[KonKorT]

It looks like Nvidia will launch a new Geforce 400 card named GTX 460 at Computex (1st June). The card will based on Nvidia's GF100 chip, but has got in comparison with GTX 470 and GTX 480 less SPs, TMUs, ROPs etc.
The card will feature 384 SPs, 48 TMUs, 32 ROPs and 256 Bit GDDR5.

GTX 460 with 384 SPs comes at 1st June?

Share|

[pogsnet]

These are models with more defective SPs and cannot clock well for GTX 480 or 470 version so they reduce them. Well maybe a good thing for the price, I hope on performance as well.

[watzup_ken]

Harvesting on the failed chips for their GTX480 and 470. But still, the cost is very high for Nvidia to offer competitive price at this level.

[junjie433]

Originally Posted by watzup_ken

Harvesting on the failed chips for their GTX480 and 470. But still, the cost is very high for Nvidia to offer competitive price at this level.

Nvidia lost its edge in price and quantity 6months ago since R5850 and R5870.

[Neo_XF]

Originally Posted by junjie433

Nvidia lost its edge in price and quantity 6months ago since R5850 and R5870.

Nvidia never had an edge in that, last generation. But they where most certainly "adequate".

Anyone remember how people where criticising the "monolithic" approach of GT200? And how they wheren't as flexible as ATI as a resault... etc etc...

[ORiN]

Originally Posted by bugga321

Need a upgrade in the coming weeks, this card sounds good to me, good value for money i feel, thanks for the news

You might want to look at ATI's offerings instead.

[power666]

I really don't think that this card will be worth it if a GF104 based chip can be launched with high enough shader clocks. nVidia would be better off using these Fermi chips with 384 enabled shaders for Quadro and Tesla lines.

[NextGen_Gamer]

Originally Posted by power666

I really don't think that this card will be worth it if a GF104 based chip can be launched with high enough shader clocks. nVidia would be better off using these Fermi chips with 384 enabled shaders for Quadro and Tesla lines.

I know you and I discussed this same issue in another virtually identical thread, but something else occurred to me: maybe GF104 is not as far along as we believe it to be. Rumors say that GF104 is taped-out, doing fine, and should be introduced in late Q2 2010 (as in June), but hey, some of those same rumors said that GF100 would come out in November 2009. It's entirely possible that either a) GF104 is not nearly as finished as we think, or b) GF104 is having just as many problems with yields and clockspeeds as GF100 did. Regardless of which of those scenarios is happening, the truth might be that GF104 is more on track for a late Q3 2010 introduction rather than a late Q2 one. And if that is the case, it kinda does make sense for NVIDIA to introduce a stop-gap solution in the form of yet another crippled GF100 chip.

They could even replace the GF100-based GeForce GTX 460 with a GF104-based one later on, kinda like the transition from the G80-based GeForce 8800 GTS (320MB/640MB favors) to the G92-based GeForce 8800 GTS (512MB one).

[josiahsuarez]

GTX460 is just another step on the path to a full top to bottom DX11 line like ATI has. it's questionable if a GF104 with 256 shaders will clock high enough to be comparable to a 384 shader part.

[power666]

Originally Posted by NextGen_Gamer

I know you and I discussed this same issue in another virtually identical thread, but something else occurred to me: maybe GF104 is not as far along as we believe it to be. Rumors say that GF104 is taped-out, doing fine, and should be introduced in late Q2 2010 (as in June), but hey, some of those same rumors said that GF100 would come out in November 2009. It's entirely possible that either a) GF104 is not nearly as finished as we think, or b) GF104 is having just as many problems with yields and clockspeeds as GF100 did.

Indeed, we did cover this before so a full rehash isn't necessary. However, I propose scenario a few additional scenarios.

c) the GF104 isn't performing on par with their expectations and a reduced GF100 would fill a hole in the line up. This can be considered an extension of b) you listed above with regards to clock speeds.

d) nVidia plans on releasing a chip containing 384 shaders down the road. The reduced GF100 would be a stop gap to that chip. I consider this scenario very unlikely as a native 384 shader part would likely be six months away at best. If anything the 384 shader part would be the successor to the GF104 but manufactured on 28 nm process. That'd be in 2011.

e) (Which you mention) nVidia will replace the GF100 part with a high clocked GF104 part later down the road.

Originally Posted by NextGen_Gamer

Regardless of which of those scenarios is happening, the truth might be that GF104 is more on track for a late Q3 2010 introduction rather than a late Q2 one. And if that is the case, it kinda does make sense for NVIDIA to introduce a stop-gap solution in the form of yet another crippled GF100 chip.

This part I figure is more variable upon how many Fermi parts they've made and how many Tesla/Quadro parts they hope to sell. I'm still of the mind that it'd be wiser to utilize the 384 shader parts for the workstation/GPGPU crowd since they'll be able to support more memory and ECC. Then again, I'm of the belief that nVidia isn't making that many Fermi parts to begin with. Perhaps this assumption is wrong and the number of wafers and yields is enough to make a 384 consumer part worth it.

I will also mention that the numerous yield numbers could all be correct. The 2% figures could be for a GF100 with full 512 shaders functional. Factoring in 480 and 448 shader parts would bring that number up drastically. 384 shaders parts would further help this number. The GTX 460 could also be the result of bins not have enough TMU's, ROP's or memory controllers functional too.