G70 here we come

[loekf2]

Tried to find a topic, but didn't see anything ... yet.

An article on computerbase.de:

http://www.computerbase.de/news/har...dia/2005/maerz/cebit05_nvidia_g70_ende_april/

is speculating on the announcement of the G70 (= successor of the NV4x) end of April. Apart from the obvious features, which I expect just to be refresh of the NV4x architecture (more pipes, higher clock speeds, faster and more memory), they are also speculating on the manufacturing process.

Hinting towards 0.13 um (0.11 um) at TSMC makes no sense for me. The NV4x is already HUGE in the current process, so I'm pretty sure Nvidia will also go the 90 nm route.

Further, releasing a product in 2005 for WFG 2.0 also makes no sense to me. I will just be DX9.0c compatible, shader model 3.0 (still).

 

[CJ]

I used the search for G70 and found several. Here's the one which also contains the link you posted (last page):

http://www.beyond3d.com/forum/viewtopic.php?t=20712&postdays=0&postorder=asc&start=0

Guess you didn't search hard enough.

 

[JD]

Don't forget, there is an opengl api which can expose hw features of an ihv card so theoretically the g70 could very well go beyond what d3d 9c offers pretty much like it was done with nv opengl register combiners.

 

[Dave Baumann]

I think it will be WGF, but there are two versions of WGF...

 

[ondaedg]

oh do share Dave.

 

[RingWraith]

Further, releasing a product in 2005 for WFG 2.0 also makes no sense to me. I will just be DX9.0c compatible, shader model 3.0 (still).

Still? We have yet to see a game that actually fully uses model 3.0. ATi has shown their choice to wait was a prudent one.

 

[Xenus]

So, Dave does this mean you've finally got some decent moles in Nvidia.

 

[Killer-Kris]

oh do share Dave.

What you probably missed was his comment about two versions of WGF. So G70 could (and will) be WGF compliant in the same sense that NV30, R300, and the newest Intel GPU all are.

 

[KimB]

So, Dave does this mean you've finally got some decent moles in Nvidia.

Haha, no. WGF 1.0 is basically DirectX 9. So he's basically saying nothing at all.

 

[Geo]

So, Dave does this mean you've finally got some decent moles in Nvidia.

You’re wrong actually – none of the [ATI] engineers will proffer pre-release information, if you care to notice they don’t post that frequently anymore (more posts are coming from behind another firewall at the moment!).

Doesn't *necessarily* mean they are whispering in his ear too, of course. . .

 

[Khronus]

Have MS even finalized WGF 2.0 yet?

 

[Titanio]

Have MS even finalized WGF 2.0 yet?

I believe not, or at least such information is not public.

 

[Megadrive1988]

An article on computerbase.de:

http://www.computerbase.de/news/har...dia/2005/maerz/cebit05_nvidia_g70_ende_april/

is speculating on the announcement of the G70 (= successor of the NV4x)

to use Jen-Hsun favorite phrase "my sense is that" the G70 seems to be an NV40 refresh, that is, a new highend NV4x part, and *not* the NV50/NV5x (or NV60) or whatever Nvidia's next-gen architecture is.

G70 is probably a SM3.0+ NV4x GPU designed to respond to ATI's forthcoming R520 Fudo.

G70 might be to NV40|GeForce 6800 what the NV25|GeForce4ti was to NV20|GeForce3

Nvidia could easily name it 'GeForce 7' even without it being a whole new architecture. as they have done several times in the past. I am not saying this is what Nvidia will do, just that it would not surprise me, given Nvidia's history.

I don't expect to see Nvidia's new architecture (NV50, NV5x or NV60) until 2006. the PS3 GPU will also be based on this new architecture. but G70 almost certainly cannot be totally new, it almost has to be a another NV4x.

 

[Geo]

to use Jen-Hsun favorite phrase "my sense is" that, G70 seems to be an NV40 refresh, that is, a new highend NV4x part, and *not* the NV50/NV5x (or NV60) or whatever Nvidia's next-gen architecture is.

G70 is probably a SM3.0+ NV4x GPU designed to respond to ATI's forthcoming R520 Fudo.

G70 might be to NV40|GeForce 6800 what the NV25|GeForce4ti was to NV20|GeForce3

Nvidia could easily name it 'GeForce 7' even without it being a whole new architecture. as they have done several times in the past. I am not saying this is what Nvidia will do, just that it would not surprise me, given Nvidia's history.

I don't expect to see Nvidia's new architecture (NV50, NV5x or NV60) until 2006. the PS3 GPU will also be based on this new architecture. but G70 almost certainly cannot be totally new, it almost has to be a another NV4x.

My spidey-sense is still tingling and has been for weeks. How in the world is a non-low-k 110nm part not going to get its clock cleaned by a similarly-featured low-k 90nm part? You could point at R300 vs NV30, but I would object on not "similarly featured" grounds.

[Not that my spidey-sense is known to be all that reliable in the graphics world. Nevertheless.]

 

[KimB]

I highly doubt that there will be any terribly significant process differences between ATI and nVidia's high-end parts for some time to come.

 

[Entropy]

I highly doubt that there will be any terribly significant process differences between ATI and nVidia's high-end parts for some time to come.

Given that they use the same fabs, that assumption makes perfect sense. If anything, nVidias willingness to go to IBM allows them access to more sophisticated lithographic processes if they really feel the need. Still, there may be short term differences, but they shouldn't matter much in terms of price/performance.

 

[Ailuros]

I highly doubt that there will be any terribly significant process differences between ATI and nVidia's high-end parts for some time to come.

Judging from Michael Hara's comments quoted on B3D's frontpage it sounds rather to me that NV is about 6 months behind when it comes to 90nm in general.

How in the world is a non-low-k 110nm part not going to get its clock cleaned by a similarly-featured low-k 90nm part?

More quads for 110nm maybe? With 6 quads @400MHz you get the same fill-rate as with 4 quads @600MHz for example.

Have MS even finalized WGF 2.0 yet?

It would be a disaster for the IHVs if that wouldn't be the case for quite some time now IMHO. At least Microsoft and IHVs are aware of the specs.

 

[KimB]

Judging from Michael Hara's comments quoted on B3D's frontpage it sounds rather to me that NV is about 6 months behind when it comes to 90nm in general.

The problem with this statement is that neither nVidia nor ATI have anything direct to do with the manufacturing.

 

[Dave Baumann]

The IHV's make different choices at different times and they also have varying sucesses with those choices - its not as though R350 was 130nm nor NV40 130nm low-k.

 

[Joe DeFuria]

Judging from Michael Hara's comments quoted on B3D's frontpage it sounds rather to me that NV is about 6 months behind when it comes to 90nm in general.

The problem with this statement is that neither nVidia nor ATI have anything direct to do with the manufacturing.

But there is a finite manufacturing capability, particularly with new, large chips. So each company looks at its road-map options, engineering capacity, and works with the manufacturers to determine when to "hit" new processes.

I don't recall any recent history of ATI and nVidia both moving to a brand new major process at the same time. One usually (typically, it was nVidia in the past with 0.15-->0.13) hits it one product cycle earlier.