Nvidia gtx 475??

[jonnyboyC]

so got bored and looked at the 4xx series cards, just to waste some time and noticed on that they had a gtx 475 on the list saying it was based on the gf 104, does anyone have any clue what this card will entail?

 

[jonnyboyC]

well i would have guessed that based on the name but does any one have a guess on cores ,clocks, memory and the like, TDP

 

[4745454b]

Wait, how do they pull this off? Without using a die shrink seeing as none is ready, how does Nvidia not only bump the frequency up, but enable all the shaders? I can see them bumping the clocks up a bit, cherry picking the chips, etc. But enable all 512 shaders? What changed that allowed them to do this?

 

[jonnyboyC]

i kinda figured there wasn't any real data on it, i just saw on wikipedia, and there was a 475 with a bunch of ??? under it,

but as for the 512, it's not really rocket science, as psychosaysdie said, yields are up, i'm sure even from the get go they were getting fully functional 512 cuda core chips, but, they were in such short supply they couldn't get enough to market, and they were already months behind anyways

 

[4745454b]

Yields might be going up but it has to be something else. Yields going up should mean that they don't have to junk as many 480s into 470 cores. The other issue you have is the 300W limit. They rate the GTX480 at 250W, but its more then that. By adding in those extra cores AND raising the clocks, how do they stay under 300W?

I can understand if they added more layers to the chip so they could bump the clocks. But I can't see how they increase core and frequency and stay under 300W. What I'm wondering is did they add a metal layer? Change the transistors? Some half node come out that I didn't hear about?

 

[TheGreatGrapeApe]

What changed that allowed them to do this?

spin Spin RE-SPIN, Baby!

And the 485 is still said to not be coming in under the 300W spec, so it likely won't be destined for OEMs.

And you should know better than that, there is no half node from 40, 40 is the half-node, and there's nothing between here and the 32/28nm until Xmas.

 

[4745454b]

So again, anyone know whats in the respin? I for one would like to know whats under the new hood. I'd probably only understand 5-10% of it, but I feel it helps me learn.

 

[TheGreatGrapeApe]

Until it surfaces, it's hard to tell what the difference is.

It's unlikely a new metal layer, but it could be and would be the easiest to figure out by nomenclature alone.

You'll likely have to wait for launch for that bit, and then for a birdie to tell one of the dynamic trio (Charlie, Fuad, Theo) otherwise.

 

[jyjjy]

I will be interesting to see what they do with the GTX 475. They are enabling all the shaders and upping the speed for the 485 but with the architecture built on 32 shader blocks they either need to leave the number of shaders the same for the GTX 475 or add 32 and give it the same number as the 480. If they leave it the same and bump up the speeds then it is just like a factory OCed 470 which is meh. If they enable 32 more it will be basically an underclocked 480 which could be a great value card for overclockers if they really only raise the price by $20-30, especially if the respin allows for higher top speeds.

 

[4745454b]

You mean GTX475 right? The GTX75 is a completely different chip. You raise a good point assuming my sleep deprived mind followed correctly. If the GTX475 is based off the new GTX485, then you could have a chip thats the GTX480. If prices only go up ~$40, then you could have the "GTX480" for around $400. Still to expensive for me, but its progress.

 

[jyjjy]

lol
Yes, I meant the the GTX 475(fixed it with an edit.)
Ironically you typo-ed in questioning me and said GTX75

 

[4745454b]

LOL. Now I see it. Think I'll leave it as a tribute to bad naming schemes.