"Inevitable Bleak Outcome for nVidia's Cuda + Physx Strategy"

[Keysplayr]

Originally posted by: evolucion8
im confused. when u guys talked about pixel shading for color sampling. is the pixel shader unit who does that job? i thought that it was performed by a fixed function thing like the ROPs. GF3/GF4 had very limited pixel shading functionality compared to Radeon 8500

ROP's are where all the data that is processed by the GPU is finally organized and values are "mixed" together, then sent to your monitor.

 

[evolucion8]

Originally posted by: Keysplayr
ROP's are where all the data that is processed by the GPU is finally organized and values are "mixed" together, then sent to your monitor.

didnt the hd2xxx series had those broken and thats why they were slow at fsaa. or it was their tmu?

 

[Keysplayr]

Originally posted by: evolucion8

Originally posted by: Keysplayr
ROP's are where all the data that is processed by the GPU is finally organized and values are "mixed" together, then sent to your monitor.

didnt the hd2xxx series had those broken and thats why they were slow at fsaa. or it was their tmu?

I think the 2xxx thru 3xxx series attempted AA via the shaders, and not dedicated hardware for AA. Something like that. Didn't do so hot. Remedied in the 4xxx series.

 

[filmguy1200]

Hmm....

 

[Melted Rabbit]

I hate to say this but, after reading this thread and watching it play out as it has, Scali's responses in a debate could be described as: "You can bring a man to information, but you can't make him think."

I thought that these three posts, by Scali, from the same thread on Beyond3d, were particularly egregious:


http://forum.beyond3d.com/show...?p=1233064&postcount=2

Scali:
Well, I've always had this theory that ATi's instructionset wasn't very efficient, and graphics were actually a pretty good case, many GPGPU tasks would be worse. Perhaps that's what we're seeing now, nVidia's scalar threading approach paying off.

Scali, you didn't have any evidence of that then, nor is there any evidence of that conclusion has any basis in reality.

http://forum.beyond3d.com/show...p=1233179&postcount=16

Scali:

Originally Posted by Jawed:
Did you read the PDF already? Jawed

You have to realize that I don't believe ANYTHING that AMD/ATi says anymore, given their past. I'll just wait and see what happens, if anything happens at all.

So, it seems that attempting debate with you is futile if the the posts on both beyond3d and here, on AT in this thread. Scali, you don't give up an inch, even other posters show deficiencies in your claims. You even seem to totally ignore the other side's evidence, this shows you are not much more than a troll. You don't come here to debate, just to flame, attract attention, and inflate your e-peen. Other non-technical forums I visit would have banned you long ago. Note this is me trying to actually be somewhat constructive as to how not to behave in civilized society and not intended as a personal attack. A personal attack would have been the statement: "You are worse than Hitler!", which you would really have to work hard at to become.

 

[BFG10K]

Originally posted by: Scali

I thought we had agreed that multisample could mean any kind of special case supersampling,

Sure, if be ?we? you mean ?Scali and no-one else?.

Only because they fudged the pattern so it still works.

Seriously, what is this babbling claptrap you keep spewing? Do you even read quotes before responding?

You repeatedly stated the texture sample wasn?t in the center and said B3D ?backed? you, despite the fact that they actually stated the exact opposite, and thus disproved you.

Your response to this was only because they fudged the pattern so it still works.

Seriously, is this debate in English, or in another language? Because your reply appears to be a random construction of words that are about as relevant as the price of fish. Am I arguing with a bot, or with some kind of sentence creation software?

Again, since words fail me, I?ll just quote one of your past responses again:

Originally posted by: Scali

It seems deliberate (I give you the benefit of the doubt that you're not really THAT thick. Sadly that means that I think you are trolling and being obnoxious on purpose).

Yep, your quote is coming in really handy. It?s amazing how self-prophetic it?s turned out to be.

Wasn't the key to MSAA that colour and z are decoupled?

They're geometry samples yes, but that's the problem... They aren't decoupled.

Do you even know what that term means in this context?

They aren't here, that's why you can only make it work with these fudged patterns, where you always need to sample in the pixel center.

Seriously, just stop it. You?re simply embarrassing yourself each time you post now, and revealing how fraudulent your arguments are.

The problem is with the Quincunx diagram. Samples 1 and 2 would have the correct texel, but 3, 4 and 5 don't (they are samples taken from adjacent pixels).

No, the problem is with you not acknowledging the difference between 2x and 2xQ.

Another problem is that the samples aren't evenly distributed. With one depth sample at the center and one at the top left corner, your average position is halfway between them. But your colour sample is based on the center.
On GF4 they corrected that by having the samples evenly distributed around the center of the pixel, as it should be.

And? The positions of the geometry samples is not a requirement to fit the definition of MSAA.

Oh please. I've proven that I know perfectly well how MSAA works in OpenGL and Direct3D. My understanding of MSAA was never in doubt and you know it.

Ahahahahaha! What a joke. And I didn?t even need to buy a ticket to chortle at this sideshow.

Decoupled from the POSITION obviously, Mr. self-proclaimed MSAA-expert.

I?m not an expert, but out of the two of us, I?m the one that understands the basic fundamentals of how MSAA works. You don?t.

You just keep pulling new ?theories? out of an orifice every time someone hands you a new scrap of information about how this stuff works, and then try to pretend you ?knew? all along how it worked, even as you interpret the new information comically wrong.

Furthermore, as you create these bullshit theories, you claim a ?deep? level of understanding no-one else has, and then claim the burden of proof is somehow on us to disprove the steaming pile you?ve created.

Again, an open forum doesn?t work like that. A personal blog with you as the sole registered user (with all responses banned) might be a more suitable home for your ?debating? skills.

Obviously patterns with a sample in the center are generally sub-optimal. Which nVidia realized and improved with the GeForce4.

Right, but that has nothing to do with whether it?s classified as MSAA or not. This is more irrelevant rhetoric on your part.

But why did they need to fix that in the GeForce4 if according to you GeForce3 already did full MSAA as we know it,

Yeah, why indeed did ATI need to fix the 2xxx by changing the sample positions when the R3xx already did full MSAA as we know it. :roll:

meaning full freedom in the sampling pattern

Uh, no. If that?s what it meant, then not a single part would be doing true MSAA using your ?definition?. But I suspect you won?t address my response properly, and will likely respond with more random sentences.

 

[Scali]

Give it up, Melted Rabbit.
I never ignore evidence. Case in point: I mistakenly said that BioShock used PhysX. Someone provided a link that showed that it used Havok. I admitted I was wrong right away.

The problem is that there IS no evidence in this discussion. There's just some people saying I'm wrong and insulting me. But when I ask them to prove their point, they fail to deliver.
That's now how a discussion works.

There's also nothing wrong with the posts you are presenting here, because as you can see, nobody on Beyond3D made a big deal out of them.
And as you know from your posts in the "25 chips" thread, if they don't agree with what you say, they'll correct you. Which by the way was a nice bit of trolling on your behalf. Obviously I haven't taken you seriously since.

 

[Scali]

Originally posted by: BFG10K
You repeatedly stated the texture sample wasn?t in the center and said B3D ?backed? you, despite the fact that they actually stated the exact opposite, and thus disproved you.

According to Beyond3D, GeForce3 samples at the first sample position by definition.
That is not the same as sampling at the pixel center by definition.
Hence the position of the first sample has to BE at the pixel center if you want it to sample at the pixel center.
Which as I already said, has some problems.

Originally posted by: BFG10K
I?m not an expert, but out of the two of us, I?m the one that understands the basic fundamentals of how MSAA works. You don?t.

If I didn't understand how MSAA works, I wouldn't be questioning whether GF3/4 do MSAA, would I?
Again, the issue is about what GF3/4 do, MSAA itself was never in question.

Then again, since you prefer insults over actual discussion, I think we're done here. You seem mostly frustrated that you can't prove your point.
That's the fun part. I know you may have a point, and I don't really care about GF3/4.
I must admit that I forgot that GF4 improved on the AA techniques of GF3. I thought they both did the same.
But it's cool to see you totally lose it because you fail to make a solid case
You know as well as I do that I know what I'm talking about, and my theories aren't just 'bullshit'. They're poking holes into your theories and claims in a very annoying way. They keep bringing up that there is some information lacking here.
See, you fell into a trap a few times. By claiming that GF3 does "MSAA as we know it today", you went wrong.

I have to admit I'm a bit disappointed in you. I was waiting for you to draw the colour-compression card, but you never did.
That's actually a significant difference between GF3/4 and R300, and affects performance in a rather direct way.
I wonder though... why didn't nVidia put it in the GF3/4? They already had z-buffer compression, the concept is very similar, and when you're doing MSAA, it's a great optimization.

 

[BFG10K]

Originally posted by: Scali

According to Beyond3D, GeForce3 samples at the first sample position by definition.
That is not the same as sampling at the pixel center by definition.

They both sample at the center which is what I said. You then claimed they didn?t and that B3D backed your claim, which they didn?t, and you're now playing semantic games without retracting your false interpretation of the article.

Which as I already said, has some problems.

It has a suboptimal grid, but that in no way stops it from being classed as MSAA. If it did, only stochastic SGMS could be classed as ?real MSAA? using that reasoning, and not a single part in consumer space does that.

Again, the issue is about what GF3/4 do, MSAA itself was never in question.

This is a blatant lie. Here?s your original quote that started this:

Originally posted by: Scali

ATi was the first with MSAA on the Radeon 9500/9700 series. nVidia only had the GeForce4, which only supported SSAA, which was too slow to actually use.

Wrong on both counts; nice to see you?ll stoop to lying to cover your tracks.

But it's cool to see you totally lose it because you fail to make a solid case

If someone kept repeatedly claiming that 1+1 = 3 while you showed them numerous evidence to the contrary, you too would have difficulty maintaining a straight face, especially if this person claimed to have ?deep? knowledge others don?t have.

 

[Scali]

Originally posted by: BFG10K
This is a blatant lie. Here?s your original quote that started this:

It started as a careless remark, I wasn't deliberately lying. When you jumped all over it, I quickly admitted that nVidia does do some kind of MSAA, but not quite the same as what the R300 did.
So don't accuse me of lying when I corrected myself immediately.

Originally posted by: BFG10K
If someone kept repeatedly claiming that 1+1 = 3 while you showed them numerous evidence to the contrary

The evidence is all circumstantial

Originally posted by: BFG10K
you too would have difficulty maintaining a straight face, especially if this person claimed to have ?deep? knowledge others don?t have.

Well, I think you do know that I know what I'm talking about.
I think you also know that even though you may know a thing or two about MSAA (but you didn't start about colour compression...?), that certainly doesn't prove you're on the same level as I am.
So if you really have the need to measure yourself against me, you'll have to prove yourself on many other subjects aswell.

 

[devione]

I've been following this article since it's first post.

And while I don't really have anything to add to the argument, I have to say that for an individual nominated as a mod to get so worked up over a (assuming he really is) troll is honestly pathetic.

BFG10K, if he really is trolling, why feed him? Leave it alone already. For all the "reporting to the senior mods" threats you've been pumping out in the previous pages, and so far no action has been taken, maybe he isn't trolling as suggested? Maybe he does have a point?

By no means I am defending anyone here, but for someone holding a modship to present themselves in this way, it's just unacceptable.

 

[SirPauly]

No offense , but when the topic was about Physx, well, the debate was refreshing, but when this has turned into the "Scali" debate --- no offense -- is it possible to move forward from this and try to stay on topic?