: I would leave the 2 MB+2 MB of DRAM for the I/O CPU and the SPU2 off-chip: you could even have a common 4 MB block and re-work some things in the PlayStation 2 I/O CPU and the SPU2 to make this possible.

Well yeah, that should be a cheap enough solution with likely most reliable compatibility results, I'm just wondering if there could be any licensing problems with integrating it all on the same chip.

I think the memory controller in the EE can be replaced by an XDR memory controller:

Yeah I kinda figure they'd do that too. Anyway, for now we can just wait and see, honestly I don't even know any decent rumours about what the setup for bw-compatibility is supposed to be like.

 

[Panajev2001a]

I'm pretty sure EE+GS@90nm right now costs around $40. I doubt it would be $1 when PS3 launches. There's no chip with 50 million transistors that cost $1. It costs money just for FAB space utilization and raw materials.

So... Intel could manufacture the 400+ mm^2 Itanium 2 at about $140 and SCE cannot manufacture the 86 mm^2 EE+GS@90 nm chip at less than $40 ?

$40 is my estimate, what's yours? Oh btw $39 is less than $40, so is $38, $37, $36, $35..

Intel's chip is 5 times bigger. A quick calculation 140/5 = $30.

Do you think that costs diminish only linearly ? You are placing chips on a round wafer... over a nice area.

Something tells me that a 500 mm^2 chip costs more than 2x to manufacture than a 250 mm^2 chip .

 

[PC-Engine]

I'm pretty sure EE+GS@90nm right now costs around $40. I doubt it would be $1 when PS3 launches. There's no chip with 50 million transistors that cost $1. It costs money just for FAB space utilization and raw materials.

So... Intel could manufacture the 400+ mm^2 Itanium 2 at about $140 and SCE cannot manufacture the 86 mm^2 EE+GS@90 nm chip at less than $40 ?

$40 is my estimate, what's yours? Oh btw $39 is less than $40, so is $38, $37, $36, $35..

Intel's chip is 5 times bigger. A quick calculation 140/5 = $30.

Do you think that costs diminish only linearly ? You are placing chips on a round wafer... over a nice area.

Something tells me that a 500 mm^2 chip costs more than 2x to manufacture than a 250 mm^2 chip .

Of course it's not as simple as that that's why I didn't bring up intel chips, you did.

You still haven't given an estimate for the chip, what's the matter too scared it won't be $10 or $1?

 

[Shifty Geezer]

Will the Cell SPU architecture be condusive to effecient emulation? I'd have thought not and software PS2 emulation would be handled entirely by PE.