NV40 Pixel shader = Checkbox or Useable?

[indio]

Will NV40 ps 3.0 be useable?

 

[davepermen]

of course it's usable. question is, will we see it used before nv40 gets dropped again?

 

[demalion]

Well, the middle two seem the reasonable ones. Where is the box for "I don't know which and I'm not going to claim I know or attack people for thinking the one of those I did not pick"?

 

[Quitch]

I said useful but limited. Assuming it runs at a decent speed, I think the fact that the main competitor doesn't support it will limit how often people use it.... then again, there's still a lot of nVidia developers out there, so I don't think nVidia would have as hard a time as ATI in getting people to pick up on things the other cards can't do.

 

[Guden Oden]

What's the point of voting when we have no info to support any of the available alternatives?

 

[Pete]

I'm with Guden. Are all these polls really nec'y? Don't we get enough chances for baseless speculation in other NV40/R420 rumor threads?

 

[Megadrive1988]

Pixel Shader 3.0 probably won't be even close to useful until NV50/R500, which will have Pixel Shader 4.0, and 4.0 won't be useful until...well you get it

 

[lost]

I would say PS3.0 will be usefull once titles come out supporting VS 3.0. VS3 is much more usefull than 2.0 and IIRC, the output format from VS3 can only be read by PS3. Because of this, maybe PS3 only features such as dynamic branching may not be used thoroughly, but PS3 will be used to compliment VS3.

 

[retsam]

didnt jc say that the first genaration of cards with a curtain feature set is only usefull to developers .... i say add one more to those choices



retsam

 

[Fodder]

didnt jc say that the first genaration of cards with a curtain feature set is only usefull to developers .... i say add one more to those choices

He did, but then R300 showed that a first generation DX9 card could actually do DX9 well.

Anyhow, what with ATI "advising" developers not to use PS3.0, that hints to me that Nv's implementation is half decent.

 

[AlphaWolf]

didnt jc say that the first genaration of cards with a curtain feature set is only usefull to developers .... i say add one more to those choices

He did, but then R300 showed that a first generation DX9 card could actually do DX9 well.

Anyhow, what with ATI "advising" developers not to use PS3.0, that hints to me that Nv's implementation is half decent.

How do you get that out of it?

If Nv's implementation is really poor would ATI suggest differently?

 

[Tim Murray]

You can't get ANYTHING out of it. If NV's implementation is to the R420 what R300 was to NV30, then of course they'll say that. If NV40 is slower than the guy who drives Miss Daisy, who cares? They'll say it anyway just because they don't support it.

 

[gunblade]

didnt jc say that the first genaration of cards with a curtain feature set is only usefull to developers .... i say add one more to those choices

He did, but then R300 showed that a first generation DX9 card could actually do DX9 well.

Anyhow, what with ATI "advising" developers not to use PS3.0, that hints to me that Nv's implementation is half decent.

Well, Ati don't have the feature, so no way would they suggest to use it regardless of whether the feature runs decently or otherwise.

 

[Xmas]

I think there's at least another step between "completely" and "very limited".

 

[indio]

how much glanularity do we need?

 

[no_way]

of course it's usable. question is, will we see it used before nv40 gets dropped again?

Thats a bold assertion. RT-patches on NV20 anyone ? Did displacement mapping ever happen on Parhelia ? Does R300 do truform ?

 

[demalion]

how much glanularity do we need?

Must...resist...testosterone joke...

How much granularity to not exclude what is likely to suitably describe NV40's PS 3.0? 3 gradations that were more general could probably do (within the framework of asking for guesswork), the problem is the extremes not the amount.

 

[indio]

Well the reason I posted the poll is this.
There is a general notion that R420 will not have PS 3.0 functionality
and NV40 will. If the previous is true , will NV40 be functional enough to program for? PS 3.0 will have a limited exposure (nv40 only) Will developers take the time and effort to program for basically a single card?
Maybe I should have polled this - NV40 .... closer to r300 or NV30?

If NV40 is as functional with PS 3.0 as R300 is with PS 2.0 , this could be a boon for Nvidia. If it's more like NV30 it's more smoke and mirrors.
I'm just trying to get a gauge of where opinions are at.

 

[Xmas]

When you program specifically for PS3.0 - i.e. not only compile the same source for both ps_2_0 and ps_3_0 targets, but really use the new possibilities - you can go two ways, one is adding more complex effects, and the other is saving performance. And while it will be a single card in a few weeks, it will be a whole product family later this year (plus maybe cards from other IHVs), and especially for the mainstream parts developers might be willing to go the optimization route.

 

[GrapeApe]

Another thing to consider are those people looking for a longer term hold, basically looking for the next R9700Pro to take them beyond the horizon. If the NV40 can perform well with initial applications that showcase it's use then the NV40 will have that advanatge to people looking for a card they can still use in 2 years time. Does it mean much short term? Of course not, but for those not looking to replace the NV40 with an NV45/50 or the R420 with an R480/500 in 6-12 months they may put a little more weight on PS 3.0 support despite it lack of practical uses in the near future.

That being said if the NV40 is as effective at running PS3.0 instensive applications as the FX5200 is at running any pixel shader intensive app now, then it will mean nothing, especially if that is exposed early on.

That's just my two frames worth.