PSX 90nm issue - Chipworks gives their opinion

[Guden Oden]

I think he was suggesting you leave the thread - and I would agree with him. If you don't like this thread then stop commenting.

Whatever for?!

Just because I don't find the topic meaningful doesn't mean I don't have just as much a right as the next person speaking out on the subject (especially since I am NOT being either abusive nor disruptive), or that my views are any less valid. Jesus! If all you want, Dave, is a bunch of backslappers, you should rename the site accordingly and kick out those with differing opinions.

Given the importance people place on process (quite correctly) and the number of links from people in this forum I've previously seen on Sony being the the first to deliver 90nm processes, discussions on how that is achieved and whether we are comparing process apples to process apples is a viable and relavent discussion.

One could think so, but if 'people' have neither the knowledge nor the education or even the means neccessary to speak out on the subject, what exactly can be accomplished? Is there any doubt there are .09u features in Sony's IC? No, I don't think anyone here or otherwise would deny that there are. So obviously, Sony's process is working. What are we to infer from the fact not ALL of the chip is done in .09 tech?

Answer is easy: nothing, and that because we lack knowledge of the reasoning leading up to the particular implementation made in that chip in question.

Except, people don't want to infer that which they should (ie, nothing), they draw far-reaching conclusions supported by MORE nothing, such as Sony's process not working, .09 tech in general will lead to no performance increases etc.

What we are doing is akin to Plato and his philosophy students debating the make-up of matter. We know NOW everything is not a combination of earth, water, fire and air, but only because we built machines telling us otherwise. However, we are Plato in this thread, we don't have the machines neccessary to take the discussion to any kind of meaningful level. It'll just be a philisophical debate arguing over wether it was ethical of Sony to call their EE+GS@90nm a .09u chip or not, and as far as THAT argument goes, we've already seen most regulars polarize into two diametrically opposite camps (probably no small reason as to why the first thread discussing this got locked I would think).

Not unexpectedly divided pretty much straight along the likes-Sony/doesn't-like-Sony rift, might be added...

So tell me again, because maybe I am a bit stupid, what exactly IS it you think can be accomplished here, really? Can we convince Sony to stop calling the GS+EE@90nm a .09u chip, perhaps? Or will there just be a whole lot of flapping of lips (or finger-taps on keyboards as it may be) and very little to show for it all?

 

[marconelly!]

Sorry but this is one of the dumbest things I've ever heard.

Your saying we shouldn't care that we were lied to . You say its not big deal. I ask what is next . THey lied about this . Why wont they lie about something else ? ANd if no one cares about it this time mabye next time it will be a bigger deal . Mabye next time they will lie about a console performance . Or mabye something else .

This is an important topic. The truth is important . I don't care for being lied to .

Then, is Intel lying as well? According to those 90nm rules, their announced specs don't fall into the 90nm guidelines as well. Or could it be that these manufacturers interpret those guidelines a bit more freely and announce what they are doing accordingly? There's *some* 90nm tech in it (or as close to it as Intel plans to get as well) and they decided to call it a 90nm chip.

What about all those silly PRs with twisted numbers Microsoft likes to put up? Last time I checked many people brush it off saying something like "It's just PR, it's meant to twist the truth". I'd say these 90nm announcements are twisting the truth as well then. Either that, or we call all of that lies.

What about the commercials for various Burger Kings or McDonalds, that all show those big, vibrant looking burgers, whereas the real thing is nothing alike? Or those Wallmart commercials where their 'employees' happily talk about how much they enjoy working there, when in reality, Wallmart has the lowest sallaries imaginable, for the types of jobs it offers? We are being 'lied' on a daily basis, and if you are going to obsess about such things every morning when you wake up and every night when you go to the bed, you might as well start living under the tree and cut yourself from the civilized world, to save yourself from insanity.

 

[notAFanB]

It'll just be a philisophical debate arguing over wether it was ethical of Sony to call their EE+GS@90nm a .09u chip or not, and as far as THAT argument goes, we've already seen most regulars polarize into two diametrically opposite camps (probably no small reason as to why the first thread discussing this got locked I would think).

which 2 are you refering to the It'sNotRelevent/YesItIs or the EEis.09u/noIt's not camp?

the first camp were talking over each other, not so much ignoreing each others points rather dscussing almost 2 entirely diferent subject in order to mask their bias.

the second however is more interesting and is up for debate.

Not unexpectedly divided pretty much straight along the likes-Sony/doesn't-like-Sony rift, might be added...

it's just you man.

So tell me again, because maybe I am a bit stupid, what exactly IS it you think can be accomplished here, really?

discussion without ridicule maybe, that and whether or not Sony were skretching their competence to impress the shareholders (thus distorting the market somewhat).

Can we convince Sony to stop calling the GS+EE@90nm a .09u chip, perhaps?

never the point of discussion.

Or will there just be a whole lot of flapping of lips (or finger-taps on keyboards as it may be) and very little to show for it all?

This IS A DISCUSSION FORUM! there is anything to show anyone except when each of us leaves this desk (for a piss mostly).


Marc:

Either that, or we call all of that lies.

I pick Lies, Damn lies and statistics myself.

 

[Dave Baumann]

Jesus! If all you want, Dave, is a bunch of backslappers, you should rename the site accordingly and kick out those with differing opinions.

99% of the topics on these forums are over discussions on differing opinions, however you are not discussing opinions related to the topic of you are just making noise that the topic of discussion is pointless, which is not contributing to the discussion at all. If you don't like the thread you are not forced to comment on it.

 

[Ug Lee]

Sheesh, talk about an overreaction.

 

[Vince]

99% of the topics on these forums are over discussions on differing opinions

Unless the discussion is between Dave and myself

 

[cthellis42]

Re: ...

If I were the president of Sony, I would not have claimed that PSX2OAC was a 90 nm product. For all purpose and intentions, it does not even deliver the benefits of a true 90 nm process, the transistor density as witnessed by 150 million transistor Dothan.

Funny, I didn't notice this part first time around. Are the 90nm chips used in the Apple Xserve G5's not truly 90nm because they only have ~58 million transistors? Or--just possibly--are there more ways to use the tech?

BTW: Does anyone know what wattage the EE+GS officially burns at? If they slipped slightly on size, they may have slipped a bit on heat as well... but I don't know if there have been tests on the released product.

 

[Geeforcer]

And Geeforcer:

Grow UP, will you?

Quite an ironic suggestion coming from a person who spend half the tread whining about the merits of the ongoing discussion. The irony is redoubled considering that one would think a person so concerned with productivity would waste so much time chastising people about their choice of discussion. Now that you’ve made your feeling about the validity of the thread abundantly clear (over...and over...and over again), feel free to leave it and peruse discussions more suited to you penchant for productivity.

 

[nobie]

cant we all just get along

 

[cthellis42]

Not if "all" amounts to two or more people. Certainly not on all subjects.

 

[Deadmeat]

...

Funny, I didn't notice this part first time around. Are the 90nm chips used in the Apple Xserve G5's not truly 90nm because they only have ~58 million transistors? Or--just possibly--are there more ways to use the tech?

By way of comparison, Glaskowsky noted that IBM's PowerPC 970FX, a 90nm part, was about half the size of the IBM PowerPC 970. "It's considerably cheaper, about a third of the manufacturing cost, although the packaging is essentially the same," he said. "It's faster, consumes less power, and enables new systems like the [Apple Computer Inc.] Xserve G5."

Source

EE and GS together measures about 130 mm2 on 0.13 micron process. They should have shrunk it down to 65 mm2 or less on a "true" 90 nm process, but the PSX2OAC doesn't measure this.

IBM's "true" 90 nm process provides the cost-saving benefits by shrinking the die size to half, SCEI's "I don't know what this really is" 90 nm process doesn't.

 

[cthellis42]

It's not a "true" 130nm process either. It's smaller than the one and larger than the other, and basically takes what it needs of both. Regardless, that was just to point out that your transistor comment makes no sense since you do not need to stack on a lot of transistors to be "truly" utilizing a process, nor to make a good-performing chip, nor using 90nm as a necessity to keep heat down, nor...

The misleading nature of their press announcements is what one can take offense at, not the nature of the chips.

 

[Deadmeat]

...

The "benefit" of a 90 nm process is its higher transistor density; you can do with 90 nm

1. Put same number of transistors in a smaller die and lower your production cost, or

2. Put more transistors into same die size. More at same price.

PSX2OAC offers neither, it is not significantly smaller than its 130 nm parts nor does it offer more for same money. In other word, its purpose of existence is questionable.

 

[Paul]

its purpose of existence is questionable.

This takes the cake. Of all the things you have said, I really think this is it...

 

[nobie]

sounds like PSX is having an existential crisis

 

[Vince]

Re: ...

The "benefit" of a 90 nm process is its higher transistor density; you can do with 90 nm

1. Put same number of transistors in a smaller die and lower your production cost, or

2. Put more transistors into same die size. More at same price.

PSX2OAC offers neither, it is not significantly smaller than its 130 nm parts nor does it offer more for same money. In other word, its purpose of existence is questionable.

What? Seriously, do you ever stop to think? Do you ever think, period?

The EE+GS@90 actually does both, which is quite ironic. Not only are they taking each "discrete" part and utilizing the higher density to minimize each's size; but after redesigning each they've taken both and combined it into one IC which crams more transistors than each IC viewed singularly. So, you're totally wrong.

The last production Graphic Synthesizer was ~73mm2, the last production Emotion Engine was ~73mm2. The total physical is ~140mm2. The EE+GS@90 is ~90mm2.

90mm2 is roughly 35% smaller than ~140mm2

It is significantly smaller and it's purpose is obvious to anyone with more than two excitable cells as it's now a single IC, let this soak into your head. It's very important. By moving to 90nm and combining the ICs, there is nolonger a division of production; OTSS no longer "just" manufactures Emotion Engines, Nagasaki no longer "just" manufactures Graphic Synthesizers. They now independently produce both, virtually doubling production and dropping costs dramatically.

It's also clearly a 90nm chip as you can see the ~45nm gate's and their Dielectric yourself. These things just aren't possible at Sony's 130nm node, this is known fact. What we're seeing (as opposed to you) is that there is a process technology that's clearly 90nm, clearly among the most advanced in the industry, and clearly way ahead of the PC competition. The EE+GS@90 is a subset of this - it's clearly capable of production only @ the 90nm node. You can fabricate an IC with a 150nm gate @ 90nm, but you can't do the inverse - the sooner you realize this the sooner you'll stop this insanity.

 

[Paul]

The EE+GS@90 will also be the IC that gives backwards compatability to the PS2 when it's put inside the PS3.

 

[Deadmeat]

...

The last production Graphic Synthesizer was ~73mm2, the last production Emotion Engine was ~73mm2. The total physical is ~140mm2. The EE+GS@90 is ~90mm2.

The last gen EE was a 0.15 micron device. Scale it down to 0.13 micron fab and you get 54 mm2

54 mm2 + 73 mm2 = 127 mm2. You can do better than this by saving pad real estate.

90mm2 is roughly 35% smaller than ~140mm2

29% to be exact. Sorry to shoot you down.

It's also clearly a 90nm chip as you can see the ~45nm gate's and their Dielectric yourself.

The dispute is not over. PSX2OAC clearly doesn't have the transistor density of a true 90 nm device. That much is agreed.

These things just aren't possible at Sony's 130nm node

I say SCEI shaved off 10 mm2 from having to eliminate some pads and eliminating external GIF bus. Going from 117 mm2 to 90 mm2 isn't that groundbreaking.

 

[Vince]

Re: ...

The last production Graphic Synthesizer was ~73mm2, the last production Emotion Engine was ~73mm2. The total physical is ~140mm2. The EE+GS@90 is ~90mm2.

The last gen EE was a 0.15 micron device. Scale it down to 0.13 micron fab and you get 54 mm2

54 mm2 + 73 mm2 = 127 mm2. You can do better than this by saving pad real estate.

Fine, we'll use your numbers - just make it one less thing for you to bitch at.

Yet, amazingly, that's still a ~30% shrink observed in your calculated numbers (127mm2 -> 90mm2). Between your 130nm calculated numbers and the physical EE+GS@90 numbers from SI/Chipworks, is roughly equal to a lithography generation.

Wow DM, what an ironic happening, eh? Sony claims it's 90nm and it just happens to get a physical benefit that's proportional to the transistor shrink of an average generational process advance. Must be upthere with getting struck by lightning twice or the chance of intelligence occuring in the universe.

90mm2 is roughly 35% smaller than ~140mm2

29% to be exact. Sorry to shoot you down

• Number 1 - You didn't shoot me down, my numbers were correct in the context they were given in. I clearly stated the sizes were based of that for the "Last production..." and my numbers were correct.
• Number 2 - Your number is only correct when put in context of your calculated horsehit 130nm-normalized areas, which isn't accurate past what the rest of us consider a back-of-the-envelope calculation. Believe it or not, I don't think you'll find a person who will agree that your 5th grade proportionalities are going to yeild a precise result.

It's also clearly a 90nm chip as you can see the ~45nm gate's and their Dielectric yourself.

The dispute is not over. PSX2OAC clearly doesn't have the transistor density of a true 90 nm device. That much is agreed.

What does that have to do with anything? They clearly fabricated a device which contains a plurality of gates and constructs which aren't physically capable of being mass produced on anything less than a 90nm production line utilizing (IIRC) 193nm lithography equiptment. The 130/150nm process with it's 248nm equiptment just isn't physically capable of creating such physical constructs reliable and in quantities necessary for mass production. Period.

Density is an implimentation dependant feature. It will vary with each ASIC and vary within each region of an IC, it isn't a function of anything important to this debate except being upper bounded/limited by the smallest feature the process/lithpgraphy equiptment can create (which is known to be ~45nm). And Chipworks has clearly proved that Sony's EE+GS contains features which are consistent with 90nm and it's been in production since last Fall. Deal with it.

These things just aren't possible at Sony's 130nm node

I say SCEI shaved off 10 mm2 from having to eliminate some pads and eliminating external GIF bus. Going from 117 mm2 to 90 mm2 isn't that groundbreaking.[Bold = Vince]

Sorry to have to shoot you down... Iceman, but you've previously stated "127mm2" and then used that number to find that whole "29%..." "exactly" to 90mm2 bit. Actually you formulated it for the rest of us here:

54 mm2 + 73 mm2 = 127 mm2

Try to atleast be consistent within your own posts, okie dokie slick? Because from where I'm standing 30% is a pretty darn significant proportion of anything.

Concerning the GIF, I question that considering it's known area in proportion to the whole, but I disgress even more since you're actually in a position to speak about that.. and do it with conjecture which is capable of being verified. Everything I've stated is verifiable with research, as opposed to everything you've stated which isn't because it's this massive web of interlocked BS streaming out of your head.

Hey, someone sent me this video - you should check it out. Kinda funny.