Nvidia GT300 core: Speculation

[no-X]

I'd say...

Eyefinity ~ 3D Vision
Stream ~ CUDA
PhysX ~ DX10.1
????? ~ DX11

 

[stevem]

No you haven't heard even a fraction of anything.

Should've seen my original quip. Much wittier (so I thought).

I'd say...

Eyefinity ~ 3D Vision
Stream ~ CUDA
PhysX ~ DX10.1
????? ~ DX11

I don't think points 2 & 3 are equivalents.

 

[trinibwoy]

Yeah not even close...

 

[no-X]

They are equivalent from the marketing standpoint at least. Majority of users must decide between DX10.1 (which is offered only by ATi now) and PhysX (which offers only nVidia).

 

[Scali]

PhysX - DX11 seems like the obvious next step (and nVidia is slowly closing the DX10.1 gap, with some GPUs now supporting it).

 

[stevem]

PhysX - DX11 seems like the obvious next step

Wowsers! That would be tops on Nvidia's wish-list for Xmas.

 

[Scali]

Wowsers! That would be tops on Nvidia's wish-list for Xmas.

I was referring to No-X's list, which had ???? next to DX11. I think you should just put PhysX there aswell.

 

[KimB]

Nah, because nVidia will also support DX11. I don't think we should consider nVidia's current generation to be able to compete much at all against ATI's next generation.

 

[nyr]

Did it ever sound to you like there were more chances to abandon the idea? I'd think though that they might have gone for a combined solution this time (dedicated units and SP/DP capable ALUs).

Would it be feasible to beef up the MUL and then fuse it with the MAD to do DP?

 

[Scali]

Nah, because nVidia will also support DX11.

Obviously... just not right now.

I don't think we should consider nVidia's current generation to be able to compete much at all against ATI's next generation.

Why not?
The preliminary benchmarks of the 5xxx that I've seen weren't THAT much better than the GTX285. And the price of the 5xxx cards seems to be higher at this point (I saw 319 euros, when you can get a GTX285 from about 260 euros up).

 

[no-X]

PhysX - DX11 seems like the obvious next step (and nVidia is slowly closing the DX10.1 gap, with some GPUs now supporting it).

The advantage of DX10.1 is (mostly) ~20% performance boost. 20% performnce boost is nice, but completely uninteresting for a value part, because moving from 10FPS to 12FPS won't make the game playable.

 

[KimB]

Why not?
The preliminary benchmarks of the 5xxx that I've seen weren't THAT much better than the GTX285. And the price of the 5xxx cards seems to be higher at this point (I saw 319 euros, when you can get a GTX285 from about 260 euros up).

Well, if ATI can't significantly outperform the current generation, then they should be really, really worried about what nVidia will come up with.

However, I don't think that'll be the case. Of course, in the performance and mainstream segments, it might be a little bit until the next generation is competitive. But that will come, and the current generation shouldn't stand a chance when it does.

 

[trinibwoy]

Would it be feasible to beef up the MUL and then fuse it with the MAD to do DP?

The MAD and MUL are completely separate pipelines.

 

[neliz]

Well, if ATI can't significantly outperform the current generation, then they should be really, really worried about what nVidia will come up with.

The upcoming GT21x parts already can't outperform AMD's 55nm parts.
Come next week, nVidia will have knocked about $150 of their top end card to be considered competing.
A Hail Mary GT212 needs to be 4 times as fast as a GTX260 to be crowned performance king.

 

[Slappi]

The upcoming GT21x parts already can't outperform AMD's 55nm parts.
Come next week, nVidia will have knocked about $150 of their top end card to be considered competing.
A Hail Mary GT212 needs to be 4 times as fast as a GTX260 to be crowned performance king.

Whatever you say Chuck!

 

[Scali]

Whatever you say Chuck!

Completely offtopic, but do I sense some Manchester-speak there?

 

[Davros]

what card is 4x a 260 ?

ps: slappi
I didnt know you are from Manchester i'm so sorry you have my deepest sympathy

 

[Blazkowicz]

The upcoming GT21x parts already can't outperform AMD's 55nm parts.

Nor can they outperform NV's 55nm parts.

 

[CarstenS]

Here is interesting article:
http://www.semiaccurate.com/2009/09/15/nvidia-gt300-yeilds-under-2/

I guess goodnight GT300 till next year maybe.....


People who voted http://forum.beyond3d.com/showthread.php?t=49120&page=132 that GT300 will not be here this year, looks correct. ATI R8xx will walk on Nvidia for awhile - if this article is accurate.

Maybe it's just semiaccurate?
*SCNR*
After all, the probable translation Bug that led the known industry expert to take month numbers for yield rates has been discussed a few pages back (or in another thread).

 

[DegustatoR]

The upcoming GT21x parts already can't outperform AMD's 55nm parts.

Which parts are these? Do you know the pricing of these parts already?

Come next week, nVidia will have knocked about $150 of their top end card to be considered competing.

That's true. And I have to agree with Chalnoth here: NV simply has no parts higher than GT21x that can compete with Evergreen line at the moment. But that doesn't mean that this situation will last long.

A Hail Mary GT212 needs to be 4 times as fast as a GTX260 to be crowned performance king.

Why the hell would it need something like that? It should have 5870 level of performance plus a little extra, that's all. And last time I checked 5870 was nowhere near GTX260x4 levels.
If you're talking about Hemlock then I have to remind you that supposable GT212 will have it's own "X2" card without any doubt.
But that's if it exists at all which I still consider highly unlikely.