Official: ATI in XBox Next

[Vince]

One word: Depends. Correct answer: all else cannot be equal by definition, so your question is either irrelevant or ignorant.

This is done all the time in the science, economics and mathmatics. All else being equal - what do you design for during your 5 year development cycle? 150nm or 130nm if both are avalable during your launch window?

Now you show me how that applies to Microsoft's model. Last time I checked, Microsoft isn't deisgning "monsterous and dedicated ICs". They're more or less picking parts off the shelf. (Of course, X-Box2 might be different than X-Box 1, but given the lead time of the ATI deal...not likely to be that different.)

You're in a console forum. You play the console game. At the present, Sony is raping the industry - they're the model to beat. As I've been showing... there are problems.

As cthellis42 has been trying to tell you, these are completely different models. Each of which, has it's own risks, advantages, and disadvantages. Sony's requires HUGE investments and lots of lead development time, but great flexibility and potential for "breakout". MS's has a lot less flexibility, a lot less risk, but less potential for some type of paradigm shift.

And this is fine, I agree. But, this still doesn't negate the fact that you need to beat your competition regardless of the model you utilize in your corperate structure. Do General Electric and ComEd not compete because GE's corperate model is completely different? Shit no.

So, ATI has 5 years to get xbox-2 chip out the door? So did nVidia?

Assuming a 2H 2005 launch they'll have had almost 3 years from first Microsoft tender IIRC. Again, this doesn't matter - if ATI can't do it in 2 years then they should have: (a) changed their model (b) found someone who could.

Which is why the X-Box GPU is a miserable failure? It only had 18 months or so from development to production? I though the X-Box console and GPu were holding their own quite well...

The game is different. The consoles should launch at the same time - development costs go up to compensate. Mfa's had several great posts on this and besides being a smart guy in the industry, he's opinion on Microsoft's chopice of ATI are interesting IMHO.

This is getting OT

 

[Vince]

t depends on the situation.

Find me a situation when a closed box targeted IC with 3 years of R&D would be better off with an inferior lithography process?

A simple exercise to (again) illustrate the point.

• According to you 65nm has more advantages, by far outweighing 0.13u or 0.09u in a closed box IC.
• No shipping closed box that I'm aware of currently has 65nm tech.

How can you reconcile those two "truisms"?

Simple Euclidian Geometry can solve the first. The second is because they're developing it now, rest assured that as semiconductor companies with it all inhouse (more like nVidia's style) they've been planning towards this process since 2001. This can be seen in their recent, combined, $7-8Billion investment in 65nm fabrication to come online in 2005.

This is what I'm talking about. You need to get a handle on lithography from a power-house (eg. IBM, Intel) and have the entire development pipeline pointed at an advanced goal. From here you can use past preformance and precedence to see the rest of my argument.

PS. Hope you don't mind, I reformatted your list.

 

[cthellis42]

As cthellis42 has been trying to tell you,

Ugh... Please folks, just cthellis! When I signed up for B3D initially my attempt somehow got screwed up and the name was locked to me.

As 42 is the answer, however, I never mind adding it to any of my nicks. Hehe...

That post by cthellis42 is great - and I agree very much with it. While the needs of MS might be met, I'm still dispointed (as are others) but it's only a console and it doesn't change my life.

Ah, funky. ^_^ I was hoping you'd at least get around to reading it. I had the odd feeling you and Joe would be sniping at each other for the next few hours...

I'm curious why you're still quite down on ATI coming from that angle, however. Wouldn't you agree that MS did not give nVidia as much time or enough resources to let them do what they might have wanted to and ENSURE it? Wouldn't it stand that ATI will be just as limited by some of the same constraints? (Though differently so due to the nature of this partnership. They do get to concentrate almost solely on the graphics chip, but they won't be involved on the same production levels as nVidia was. And though doubtless they have major input, the ULTIMATE decisions will rest with Microsoft.)

This is done all the time in the science, economics and mathmatics. All else being equal - what do you design for during your 5 year development cycle? 150nm or 130nm if both are avalable during your launch window?

You use 130nm. But as time has shown, things are rarely if ever equal...

Assuming a 2H 2005 launch they'll have had almost 3 years from first Microsoft tender IIRC. Again, this doesn't matter - if ATI can't do it in 2 years then they should have: (a) changed their model (b) found someone who could.

So in this case it would seem they have more time to spend on design, more time to work out with Microsoft who should fab, more time for Microsoft to determine if it should leverage it's weight with a fab to ensure proper production at the desired process... But the money isn't ultimately being leveraged by ATI, so they are dependant on others (as ATI and nVidia in general were dependant on TSMC)--most specifically Microsoft--and have both parameters to dictate and parameters they are GIVEN to work within.

I expect the Xbox2 won't be quite the mish-mash as the original, but they still have many things to cover, and are spending less time than their competition which will almost assuredly have the jump to market at this point...

In this whole situation I don't see ATI or nVidia being particularly superior is all. The critical design processes rest in one place--Microsoft.

 

[nelg]

What could be intresting is if Microsoft purchased the IP from ATI and gave the CPU contract to Intel with the condition of Intel fabing the VPU.

 

[Joe DeFuria]

This is done all the time in the science, economics and mathmatics. All else being equal - what do you design for during your 5 year development cycle? 150nm or 130nm if both are avalable during your launch window?

If you have a 5 year development cycle BOTH 150nm and 130m are likely not to exist. You are also willing to be a bit flexible on your exact launch window....probably about a 1 year time frame. That launch time frame will diminish in scope as time passes, and things start to pan out.

Given that, and the fact that you have some ability to adjust early on, I would probably try and shoot for 130nm initially.

You're in a console forum. You play the console game.

Last time I checked, X-Box was a console, and Microsoft is the console vendor. Did I miss something?

At the present, Sony is raping the industry - they're the model to beat.

And when you're the market leader, one of the things you enjoy is the ability to take higher risks. That doesn't mean the risks always pan out. Do you think Sony is as concerned with the timing of their next console, as much as Microsoft is?

You're always the "model to beat" right until the time that you get beaten. Not saying or predicting that this will happen, of course.

And this is fine, I agree. But, this still doesn't negate the fact that you need to beat your competition regardless of the model you utilize in your corperate structure.

That depends.

Is there room for 2 consoles? If so, then you don't have to BEAT them, you just have to be profitable. In any case, you don't have to use the same busines model as your competitor to compete or beat them.

Do General Electric and ComEd not compete because GE's corperate model is completely different? Shit no.

Did I say they didn't have to compete?

- if ATI can't do it in 2 years then they should have: (a) changed their model (b) found someone who could.

Agreed. What's your point? Do you think ATI could do in 2-3 years what Sony/T/I is doing over 5-6 years?

Yes, this is a console forum. It just so happens that ATI and nVidia were competing for console business that is nothing like Sony's model. Because Sony is the leader, doesn't negate the fact that other models exist.

 

[Megadrive1988]

You're in a console forum. You play the console game. At the present, Sony is raping the industry

sort of. Sony isnt dominating, raping and staggnating the console industry the way Nintendo did in the late 1980s with their 8-bit Famicom / NES. Sega put a stop to that in the U.S. with the Genesis and NEC put a stop to it, to some degree, in Japan with the P.C. Engine.

Sony is not that far off from doing what Nintendo did the late 80s, though, and I am certain that Sony would, if it could. thankfully 2 strong competitors are there to ensure it doesn't happen. hopefully it will remain that way well into the near-term, as well as the long-term. competition and more than 1 or 2 console hardware providers is desperately needed. In the mid-to-late 80s, their was only one main console player (Nintendo NES)
with two extremely weak players (SEGA-SMS, Atari-7800). The Playstation1 did extremely well, but there were two other consoles that did moderately well in at least one major world territory, Saturn in Japan and N64 in the U.S. - and now with the current generation, we have PS2 doing extremely well, but two other players with 10-20% of the market each (i dont know the exact figures on Xbox and GC's marketshare).


With the next generation, I expect both MS's and Nintendo's marketshare to increase by at least 5% each. while not a stunning improvement, it will help to further balence off Sony. There is no doubt that PS3 will be very successful, and I'd automatically assume PS3 will get at least 50% of the market. probably 55-60%, but Nintendo and MS's combined marketshare should be pretty decent. Again, all just my speculation. but I think it's quite reasonable.

 

[Nexiss]

Simple Euclidian Geometry can solve the first.
The second is because they're developing it now, rest assured that as semiconductor companies with it all inhouse (more like nVidia's style) they've been planning towards this process since 2001. This can be seen in their recent, combined, $7-8Billion investment in 65nm fabrication to come online in 2005.

All else being equal, obviously it is superior. Problem is, all else isn't equal, and this is a business. If going with the more advanced process has too many problems and leaves a nice big window of opportunity for your competitor, then no, that is not clearly the best process to choose and you'll want to seriously weigh the potential benefits and risks.
It may not be quite the same in the console space as in the PC space, but it still applies. Going with a more advanced process could lead to all sorts of problems for Sony, but they are clearly willing to take the risk.

Anyway, ATI clearly has engineering talent. Even if they aren't pushing ahead with the lithography as hard as Nvidia, that doesn't mean they couldn't. I really don't see any problem with MS choosing them, and MS, with more information than any of us here undoubtedly, obviously agrees.

 

[Joe DeFuria]

Find me a situation when a closed box targeted IC with 3 years of R&D would be better off with an inferior lithography process?

One where 3 years after R&D begins, it turns out the "superior" lithography process "ain't ready yet."

A simple exercise to (again) illustrate the point.

• According to you 65nm has more advantages, by far outweighing 0.13u or 0.09u in a closed box IC.
• No shipping closed box that I'm aware of currently has 65nm tech.

How can you reconcile those two "truisms"?

Simple Euclidian Geometry can solve the first. The second is because they're developing it now...

No, I said reconcile the two statements. Not tell me how individually each one is true.

In other words, if 65nm is by far more advantageous than 0.13u and 0.09u, how come there are no shipping consoles with such parts?

Never mind...I'll answer: because the first is actually NOT a truism. 65nm doesn't always have more advantages. Time to market can rule in, or rule out a process, so that, by definition, can be more important than more advanced lithography. Choosing an advanced Process CAN turn out to be a hindrance, if it ends up hurting time to market in a way that is more detrimental than a "lower performing, but earlier to market" console based on a less advanced process.

 

[Megadrive1988]

I haven't followed every post of this thread. is it the current understanding that Sony will be on .065 for PS3 and ATI will be on .09 for Xbox2 ?

hopefully ATI will find a fab that can do .065 because Sony will be migrating from .065 to .045 ASAP.

 

[Vince]

Is there room for 2 consoles? If so, then you don't have to BEAT them, you just have to be profitable. In any case, you don't have to use the same busines model as your competitor to compete or beat them

Yes, this is a console forum. It just so happens that ATI and nVidia were competing for console business that is nothing like Sony's model. Because Sony is the leader, doesn't negate the fact that other models exist

While I realise your here for ATI's stake and not Microsoft's - IMHO this is Microsoft's last change to dislodge Sony from the "connected" Livingroom paradigm for quite awhile. After Cell gets adopted as a pseudo-standard across Sony Group/Toshiba it's just going to spread and Microsoft will be in a horrible position to leverage their PC-centric buisness model into the livingroom and pervasive computing world.

Microsoft's model is horrible. The company has the money, we all know why they want with XBox - I can't imagine why they're not going the distance like cthellis touched on. I'll make this prediction, if Cell is commercially sucessfull as a pervasive computing device (eg. same scalable architecture/ISA in TV's, BVD's, PS3, Cameras, Cell Phones, etc) then Microsoft has lost massivly.

So yes, Microsoft has to beat them.

PS. About your Cell info, I'll PM you sometime.

 

[Joe DeFuria]

I haven't followed every post of this thread. is it the current understanding that Sony will be on .065 for PS3 and ATI will be on .09 for Xbox2 ?

Certainly, it is understood that PS3 will be on 0.065.

With ATI...it's anyone's guess right now. One monkey wrench is: it's not clear that ATI is even dictating the fab! MS may have already made the decision of target fab and process.

Side question: what process was the original PS2 chips fabbed in?

 

[Vince]

Never mind...I'll answer: because the first is actually NOT a truism. 65nm doesn't always have more advantages. Time to market can rule in, or rule out a process, so that, by definition, can be more important than more advanced lithography.

Not at all. If 65nm was ready for production today then you bet your ass the likes of Intel, nVidia, SCE, etc would jump on it. Just like they did with 130nm.

Time to market only decides which advanced lithography process will be available and which to shoot for. For example, when desiging the R300, hypothetically, they could have chosen the 130nm libraries instead of the 150nm ones.

Not to mention when your designing under a 5 year umbrella like you stated yourself for a console. You target a process technology (eg. 65nm) early on that your R&D tells you will be capable of being ready by then. Then you design with that in mind and prepare to kick ass.

Choosing an advanced Process CAN turn out to be a hindrance, if it ends up hurting time to market in a way that is more detrimental than a "lower performing, but earlier to market" console based on a less advanced process.

This is a specific case scenario Joe and as such it's irrelevent. You don't plan your vacation for next summer around if your otherwise healthy Aunt may or may not decide to drop dead during that period. You put your ass on the line and go for it.

Chekib Akrout, an IBM VP of the PowerPC division and now STI/Austin said the following: "The American attitude seems to be that 'when you go for the leading edge, you're always going to be close to the edge of the cliff.'"

I think the man's on the path of truth...

 

[Vince]

Side question: what process was the original PS2 chips fabbed in?

Jerk doesn't look at my pictures. Click on this to expand:

<img src=http://pc.watch.impress.co.jp/docs/2003/0421/sony1_03.jpg width=150>

And to preempt any of your parallels about if the PS2's EmotionEngine or Graphic Synthesizer was "bleeding-edge" - look at the sheer size of the ICs. And if you still feel there is even debate here then realize that SCE and Toshiba are more system integrators than a R&amp;D powerhouse like IBM. If SCE/Toshiab pushed the 0.25um process that far (especially the GS and it's eDRAM), with that big of an IC - what do you think STI is going to do? Have you seen the Cell patent?

 

[TheMightyPuck]

Side question: what process was the original PS2 chips fabbed in?

Bingo, we have a winner.

 

[cthellis42]

that doesn't mean they couldn't.

...or aren't...

With the next generation, I expect both MS's and Nintendo's marketshare to increase by at least 5% each. while not a stunning improvement, it will help to further balence off Sony. There is no doubt that PS3 will be very successful, and I'd automatically assume PS3 will get at least 50% of the market. probably 55-60%, but Nintendo and MS's combined marketshare should be pretty decent. Again, all just my speculation. but I think it's quite reasonable.

Actually, I'm kind of expecting this next generation to be "make it or break it" in a lot of ways. Sony is playing a game much bigger than consoles, and if it plays out for them properly, could lead to HUGE things in all sorts of areas by the end of next generation. I'm rather expecting a lot more convergence by the end, and Microsoft seems a lot more prone to merge with a PC ideal, which may not be what developers (and other companies) have in mind. (It may also screw a bit with their "Media Center PC" initiative.) Nintendo I don't think dreams big enough, so I expect to see them focues pretty purely on gaming (with obviously more attention paid to online capabilities and portable interconnectivity) and though I can see them sticking to their word and pretty much always continuing their console tradition, I can just as well see them becoming a niche market where they are mainly a vehicle for their own games.

Sony, meanwhile, (should Cell function as it should and they can tap this well from ~2005-2010) will be creating their own unique meld of a host of games and entertainment and a range of hardware devices--perhaps even computers later on--and quite possibly reshaping the way one thinks about all of it.

Microsoft, should they not gain much headway, I can see shifting the Xbox design to an uber-dedicated gaming PC with some media center purposing, but mainly hoping to draw Xbox developers into the PC market full-strength (where it will no doubt continue to reign supreme) and push more linking that way. Else they will HAVE to break off totally, perhaps snatch Intel, and pull what Sony/IBM/Toshiba are now and REALLY rethink things from a purely console market, because if they don't gain much footing on Sony after two generations they're not likely to succeed in a third. (And on the whole I can't see them doing this.)

...

... ... All subject to change, of course.

 

[Joe DeFuria]

While I realise your here for ATI's stake and not Microsoft's - /quote]

I'm here for myself, thank you.

IMHO this is Microsoft's last change to dislodge Sony from the "connected" Livingroom paradigm for quite awhile. After Cell gets adopted as a pseudo-standard across Sony Group/Toshiba it's just going to spread and Microsoft will be in a horrible position to leverage their PC-centric buisness model into the livingroom and pervasive computing world.

That's quite possible, sure.

It's also possible that like many other "paraigm changing efforts", Cell just doesn't really come to fruition. It ends up being 2 years late due to production and issues and fab issues. It results in a difficult to program for architecture, too high in power consumption, and too expensive to displace "simpler, more fixed funtion" ICs in devices.

Distributed computing turns out to be 20 years ahead of it's time. And at that time X-Box 6 uses what old PS3 hacks call "cell rip-off technology...why can't MS do anything original?"

Microsoft's model is horrible.

Less risky, for sure. History will be the judge on how horrible it is.

 

[Joe DeFuria]

Not at all. If 65nm was ready for production today then you bet your ass the likes of Intel, nVidia, SCE, etc would jump on it. Just like they did with 130nm.

The point is (can't believe you refuse to acknowledge it), 65nm is NOT ready today. Hence: all else is not equal.

Time to market only decides which advanced lithography process will be available and which to shoot for.

Or conversely, sensitivity to time to market can dictate which lithography process you shoot for.

For example, when desiging the R300, hypothetically, they could have chosen the 130nm libraries instead of the 150nm ones.

Right. And ATI likely felt that lower time-to-market risks of 150nm outweighed any other potetial advantages of 130nm. And in that case, they guessed right.

Not to mention when your designing under a 5 year umbrella like you stated yourself for a console.

ATI is not designing in a 5 year umbrella, like I stated.

You target a process technology (eg. 65nm) early on that your R&amp;D tells you will be capable of being ready by then. Then you design with that in mind and prepare to kick ass.

And if 65nm turns out to take a year longer to be able to get the product out in quantity than was anticipated 5 years ago...then what?

The answer is: it depends. Being late to market is only a real issue if your competitor steps up. If he does, you might end up getting screwed. Ifhe doesn't you might be OK.

At the same time R&amp;D is telling you what processes are "obtainable", someone else in management is telling you how flexible they think the ship window is, before you run the risk of getting into "trouble."

Choosing an advanced Process CAN turn out to be a hindrance, if it ends up hurting time to market in a way that is more detrimental than a "lower performing, but earlier to market" console based on a less advanced process.

This is a specific case scenario Joe and as such it's irrelevent.

And it's irrelevant because....you say so? Or because we always know 5 years in advance the state of a particular fab process will be in?

You don't plan your vacation for next summer around if your otherwise healthy Aunt may or may not decide to drop dead during that period. You put your ass on the line and go for it.

Um, if my Aunt DOES drop dead, then guess what? My vacation is screwed. The fact that I put my ass on the line doesn't make me any less screwed. (We'll ignore the fact that my Aunt is really the one that got screwed here. )

Chekib Akrout, an IBM VP of the PowerPC division and now STI/Austin said the following: "The American attitude seems to be that 'when you go for the leading edge, you're always going to be close to the edge of the cliff.'"

I think the man's on the path of truth...

Sure...and I really admire risk-takers. The key is recongizing and evaluating the risks, not just ignoring them. That's what gets you killed.

 

[Vince]

It's also possible that like many other "paraigm changing efforts", Cell just doesn't really come to fruition. It ends up being 2 years late due to production and issues and fab issues. It results in a difficult to program for architecture, too high in power consumption, and too expensive to displace "simpler, more fixed funtion" ICs in devices.

Of course, a very real possibility. But companies won't win by planning on their competitors failing.

 

[Joe DeFuria]

And to preempt any of your parallels about if the PS2's EmotionEngine or Graphic Synthesizer was "bleeding-edge" - look at the sheer size of the ICs.

That is exactly my point in the related side argument...look at the sheer size. Certainly more complex than the Intel and nVidia ICs, right?

And yet, "how much more powerful" is the PS2 than the X-box?

Last I heard, they each had their pros and cons, but are generally considered on par. So the PS2 "architecture" requires the large and complex Ics to compete with the x-box 1 architecture.

(This isn't a knock against either architecture...it's merely an acknowledgement of how different they are.)

If PS3 is on 0.065, and if Xbox2 is on 0.9, that doesn't say much about overall performance / features. The architectures are just so vastly different....

 

[Joe DeFuria]

Of course, a very real possibility. But companies won't win by planning on their competitors failing.

Nor do they win by underestimating their competitors either.