Nvidia GT300 core: Speculation
- Thread starter Shtal
- Start date
- Status
Maybe I should quote you from a few weeks back.
My layman's personal opinion has obviously nothing to do with professional design decisions.
not quiet sure what you mean by its dead jim.
Maybe it would have more function if it was placed on a board? Surely, nvidia wouldn't want you to spoon a functioning GPU off of a board.
When was the last time they put A1 silicon on a board? (***edit: at least that's what the link states I can only see a black nothing there...)
Yeah, if you're talking about the photo then I have some severe doubts about it being G300... It looks too much like the photo I saw a year ago 8)
ah, the stupid chinese.
http://www.overclockers.ru/images/news/2008/08/25/gt200_01.jpg
http://www.overclockers.ru/images/news/2008/08/25/gt200_01.jpg
but then, there's not too much difference in size anyway.
Yeah probably not. But your enthusiasm for negative Nvidia news is amusing
Who cares about the die sizes anyway as long as its not much bigger than GT200b and the complexity twice as high?
Yeah probably not. But your enthusiasm for negative Nvidia news is amusing
I have studied Chinese to make sure I root out the smallest mistakes by Jen-Hsun's squad.
Then remove the chopsticks out of your nose; one is not supposed to use them that way *snicker*
I think one interpretation of Nvidia insistence that DX11 graphics will not be the prime mover of new graphics cards (this year) is that (unlike ATI) they will only have one full DX11 part this year - GT300 on the high end (not counting SP-disabled top end parts that are one rung down on the performance ladder as a separate part.)
They will want to compete against the 57xx with their new G200 respin parts, and those are not full DX11 compliant (although I bet a little bit of market haze on just how much they 'support' DX11 is to be expected.)
Hence they will insist that CUDA / DirectCompute (which will run in 4.x guise on both last gen parts and the rumored GT21x parts) are more of a motivation to buy video cards. After all, the $200 and under parts probably account for a lot more of the market than the $250+ market, so winning the top end of the market alone, while taking a real beating in the performance mid-range will not be good enough for Nvidia.
I think they may have been blind-sided by the 3 way and 6 way output cards from ATI - that's a real market differentiator that they may not have seen coming in time.
They will want to compete against the 57xx with their new G200 respin parts, and those are not full DX11 compliant (although I bet a little bit of market haze on just how much they 'support' DX11 is to be expected.)
Hence they will insist that CUDA / DirectCompute (which will run in 4.x guise on both last gen parts and the rumored GT21x parts) are more of a motivation to buy video cards. After all, the $200 and under parts probably account for a lot more of the market than the $250+ market, so winning the top end of the market alone, while taking a real beating in the performance mid-range will not be good enough for Nvidia.
I think they may have been blind-sided by the 3 way and 6 way output cards from ATI - that's a real market differentiator that they may not have seen coming in time.
I don't think Nvidia has anything to worry about from Eyefinity. But I think you're spot on with the G200 respins. Those are going to be DX10.1 parts going up against AMD's DX11 lineup, and from early signs they aren't very fast either (at least the entry level stuff). So it looks like AMD will be having a very merry christmas this year.
Makes sense. Very possible what they are attempting to do.
Sorry to disappoint you in your enthusiasm for G(T)300 being DOA. The ASIC shown is most definitely a G80 since number an placement of the SMD-parts on the package do exactly match.
Granted, the picture was chosen quite cleverly, since rumored die-size for G(T)300 is about the same as G80 and there's not exactly an abundance of naked G80-shots floating around to compare it with.
auntie edit says:
Here are some pics of a naked G80-chip; a GT200 looks like this
Last edited by a moderator:
Can those G200-based GPUs really stand against 57xx?
How would CUDA help move hardware, Physx? Hopefully that's not the only reason.
- Status