[H]'s take on SM3 pros/cons

Sanctusx2

Newcomer
I thought this was a great read. It's the most unbiased report on the whole SM3 better than SM2 debate that I've seen. They discuss the new effects, how it will affect gamers directly, and what this technology brings to the table. All without succumbing to the temptation of being heavily biased for or against the NV40. Not too much new that you can't get by sifting through forums, but it's nice to see it all laid out clearly and concisely.

edit: oops, article is on SM3 not just PS3, so changed words accordingly. ^^
 
The whole comparison is based on screenshots and a misunderstanding of what effects were being done (offset mapping, not displacement mapping)

I don't think the article was very informative.
 
from the Hardocp article...

"Going back and reading exactly what we asked NVIDIA, "What exact PS 3.0 features are (Crytek) using?" They responded, "As stated earlier, displacement mapping is used for the walls and stone textures like the Buddha." Funny enough, as noted earlier, Displacement Mapping is a Vertex Shader feature not a Pixel Shader feature. So all in all, I would have to believe what Tamasi is quoted as saying in the TR interview as true."
--------------------------------------------------------------
from the Tr interview....

"TR: We've seen the Far Cry screenshots you all released with Shader Model 3.0 effects.

Tamasi: Actually, those are Shader 2 or Shader 3. That's right.

TR: One of the effects we're seeing is a "pseudo displacement mapping" effect, isn't it?

Tamasi: Yeah. "Virtual displacement mapping," "parallax mapping," there's been a number of terms for that.

TR: Any idea how many instructions long the shader program is that produces this effect?

Tamasi: That effect actually is reasonably inexpensive from a number of... I think it's less than ten for that one particular piece of that effect. It's actually less than ten shader instructions to do that.

TR: Will we see a Shader Model 2.0 path for GeForce FX with this same effect in Far Cry?

Tamasi: Yeah, the images that you've seen from Far Cry, the current path, those are actually Shader Model 2.0, and anything that runs Shader Model 2.0 should be able to produce those images. "
---------------------------------------------------------------
Does that mean that those Ps3.0 screenies shown by Nvidia are actually Ps2.0?????

Seems like it to me.
 
As much as I like to blame [H] for anything...I can't blame them for that misunderstanding...

[H said:
]Tech Report posted an interview with Tony Tamasi of NVIDIA and they comment on the FarCry SM3.0 screenshots and they make some statements that might lead you to beleive that no real SM3.0 operations are being done in the "PS3.0 Screenshots" from NVIDIA. We are a bit familiar with Virtual Displacement Mapping and Parrallx Mapping, but we were not lead to believe by our NVIDIA contacts that is what is going on in the screenshots from NVIDIA. And quite frankly it is not really that important. This seems to simply further our thoughts that SM3.0 is not bringing much in terms of Image Quality to the table currently that cannot be done in SM2.0 and that is what will be important to the gamers buying the video cards.

Going back and reading exactly what we asked NVIDIA, "What exact PS 3.0 features are (Crytek) using?" They responded, "As stated earlier, displacement mapping is used for the walls and stone textures like the Buddha." Funny enough, as noted earlier, Displacement Mapping is a Vertex Shader feature not a Pixel Shader feature. So all in all, I would have to believe what Tamasi is quoted as saying in the TR interview as true.

It seems that once again getting the exact truth out of NVIDIA can be a painstakingly complex exercise.

So while they should've known that displacement mapping wasn't the feature in use...you should really be saying that of nVidia PR...
 
Go back and read the original launch party show report I posted here. Yes, they are not PS3.0 only, they are using plain 'ol offset mapping, go the Humus's demo site if you want to see it in action on current generation cards.
 
DemoCoder said:
Go back and read the original launch party show report I posted here. Yes, they are not PS3.0 only, they are using plain 'ol offset mapping, go the Humus's demo site if you want to see it in action on current generation cards.

Yes, but that's not what nV told [H], apparently.
 
Crytek gave all the information needed on Launch day.

It's not surprising to get different information out of different people at a large company, especially if you talk to non-technical people, or people who are not immediately involved in the subject at hand and are passing on third party info. Even Tomasi doesn't know enough about offset mapping and how many instructions it requires.


No doubt some of the people at NV are feeding incorrect info, but that doesn't excuse [H] from writing a shoddy article on SM3.0. He needs to get someone more technical to help him, instead of firing off questions to PR people.

As an example: Look at B3D's original 3dfx-sponsored FSAA article. That's a well written article in AA. I expect articles on SM3.0 to include real analysis, even if the general public can't understand it as easy. Trying to base an article about shader models on a few screenshots is lame.
 
It's obviously an attempt to correct the misapprehension that has been propogated using this "here's a rubbish picture without SM3.0, and here's a really great picture using SM.3.0" during the NV40 launch. It explains for the layman that as far as consumers are concerned, with regards to visuals, SM2.0 gets you a lot of the same benefits that SM3.0 does when compared to no use of SM at all.
 
Bouncing Zabaglione Bros. said:
"here's a rubbish picture without SM3.0, and here's a really great picture using SM.3.0" during the NV40 launch.

We know the "before" is Ps1.1 not Ps2.0.

It the "after" picture really Ps2.0 and not even really Ps3.0???
 
Having played FarCry myself for the past few weeks, I wonder what graphics settings those supposed PS 1.1 shots were taken under.

I sure haven't see anything quite as crappy on my GF4Ti.

... off to test lowest settings ...
 
Moose said:
Bouncing Zabaglione Bros. said:
"here's a rubbish picture without SM3.0, and here's a really great picture using SM.3.0" during the NV40 launch.

We know the "before" is Ps1.1 not Ps2.0.

It the "after" picture really Ps2.0 and not even really Ps3.0???

We do now, but there was strong implication at the time that this was a PS2.0 vs PS3.0 comparison, when it was nothing of the sort. The pictures were even labelled that way at one of the Euro launches, and some websites picked this up and propogated the incorrect information.
 
Looks like they should have been labeled Ps1.1/Ps2.0.

I guess you won't sell many cards based on Ps3.0 support if you label them that way huh. ;)
 
DemoCoder said:
No doubt some of the people at NV are feeding incorrect info, but that doesn't excuse [H] from writing a shoddy article on SM3.0. He needs to get someone more technical to help him, instead of firing off questions to PR people.

Well, not everyone can be B3D. At least they made an update to their article when they found it to be in error. That's pretty much what I would expect.

And to be blunt, it's one thing to expect that PR folks are going to spin and put thiings in a positive light. It's another to make the step and assume that whatever they're telling you is flat-out wrong.

Trying to base an article about shader models on a few screenshots is lame.

Agreed...but nVidia trying to make their case with screenshots is equally as lame.

The bottom line is...if nVidia is having such a hard time showing the advantages of PS 3.0 vs. 2.0....well, that speaks volumes right there IMO.
 
I can't even beleive they are PS1.1. I haven't seen water look that bad in years.

Plus the statue, walls are all low detail. It looks like somebody turned it on to low deatil with postive LOD bias and bilinear filtering.

A massive massive PR act of FUD.
 
incurable said:
I sure haven't see anything quite as crappy on my GF4Ti.

... off to test lowest settings ...
Ok, testing confirmed that those shots look similar to the output of my GF4 with all graphics options in-game turned to low.

Whether it actually still uses any kind of shaders under those settings I don't know (I'll try to test it, though), but labeling the shots PS 1.1 and implying that this is the best DX8 hardware can do IQ-wise is just plain wrong. :rolleyes:

cu

incurable
 
Joe DeFuria said:
The bottom line is...if nVidia is having such a hard time showing the advantages of PS 3.0 vs. 2.0....well, that speaks volumes right there IMO.

Even Nvidia themselves admit that the major thing with SM3.0 is to make it easier on developers (Tony Tamasi interview):

There's a lot of things like that that are possible with Shader Model 3, but frankly, I think the biggest win for Shader Model 3, and from what you've read from developers or if you've talked to them you'll hear pretty much the same thing, is that Shader Model 3 fundamentally just makes it easier on developers. As far as I can tell, that's the biggest win for everyone, because it gives them a real programming model that they're used to.They don't have to worry about instruction set limits and what I call "coding inside out." They can just kind of write their shaders and not have to worry about, "Gee, is this 96 instructions?" or whatnot. And frankly, the feature set is complete enough that they can just kind of code away and get the effect that they want. And frankly, it can be completed simpler and easier in Shader Model 3, so from a productivity perspective, they're going to be much happier.
 
Especially while ignoring that r300s of about 1.5 years old run it just fine with the full ps2.0 settings.
 
This stuff is really getting old. Here I am trying to put up a good preview/coverage of the NVIDIA NV40 event, and I have had to re-write many sections dealing with the new features and especially SM 3.0 support. Of course we left the event feeling very good about the future of 3D, and how well things will look with SM 3.0. Then I started to hear that SM 3.0 doesn't really advance the overall look and feel of 3D rendering all that much, but rather it can provide some real performance enhancements. Now I am getting the distinct impression that while there are some performance advantages to SM 3.0, it is not the brave new world that we were all led to believe. Yes, it adds loops and branches, but it is starting to look like the performance penalties for many of these operations makes it undesirable in many situations. Sure, Nalu has a single shader for her skin/scales, and the overall effect is very impressive, but not all situations are going to reflect that kind of shader use. As the Far Cry example is showing, most everything in the SM 3.0 scene can be done in SM 2.0. Perhaps the most popular feature will be the geometry instancing, and it will probably get a lot of use in RTS games the features lots of units. After that, perhaps vertex texture reads? I can't really say for sure.

Basically, I am really unhappy now with NVIDIA marketing for making my job so damn tough to do, and yet try to remain respectable in the media field. Its damn hard to get a straight answer from anybody, mainly because there are so many strings connecting everyone together that nobody wants to tread on their main partner's toes!

At this moment, if I were to try to sum things up, I think it would be: SM 2.0 will be the defacto standard for some time, and while SM 3.0 offers some nice advancements, I feel that it will not be widely accepted until significantly faster hardware is developed to take advantage of its more complex features and procedures. Would this be "mainly" true?

Edit: a "tense" problem
 
Me said:
The way DX is going, it looks like DX10 might be the first possible version where we start to approach the "graceful performance degredation" model.

Simon F jogged my memory, and even that is likely wrong.

http://www.beyond3d.com/forum/viewtopic.php?t=8959

One of the main "points" made with the DX10 model is that while shading instructions / coding is supposed to be "limitless", there will be definite indications / guidelines as to at what point you can expect to go from "real-time" to "off-line" type performance.

Coders will continue to have to be aware of these limits, and it can still vary from architecture to architecture...AFAIK.
 
Back
Top