Predicting the Xbox-2

[Brimstone]

Spong as a source is bizzare. Who runs spong anyway? The same people that run the inquirer?

Having dual R500's would seem like a path to take. A single R500 would result in a console with a much lower performance level than the PS 3 will have if the speculation of Panjev is somewhat correct. Although I could see them ripping apart the R500, and splitting the core up like 3dfx did with the Voodoo or planned to do with the Rampage. If the X-Box 2 had a chipset all split up like that, they could go for the good yields at .9nm and then go for cost reduction at .065nm like you've suggested.

 

[Paul]

Spong isn't a news source it's entertainment, much like those "newspapers" you see while waiting on line to pay for groceries. "Boy grows Third arm"

 

[Paul]

Judging from what R500 is going to become, no way in hell they could put 2 inside an Xbox and expect it not to overheat like hell. Just trust me on this, 1 R500 will be PLENTY. And I mean that, underline plenty.

 

[Josiah]

I don't think they would use the same R500 you'll see in video cards, but rather a modfied version of R500 (or more likely, a modified version of R550)

 

[nonamer]

Anything that goes into the Xbox-"2" will be DX10 compliant at least. Any knowledge about what type of DX will the R500 be compliant?

Chap, not to burst your buble or anything - but that roadmap is wrong. For starters, there is nobody with 110nm lines. IMHO, they will utilize 130nm untill ~2005 when they go 90nm, with the lithography process being the upperbounds on preformance for the IHVs untill then.

I find it hard to believe that they'll wait until 2005 to go to 90nm. There's gonna be a huge glut of fab capacity at 90nm by 2004 AFAIK. There's something like a half a dozen companies or so with their own 300mm fullsize fab plants IIRC.

 

[jvd]

Anything that goes into the Xbox-"2" will be DX10 compliant at least. Any knowledge about what type of DX will the R500 be compliant?

Chap, not to burst your buble or anything - but that roadmap is wrong. For starters, there is nobody with 110nm lines. IMHO, they will utilize 130nm untill ~2005 when they go 90nm, with the lithography process being the upperbounds on preformance for the IHVs untill then.

I find it hard to believe that they'll wait until 2005 to go to 90nm. There's gonna be a huge glut of fab capacity at 90nm by 2004 AFAIK. There's something like a half a dozen companies or so with their own 300mm fullsize fab plants IIRC.

remember they will wanted a known micrion. Or something they have lots of experiance with so they will have enough chips at launch. All they need is an nvidia and have to launch late

 

[jvd]

Judging from what R500 is going to become, no way in hell they could put 2 inside an Xbox and expect it not to overheat like hell. Just trust me on this, 1 R500 will be PLENTY. And I mean that, underline plenty.

I'm not suggesting they run them at crazy speeds . I'm sugesting they use 2 so they don't have to push the clock speed boundrys that far. That way they can lower the voltage of the chip to keep it cooler. I'm sure my lodgic is on track. And running hot . Well i'm sure it will be cooler than whatever nvidia would make.

 

[Vince]

I find it hard to believe that they'll wait until 2005 to go to 90nm. There's gonna be a huge glut of fab capacity at 90nm by 2004 AFAIK. There's something like a half a dozen companies or so with their own 300mm fullsize fab plants IIRC.

Woops, thanks. I confused 90nm with 65nm.. my appologies.

TSMC will be mass producing 90nm in early/mid 2004 IIRC, I think early risk-production is late 2003. Behind Intel/Sony by ~6 months for mass production it would seem.

 

[zurich]

I find it hard to believe that they'll wait until 2005 to go to 90nm. There's gonna be a huge glut of fab capacity at 90nm by 2004 AFAIK. There's something like a half a dozen companies or so with their own 300mm fullsize fab plants IIRC.

Woops, thanks. I confused 90nm with 65nm.. my appologies.

TSMC will be mass producing 90nm in early/mid 2004 IIRC, I think early risk-production is late 2003. Behind Intel/Sony by ~6 months for mass production it would seem.

What about IBM? Wonder if MS wouldn't possibly use them as a 'fail safe' launch measure.

 

[Vince]

What about IBM? Wonder if MS wouldn't possibly use them as a 'fail safe' launch measure.

I don't know - the situation gets very peculiar when IBM is entered into the picture.

I question what the terms of the joint SCE-Sony-IBM-Toshiba alliance were concerning the sharing of the process technology they co-developed with 3rd parties. There must be some clause limiting it's use or what's the point of Sony Group investing the $5Billion or whatever it's upto now along with Toshiba's investment in Oita just so IBM can turn around and fab a Microsoft or Nintendo or Panasonic designed IC.

But, if this is so (as logic would assume) then how does this impact the nVidia-IBM dealings? This, along with the fact that Dave seems keen on posting these rumors have me leaning towards nVidia out of the Next Box. This is all together quite shady.

What do you think Zurich?

 

[jvd]

I find it hard to believe that they'll wait until 2005 to go to 90nm. There's gonna be a huge glut of fab capacity at 90nm by 2004 AFAIK. There's something like a half a dozen companies or so with their own 300mm fullsize fab plants IIRC.

Woops, thanks. I confused 90nm with 65nm.. my appologies.

TSMC will be mass producing 90nm in early/mid 2004 IIRC, I think early risk-production is late 2003. Behind Intel/Sony by ~6 months for mass production it would seem.

vince they are still having trouble with .13 .... so who knows if .9 will be ready in 2004

 

[zurich]

I don't know, IBM is a bit of a dark horse IMO.

On one hand, I don't know if TSMC can step up to the plate and be competitive against Sony/Toshiba's fab process in the context of a next gen system (in yields, .13 really shook their credability, not to mention the economic problems the island must be facing from the SARS hoopla which im sure are affecting their semiconductor industry), but on the other, IBM is more than able - it just depends on you said, about the terms of their partnership with Sony/Toshiba.

MS/NV is looking like an IBM deal, while MS/ATI would be a TSMC one. I don't know if either IHV would be happy with Intel fabbing their parts, as they are, in essence, a competitor of sorts (integrated graphics market). It'd be nice to bring Intel into the mix, but I don't think that would happen.

Whatever does end up happening, I have a feeling that the costs MS incurred on the Xbox will seem like punk change compared to what they'll have to spend on the Xbox2 in order to be competitive.

But such is the price of their off-the-shelf model..

 

[Dave Baumann]

On one hand, I don't know if TSMC can step up to the plate and be competitive against Sony/Toshiba's fab process in the context of a next gen system (in yields, .13 really shook their credability, not to mention the economic problems the island must be facing from the SARS hoopla which im sure are affecting their semiconductor industry), but on the other, IBM is more than able - it just depends on you said, about the terms of their partnership with Sony/Toshiba.

TSMC's business depends on it, of course they will sort it out. Just because there was one highly publisised incident with NVIAI's 130nm migration (despite warnd from TSMC to NVIDIA and ATI that what NVIDIA were trying wasn't ready) doesn't mean they are second rate as far as customer fabs go.

 

[Dave Baumann]

Having dual R500's would seem like a path to take. A single R500 would result in a console with a much lower performance level than the PS 3 will have if the speculation of Panjev is somewhat correct.

Oh good god. So now we all know the perfprmance and capabilites of parts that haven't even been talked about yet?

 

[archie4oz]

Oh good god. So now we all know the perfprmance and capabilites of parts that haven't even been talked about yet?

Hehe, I'm surprised you didn't notice that it's become the norm around here...

 

[Paul]

I am certain that an Xbox2 with a R500 could compete with PS3 on a visual basis(What you see on screen) But PS3 would probably beat it in raw specs. It would take a while before ps3 would really show it's true stuff.

I'm sugesting they use 2 so they don't have to push the clock speed boundrys that far. That way they can lower the voltage of the chip to keep it cooler.

Even then, your talking about 2 GPU's with 1Ghz clocks and memory speeds at 1Ghz.

Radeon2(R500) is currently penciled in to be 2Ghz core and memory speeds. Hence the 2.

 

[Vince]

Oh good god. So now we all know the perfprmance and capabilites of parts that haven't even been talked about yet?

Hehe, I'm surprised you didn't notice that it's become the norm around here...

"Preformance," "Capability"?!.... <shrugs> Everyone knows that two is better that one... seesh.

 

[Brimstone]

Having dual R500's would seem like a path to take. A single R500 would result in a console with a much lower performance level than the PS 3 will have if the speculation of Panjev is somewhat correct.

Oh good god. So now we all know the perfprmance and capabilites of parts that haven't even been talked about yet?

No...but it makes for more incendiary discussions.

 

[zurich]

Having dual R500's would seem like a path to take. A single R500 would result in a console with a much lower performance level than the PS 3 will have if the speculation of Panjev is somewhat correct.

Oh good god. So now we all know the perfprmance and capabilites of parts that haven't even been talked about yet?

Yea, must be something in the air around here :?

 

[zurich]

TSMC's business depends on it, of course they will sort it out. Just because there was one highly publisised incident with NVIAI's 130nm migration (despite warnd from TSMC to NVIDIA and ATI that what NVIDIA were trying wasn't ready) doesn't mean they are second rate as far as customer fabs go.

It may be their business, but they're still trailing IBM, Sony/Tosh, and Intel (more or less). Sony/Tosh is the one to catch. The last things MS wants to do is either have 1) highly overclocked dustbuster part with super low yields, but effectively setting a high performance bar for the systems life span, 2) have a much lower clocked part that can supply good yields from the get go (ie: launch). High perf high yields are ofcourse, that much better