PowerVR Series 6?

[Ailuros]

PVR-5

TBDR

24 PIPES

800MHZ

512MB

1 BILLION POLYGONS PER SECOND

300-350 MILLION TRANSISTORS

[/DREAM]

ROFL

It's been said before that effective theoretical fill-rates can be highly misleading and I'm only pointing it out because it's the probable origin of this dumb speculation.

If PowerVR would need such a configuration to compete with current high end sollutions, then it would actually kill the very purpose of using a fundamentally different architecture in the first place. Look back to the Dreamcast specs, it might help.

 

[SegaR&D]

If they used these specs plus TBDR, then the competition wouldn't have a comparable part for atleast 18-24months

 

[Ailuros]

The fact that the cost would equal that of a small island is probably irrelevant.

In any case with all the quiecy surrounding PowerVR, it's just one more senseless conditional.

 

[SegaR&D]

Probably around the $500 mark for a 2005 release.

 

[Ty]

No, no no. By then I'm sure they could price it for $250 easy. Remember this is TBDR so it's really really REALLY cheap but ultra-super-sophisticated-powerful.

 

[ZoinKs!]

No, no no. By then I'm sure they could price it for $250 easy. Remember this is TBDR so it's really really REALLY cheap but ultra-super-sophisticated-powerful.

Which is exactly why every manufacturer has been in such a hurry to embrace it.

 

[kemosabe]

Well according to this morning's National Post here in Canada, it seems at least one market analyst thinks ATI should buy them out:

"Graphics chip manufacturer ATI Technologies Inc. appears to be flush with cash these days, holding about US$400-million more than it needs to run its operations. The company could use the money to repurchase shares or make an acquisition. Gus Papageorgiou, an analyst at Scotia Capital, thinks ATI should expand into mobile phone processor applications. In particular, the analyst believes that ATI should buy Imagination Technologies Group PLC, which trades on the London Stock Exchange."

Release the rumour mill hounds.

 

[Ailuros]

Probably around the $500 mark for a 2005 release.

Sure anything you say. You obviously aren't refering to a TBDR and you obviously haven't a clue what the essentials of S5 could look like. That config of yours would starve abysmally from bandwidth, even it being theoretically a DR. Plus 6 quads @ 800MHz equals an oven even for 2005. Dream on.

kemosabe,

That would be one expensive venture to buy out IMG's success in the PDA/mobile market and it's not that ATI couldn't yield a better foothold in that particular market would they want to w/o IMG. All they need is a more competitive design than Imageon and start selling IP for that particular market, like NV did with AR10.

 

[SegaR&D]

I think Sega Sammy should buy a stake in the company

 

[SegaR&D]

Probably around the $500 mark for a 2005 release.

Sure anything you say. You obviously aren't refering to a TBDR and you obviously haven't a clue what the essentials of S5 could look like. That config of yours would starve abysmally from bandwidth, even it being theoretically a DR. Plus 6 quads @ 800MHz equals an oven even for 2005. Dream on.

What about 8x1 at 800Mhz - 1Ghz?

Would that work?

 

[arjan de lumens]

What about 8x1 at 800Mhz - 1Ghz?

Would that work?

The main problem with very hgh clock speed GPUs is that the power consumption - per drawn pixel - drawn rises nearly proportionally to the clock speed targeted for the design, due to the increased number of pipeline registers every pixel has to ripple through. An 8x1 at 1 GHz may be smaller than a 16x1 at 500MHz, but it will draw much more power.

 

[SegaR&D]

How much more?

Would it be reasonable for a PC?

 

[arjan de lumens]

How much more?

Would it be reasonable for a PC?

About 40-70% more, as all clock-related power consumption would double. Also, you would need to transport the resulting heat away from a reduced die area, requiring an overall power supply and cooling solution substantially larger and more expensive than, say, that of the Geforce6800 Ultra.