Shader Models 2 and 4

DaveBaumann said:
No, its short sighted to assume that somehing that makes full use of next gen consoles will equate to SM3.0.
What's the meaning of "short-sighted"? (I'm serious... it's a term so abused that I've forgotten what it actually means!)

I think we should forget about stuff like "full use". It's something that isn't just about reaching the spec limits of a graphics chip since there are other factors.

DaveBaumann said:
This whole thread is full of assumuptions and generalities that bears very little reflection on reality.
In a way, so are theoretical and synthetic benchmarks. Which this site uses :)

This thread, like many others, is about discussing the possibilities of eventualities.
 
I think we should forget about stuff like "full use". It's something that isn't just about reaching the spec limits of a graphics chip since there are other factors.

You don't have to reach the spec limits of a graphics chip, you just have to use some of the capabilities that don't exist elsewhere at the current time - DST Shadowing is an example of this as this is something that is fully exposed on XBox but doesn't exist within the specifications for DirectX 8/9, it just so happens that it operates on NVIDIA's PC chips in the same fashion that it operates on the XBox. Inevitably there will be elements that are exposed on both consoles that are not available in the contemporary DX Shader Models and in the case of Xenon may be closer to WGF2.0 specifications in some areas (and quite probably behind current SM3.0 specs in others).

In a way, so are theoretical and synthetic benchmarks. Which this site uses

Synthetic Benchmarks provide empirical performance data which is generally used to adjudge the capabilities of a board in specific areas in relation to its predecessors or peers! :D
 
It still seems to me you don't throw away work you've already done (sm3) just because you have some features that are outside the spec. Those same features are also outside sm2 --you'll have to solve that problem anyway, and can incorporate the solution in your sm3 path as well, even if that means your "sm3 path" is really a mixture of sm3 and sm2 (for the "outside the spec" features).
 
Geo, and that I just another illustration of the point: its not as simple as "Next Gen consoles are SM3.0, ergo PC games will be SM3.0" as it doesn't take into account a multitude of factors, such as the actual capabilities of the graphics of the consoles coming up, the marketshares of PC boards already out, the waxing an waning importance of PC technologies vs Console technologies due to their vastly differing lifecycles, etc., etc. IF what was being prosed in this thread were the case then we wouldn’t have been running DX9 benchmarks with Tomb Raider 18 months ago and Doom3, Source and UE3 engine wouldn’t have SM2.0 paths in them.
 
If what you just said was this conversation is way premature to reach dispositive conclusions, I'll agree. I would think at least we need to see some high-profile mid-life console games (as you say, early ones will be heavily impacted by the hardware they were designed on) to really start putting holes near the bullseye. But then, of course, we'll *know*, and now we're having fun kicking it around.

Btw, I'm arguing for all three paths, so I'm not sounding a death knell on anybody, or in any way contravening forward-looking partial api support from a title not originally designed for most of its development lifecycle around that api.
 
The majority of all graphic chips in use is still DX7. The best-selling cards are DX7-DX8 (effectively). Longhorn might change that, but it will probably just run on that hardware as well, without the new effects. And the low-end videochips that allow Longhorn to use WGF will probably default to DX8 for games by that time.

If you develop games and use a HLSL, you might be best served by just using the "advanced" features that compile on any DX9+ model, with some optional eye-candy for SM3.0 and up. Three models, DX7, DX8 and DX9, with some additional stuff for DX8.1 and SM3.0, if it is easy and convenient to do so?

YMMV. :D
 
DiGuru said:
The majority of all graphic chips in use is still DX7.
Going by the trusty Steam survey results, which I think are still the best representation of what cards game buyers have, DX9 cards (9500+, 5900+) hold a little over a third of the market and are ahead of both DX7 and DX8 cards. These numbers are also a good 6 months old, so are from a period when 6800 and X800 availability was still very average and the 6200/6600/X700 lines were yet to launch.
 
On first login of a new Steam account you get a prompt saying "Do you want to take part in the Steam Survey?". It's automated (like a dxdiag) but optional. As it also only shows up for new users (or when there's a new survey, as there was a month before HL2 came out) that skews the results a little towards older cards as well.
 
I fully agree with Dave. There's an enormous difference between the feature set / architecture of console chips resembling SM3-capable PC GPUs and actually being SM3-capable PC GPUs (and that's before you consider the huge differences in the overall platforms). My expectation is that SM3 probably won't receive a huge amount of attention from engine builders. I think one further aspect that hasn't been mentioned goes something like this: If you add up all the R3xx, R4xx and NV4x chips you get a whole bunch of GPUs that provide reliable SM2 performance. However, do the NV4x chips qualify as "reliably fast" SM3 products? If you observe the rather painful hit that NV40 takes running HDR lighting effects you might conclude they do not. Or look at it another way - in many respects Doom 3 is the ultimate DX8/sm1 engine (ok, it's much more complicated than that since its openGL not DX and since its leverages some NV-only shadow rendering technilogy). And that didnt appear until well into the life cycle of DX9 hardware. In fact, there were very few games put out that heavily leveraged the features available in DX8/sm1 before DX9/sm2 came along. Perhaps the industry goes incycles, missing out alternate shader models. In that scenario we go from DX7 and no shaders, we largely skip dx8 / sm1 support, see widespread coding for sm2, once again we largely skip sm3 and see widespread support for sm4. That's an over simplification and that's notreally how I see it, but I think there's an element of truth in there.
 
HDR lighting effects are not related to SM3.0. It's just that some hardware that supports SM3.0 happens to be hardware that supports FP16 blending and filtering as well. NV44 does not, and we don't exactly know what R520 and XG45 bring to the table.

I'm pretty optimistic that many console developers don't want to throw away their shaders/content for a PC port, and I'm not sure there's a lot these consoles can do that SM3.0 can't.
 
SM3 is alot nearer what the consoles will be doing, given the completely shit state of the PC business I can see us ending up with 3 bands of PC graphics for a good portion of this (next? to me its this but I guess everybody else stills think its next...) generation of consoles.

1) Crap, PS1.1. Embedded motherboard stuff, Keeps the bean counters happy, but gamers completely forget that its there.

2) PS 3.0 quick ports of the console stuff

3) (Later on its the life cycle) Prestige, SM4 high end PC stuff that gets you IHV deals and keeps the hardcore PC salivating over benchmarks.

Not really any different from now, except its currently Fixed, PS1.1 and PS2.0... Again roughly aligned with the console market.

Of course in actual fact I bet that only type 3 will be true console ports. The consoles can do stuff beyond SM3.0 but reducing a few fancy bits and fitting it all to SM3 will be the quick/cheap port method.

But I'm talking about 2-3 years from now, not sure what the time-lines everybody else is talking about. SM2.0 will be dominant PC until next-gen console is old-hat...
 
Xmas said:
HDR lighting effects are not related to SM3.0.

OK, bad example - but I suspect that NV40 is probably not the bomb for adavnced sm3 rendering. Perhaps a better example would have been branching.
 
Back
Top