there is no NV47, but G70 is on the way

pixel pipelines are faster than before (more work per clock cycle), ROPs are faster too!
 
I remember GF4 being much more efficient than GF3 just with some tweaks here and there ;)
 
DegustatoR said:
nAo said:
G70 is not a NV40 with more pipelines, there's more than that.
Not much more actually :) Basically it's still NV4x in it's architecture with some tweaks here and there.

If those tweaks as you call them get rid of quite sizeable bottlenecks overall performance can increase a lot more under certain conditions, besides the usual +X% estimate due to the added quads.
 
neliz said:
g70=nv47

Well, the G70Ultra is supposed to have 2x256bit memory bus, right?
Now what would that imply? I think it's a child of "sli2" where two processors are able to share the same graphics memory, unlike current amr or sli setups.

This would open the doors to the long rumoured dual processors GPU's

While I think the idea of a dual core GPU, even with "SLI2 on chip" is ridiculous, I guess there could be indeed in the near future a "SLI2" or "SLI3" or whatever which would be a NUMA, with a coherent FlexIO link between the two GPUs.
Maybe textures would still be fully duplicated (maybe not), but at least the "renders to textures" would be somewhat sync'ed so there's no game anymore where SLI is useless. (and whatever stuff would be done for better usage of memory bandwith and capacity)

someone could even make a bad ass dual RSX card for the PC? :)
 
Blazkowicz_ said:
someone could even make a bad ass dual RSX card for the PC? :)

Only if someone actually made a RSX processor..
By the time rsx hits the street.. there is more badass stuff available no pc level.. and you wouldn't want a non wgf2.0 compatible card in late 2006 now would you?
 
I'm not so sure WGF 2.0 cards will be available before then. If they are, it'll be the first generation of them.
 
Chalnoth said:
I'm not so sure WGF 2.0 cards will be available before then. If they are, it'll be the first generation of them.

For me, a european it should be a close thing. with M$ rescheduling longhorns launchdate back to summer 2006 and the PS3 not being introduced untill late 2006 (december in europe ) It won't be a certain bet on who's first anyway.
 
DaveBaumann said:
IMO, there will be WGF2.0 capable boards significantly before WGF2.0 is available.

That is understandable; after all DX9 boards were out before DX9 was released too ;) Hardware is evolving in a more rapid pace than the software, so the companies have to make their products sound "future-proof".
 
DaveBaumann said:
IMO, there will be WGF2.0 capable boards significantly before WGF2.0 is available.
Well, it won't be before Spring of next year, at the absolute soonest. Availability probably won't be really good until the Fall (if recent history is any indication). But either way we'll still be in the first generation of WGF2.0 products, so I don't believe that most enthusiasts will consider WGF2.0 to be a must-have by the end of next year. I'd give it another year before that happens.
 
DaveBaumann said:
IMO, there will be WGF2.0 capable boards significantly before WGF2.0 is available.

That's a bit of a shift for you, isn't it? Altho our previous discusion a couple months ago on this was aimed more R600-ish, so perhaps you have "someone else" in mind this time. . .
 
neliz said:
trinibwoy said:
Megadrive1988 said:
so according to Inquirier (they do manage to get SOME things right) the G70 is the Nv47 as most people have obviously linked together as being the same thing.

But wouldn't that mean that RSX is also based on NV47 ?

Either the G70 or G80, with the time it takes to develop the PS3, I think it will be based on the G80, so a dual core GPU yes..

And, for those that are just to ignorant, referring to a chipset's code number, when it starts with C, it doesn't automagically mean it's nvidia..

RSX is probably not G80. but more like custom G70 with some things removed and other elements added. maybe some aspects of G80 are in RSX but that doesnt make RSX a G80 or a dual-core GPU
 
nAo said:
G70 is not a NV40 with more pipelines, there's more than that.

exactly. the G70 is not just NV40 with 2 more vertex shaders and 8 more pipelines. I'd expect G70 to have some fairly significant tweaks, a few changes, a few additions, etc.

the G70 can still probably be concidered an NV4x architecture though, and not NV5x


will G80 be an NV5x then? I guess we'll have to wait some time to find out.

It doesnt look like Nvidia will be showing NV5x GPUs this year. that does not mean that RSX does not have some elements of NV5x. we just don't know yet.

my understanding based on the limited information we have about Nvidia's near-term roadmap:

G70 = either NV47 or NV48

RSX = a significantly modified & customized G70

G70 and RSX are not a complete generational architectural change from NV40, and thus are NV4x architectures, with RSX possibily having some elements of NV5x (much like NV2A having a few elements of NV30 but that did not make NV2A an NV3x GPU).

G80 is a big question mark as to what it is.

Nvidia is probably, or might be saving NV5x for Longhorn, WGF 2.0, Vertex Shader 4.0, Pixel Shader 4.0 DirectX 10 aka DirectX Next which will be a huge inflection point for a completely new architecture. I don't expect to see GPUs that fully support this until fall 2006 since Longhorn and everything it will bring probably wont be out until then, if not somewhat later.


G80 might possibily be an early NV5x chip - that is, if G80 does not = NV48
 
Vertex Shader 4.0, Pixel Shader 4.0

Do they even have a version number for Shaders in WGF2.0?

------------------------------------------------------------------------------

As for the topic if one or both IHVs are going to release WGF2.0 GPUs ~mid 2006 after all, then I wonder if there actually would be a reason for a G7x@90nm and R580 after all. Hmmm..... :rolleyes:
 
Ailuros said:
Vertex Shader 4.0, Pixel Shader 4.0

Do they even have a version number for Shaders in WGF2.0?

------------------------------------------------------------------------------

As for the topic if one or both IHVs are going to release WGF2.0 GPUs ~mid 2006 after all, then I wonder if there actually would be a reason for a G7x@90nm and R580 after all. Hmmm..... :rolleyes:

unified shaders probably has nothing to do with sm3 or 4.

Product refreshes are there to lower manufacturing costs since we won't see the g70 and r520 magically dissapear one r600 and.. g90? appear.
 
DaveBaumann said:
IMO, there will be WGF2.0 capable boards significantly before WGF2.0 is available.

I heard that microsoft is already planing for a situation were there are only WGF 1.0 hardware is available.

If I am take in account that we still don't have an offial WGF 2.0 darft this situation is IMHO highly possible. We can not compare WGF 2.0 with DX/WGF 1.0 were every IHV can select they features they want support. WGF 2.0 don't have cpas anymore. If you don't support a feature you are out of the game. If you support features that don't make it in the WGF 2.0 spec you have waste DIE size. You can still use them with OpenGL but this API have only limiteded relevance in the game bussines today. Maybe this will change but IMHO not in the timeframe of the first WGF 2.0 chips.
 
From the sounds of it Longhorn will ship with WGF1.0 (DX9), but I suspect that there will be a situation like there was with DX9 and both IHV's will have hardware available before WGF2.0 is released.
 
DaveBaumann said:
From the sounds of it Longhorn will ship with WGF1.0 (DX9), but I suspect that there will be a situation like there was with DX9 and both IHV's will have hardware available before WGF2.0 is released.

Now I am understand. You think that MS will release longhorn without WGF 2.0 and we have to wait for a servicepack (like WinFS)?

In this case I agree that it is possible that we will see offical hardware before offical API support.
 
Back
Top