no 6800ultra Extreme, still the same 6800ultra

1. FP frame buffer blending/texture filtering: much easier/more robust use of high dynamic range for developers. I expect more games to support HDR for SM3 than for SM2.
2. Vertex textures: required for displacement mapping. Can be used to great effect for nearly-flat animated surfaces (such as water).
3. Pixel Shader 3.0: performance improvements in future games.

And then for me, I plan to try to do some programming. With the impressive FP32 processing performance of the video card, I think I may be able to use it for some computational work.

 

Well, if you have had 3 hard drives die on you, you may want to go ahead and replace the PSU anyway....

 

HL2 uses HDR and Valve have no plans on incorporating SM3 atm. SM2 can run HDR, SM3 might just be slightly faster.

Again, is possible with SM2.

Dynamic branching can actually be detrimental to performance on shaders that don't have an extensive amount of instructions. I doubt there will be any games in the next year or year and a half that will benefit from it. And by the time games do start to use it the NV40 will be incapable of running them at decent speeds.

If the X800 is faster then the 6800 that defeats the purpose of SM3 altogether.

 

:rofl:

I'm sorry Chal, but please don't try and tell me that one is original to ya....it sounds like someone else I know, BBut I can't quite recall who.

I guess it all DePends on what you plan to compute with it.

 

It has nothing to do with speed in this case. In order to support HDR on pre-SM3 hardware, you have to use hacks and workarounds. With SM3, you can implement HDR by simply using a FP16 framebuffer and the appropriate lighting and/or textures.

Not really. Only with CPU emulation.

It's not just dynamic branching. The facing register, for example, will allow for fewer passes on some shaders. The Predicate register is another way of branching that is more robust than the compares offered previously, but should have similar performance characteristics.

 

I want to try matrix diagonalization. If it's to be useful, I need to do it on large matrices, which means that precision is a big issue. I'm not even sure yet if FP32 can be enough, even with the most numerically-stable technique.

 

I understand SM3 makes it slightly easier for developers but I said "benefits the end consumer".

It is possible. ATI has already demonstrated this.

All serving to increase performance in only miniscule amounts, if at all.

 

Easier on developers=more developers will add support.

Possible? Yes. Useful? No. If it were useful, games would have supported it for some time.

No. Only certain parts of SM3 are primarily related to performance, and the performance differences for some algorithms can be huge (i.e. orders of magnitude). Besides, with attitudes like this, it's a wonder technology advances at all.

 

So ps2.0 is useless also? I feel so cheated.

 

Um, I was talking about displacement mapping. PS2 has been used in games, to some extent, for a while now. Displacement mapping isn't useful unless the hardware can interpret the displacement map (i.e. support vertex textures)

 

Doesn't the R420 have this?

 

Considering that it's not an option in PS 2.x, I'd say no.

 

Chalnoth, if I get a 6800 will you be my girlfriend? Just to prove em wrong, of course 8)





someone had to...

 

What does being easier on developers have to do with benefits to the consumer? So what if there is more support? SM3 still holds no advantages over SM2 for the end user.

How do you define useful? It may take slightly longer to implement using SM2 but the end result is the same.

No, SM3 is primarily performance related through more efficient means of handling the code. The only other change, besides making it easier for the developer, is the increase in the maximum allowable instructions per shader which translates to completely useless. Unless of course Microsoft planned to stop development on direct-x completely, which isn't the case. And if the performance differences were 'orders of magnitude' don't you think nvidia would have demonstrated this by now? The only demo they've shown using SM3 was Far Cry which was used to show off displacement mapping, not performance enhancments.

 

SM 3.0 is important for one reason and one reason only, Nvidia did it first.

Period, end of Sentence.

I personally think it's a load of crap when Nvidia says the jump from SM2.0 to SM3.0 is so important that it has to be done immediately and the Nvidiots parrot it mindlessly. The jump from PS1.x to SM2.0 was much bigger, but that wasn't important, since Nvidia went around telling everyone to replace PS2.0 shaders with PS1.1 ones so FX cards could run it faster.

I understand why Nvidia is saying it, but I have serious problems with the hypocrisy of people who bought FX cards and are now saying SM3.0 is soooo important. None of the other improvements on PS1.3 seem to have mattered, but this one is important, just because Nvidia PR says it is.

Nvidia has been pushing SM3.0 for a bit now, but nobody has ever been able to give one real reason why it is so important to get SM3.0 takeup immediately, when SM2.0 takeup and PS1.4 takeup weren't so important.

 

I know what you were talking about. It's called sarcasm.

Just because a feature is possible it doesn't mean it will run at useful speeds on lowest common denominator hardware. Many of the ps2.0 features, become unusable on FX cards because of the performance hit. Higher performance may give developers the ability to use features that weren't used in the past.

How fast sm3.0 runs on nvidia's $200 and under offerings (whenever they release them) and how well received they are by consumers will give a pretty good idea of the level of sm3.0 adoption we can expect from developers in the short term.

 

Because it can be done now.

I mean, it's really simple. ATI held back the technology of their product in order to maintain performance leadership. I see this as underhanded and just plain wrong. ATI is holding back game technology in an attempt to hold onto marketshare.

Besides, I don't see where anybody said that SM2 wasn't important. It was a very significant step up from the previous.

 

For anyone slightly confused the part I've bolded above pertains to vertex displacement mapping. See quote below for more context

No, it isn't possible on ATi hardware. Either the CPU must be used in which case it is a form of software rendering, not hardware acceleration, or the effect is approximated in the pixel shaders, which is irrelevant since we are discussing vertex displacement mapping & textures.

 

ATI is not holding technology back, nvidia is jumping ahead. We saw the same thing happen with the NV30, only it didn't turn out the way they had planned. ATI researched the cost to benefit ratio of supporting SM3 and came to the conclusion that it wasn't worth implementing. SM2 has been out for two years now and games are only starting to use it, why would the situation be any different with SM3? On top of that SM3 is hardly a big jump, certainly nowhere near the likes of SM1.3 to 2.0. The advantages of SM3 are marginal and considering ATI would have had to increase their transistor count 33% it would have also effected cost and yields, as we are seeing with the 6800. SM3 is just an interim build for DXNext; aka PS1.4 all over again.

 

Technically a two pass mechanism can be used whereby the pixel shader will lookup the displacement values from the displacement map texture and then passed back to the vertex shader for the actual displacment.