So In Terms of Polygon Performance The X800 Beats Xenos

[3roxor]

I think normal maps will be the big thing . Those e3 videos from 2004 have the epic guys talking about 200 million polygon buildings and backrounds being made into 2 million polygon normal maps and keeping almost the smae quality . Character models wont benfit as much as you can still see blocky body parts however with a higher polygon count and normal maps that should be reduced. This generation will be more about shaders imo

I think for both the rsx and xenos this will reduce the need for insane polygon counts

But then why are those PGR3 devs pimpin with their poly count..

 

[Hardknock]

I think normal maps will be the big thing . Those e3 videos from 2004 have the epic guys talking about 200 million polygon buildings and backrounds being made into 2 million polygon normal maps and keeping almost the smae quality . Character models wont benfit as much as you can still see blocky body parts however with a higher polygon count and normal maps that should be reduced. This generation will be more about shaders imo

I think for both the rsx and xenos this will reduce the need for insane polygon counts

Yep I remember this aswell. And it turned out that the enrionments and characters they were showing, were actually Gears of War assets

 

[Hardknock]

Here's a link folks: http://www.unrealtechnology.com/html/technology/ue30.shtml

And it shows how toward the end how a 100 Million poly scene was compressed into a 500k triangle scene

Here's a video of it in action:

http://www.fileplanet.com/140731/140000/fileinfo/Unreal-E3-2004-Engine-Demonstration

 

[jvd]

thanks for posting the link can't find the videos on ign anymore

 

[jvd]

But then why are those PGR3 devs pimpin with their poly count..

because gamers are still on the polygon count kick. You can get better results from normal maps than increasing the polygon count . You can get the best results by doing both . The problem you run into is cost . Normal maps will cost alot more than just upping polygon counts . Of course i rather take a 500k model with normal maps made from a 10 million polygon model than a 3 million polygon model .

Same as i'd rather take a 100million polygon model of a building and use normal maps and get the actual polygon counts down to 500k than to use up say 10 million polygons with no normal maps .

 

[Arun]

Here you're assuming that bumpmapping (AKA Normal Mapping) comes free, VS-wise. That's not exactly true; if you do it in tangent-space, you're going to have to transform the eye vector (if using specular) and the light vector (for all lights, even with only diffuse lighting) using the appropriate rotation matrix.

While the bottleneck of this operation will most likely remain on the Pixel Shader (on conventional architectures), you shouldn't think it's completely free VS-wise, so on closed architectures I wouldn't expect a ridiculous advantage for unified shading, but I'd love to be proved here.

Of course, the performance hit on the VS is hardly "an order of magnitude", and it will completely depend on what other VS operations have to be done for the same shading program, but most implementations don't have it as a "free" addition, VS-wise. Doesn't mean it's not more than worth it, though!


Uttar

 

[Titanio]

You doubt or you hope?

Both.

I doubt it, because the suggestion seems to be it's tied to clockrate. So as long as RSX was setting up one tri per clock, it wouldn't be any lower than Xenos's figure.

I hope it's not any lower, because like most normal technology enthusiasts, the idea of newer hardware being more powerful/capable appeals. I think it's better to hold that kind of hope, than to, for example, hope that nothing could ever outperform Xenos until MS gets round to releasing a new console again. Eh, AlphaSpartan?

That said, I don't think the setup rate is particularly important, as long as it is "enough".

 

[Alpha_Spartan]

What kind of replies are thoses exactly?

If you think someone is making an incorrect statement, post your justification, and make your own point from that. Posting stuff like "Prove it!" and leave it at that is considered as being "noise" and therefore can (and will) be pruned.

Don't be so hostile. There was nothing wrong with me asking him to prove it. It's funny how some here get a pass while everyone else has to provide proof. The burden of proof doesn't lie with me because I didn't make the statement. Asking for proof does not imply that I didn't believe what he said or that what he said was wrong but rather that I want proof as to how he reached that conclusion. In fact, I lack the information to conclude that he's wrong. I simply asked him to provide that proof. Somehow you took offense.

People should contribute in a positive manner to any discussions, by either addressing a point or by bringing another point (On the topic).

I'm sorry you took asking for proof as something negative. I guess I made the mistake of thinking that people have to prove their statements.

To elaborate, I want to know how he came to the conclusion that the 800 million vertices per second is the G70's polygon set-up rate and not comparable to the X800's 700 million vertices per second. Again, this doesn't imply that he's wrong and that I'm right, but I want to know how he arrived at that conclusion.

 

[Farid]

Don't be so hostile. There was nothing wrong with me asking him to prove it.

There was nothing wrong with Mintmaster, or Hardknock asking the poster to elaborate further on his statements. There's definitely something wrong with the way you asked.

It's funny how some here get a pass while everyone else has to provide proof.

Yeah, it was all a setup against you and the truth...

To elaborate, I want to know how he came to the conclusion that the 800 million vertices per second is the G70's polygon set-up rate and not comparable to the X800's 700 million vertices per second. Again, this doesn't imply that he's wrong and that I'm right, but I want to know how he arrived at that conclusion.

Now, you're posting the way everyone is supposed to. In an elaborated manner.

Nobody have a problem with you or with what you asked, it was, I repeat, simply the manner that you asked that was problematic. If everybody start posting one liners to contradict somebody's point, the forum will turn, fast, into a bashfest, where no discussions can take place... And we're all trying to avoid that.

I hope you understand what the moderation team is trying to do.
Don't take thoses remarks as hostile remarks from us, they're simply reminders.
A lot of people do also post on other forums, principaly gaming forums, where hostile behaviors are common. That's why we remind people that B3D have a different policy when it comes to posting behavior.

Now enough with this, let's discuss the topic! Topic, the!

 

[Almasy]

Where's your proof?

I don´t have a link, but I do remember that information being posted here, IIRC marconelly was the one who mentioned it. That information wasn´t challenged, so I assumed it to be true.

AFAIK it was an extrapolation based on an overclocked G70.

 

[Alpha_Spartan]

Okay. So is the 7800 GTX architechture radically different from the Geforce 6 series? The 6800 Ultra can peak out at 600 million vertices per second. So if the architectures are similar, then my overclocked GT can set up more polygons than the Xenos. I must say that if this is true, it's a bit disappointing.

 

[one]

On a slightly related note...

http://www.theinquirer.net/?article=25197

The guy in charge of the demo told us that the graphic part is much more powerful than even the upcoming R5XX, series and that ATI's desktop unit will match Xbox 360 graphics with a next generation scheduled for next year.

The Rubi model was rendered with 70,000 polygons while some of its elements had as many as 120,000 polygons per character. The most demanding scene had the dreamlike number of one million polygons per scene.

We looked at the demo and it did look amazing but we have to add that we haven’t seen any new cool effect feature that we haven’t seen on a PC. We still haven’t seen so many polygons on a PC, however, and the guy suggested that we should focus on the power of rendering when it comes to this console. A PC cannot match this now, but of course will match the performance.

 

[Shogmaster]

Okay. So is the 7800 GTX architechture radically different from the Geforce 6 series? The 6800 Ultra can peak out at 600 million vertices per second. So if the architectures are similar, then my overclocked GT can set up more polygons than the Xenos. I must say that if this is true, it's a bit disappointing.

Haves you seen any games running on your overclocked GT anywhere close to 600 million pps? Come on, dude....

That 600 million figure for 6800U gotta be for transform only, not set up and drawn.

 

[Titanio]

The setup rate is, as far as I can see, inconsequential as long as there's "enough". I mean, a setup rate of 500pps could allow up to 8.3m polys per frame (at 60fps) - which at 720p could mean multiple polys per pixel (though that is assuming no overdraw..or you could consider it as polygons the size of pixels with a certain level of overdraw).

I doubt any game next-gen will come close to that limit. With next-gen shaders, you're not going to get ~8m polys per frame. It isn't a bottleneck, so I would not invest too much time fretting over it.

 

[Alpha_Spartan]

Haves you seen any games running on your overclocked GT anywhere close to 600 million pps? Come on, dude....

That 600 million figure for 6800U gotta be for transform only, not set up and drawn.

So if that's true, then the Geforce 7 has to be radically different from the Geforce 6. I mean, that's a huge difference from 600 million transform to 800 million set-up. The G70 (RSX) has to be more revolutionary than we thought it was!

 

[Titanio]

So if that's true, then the Geforce 7 has to be radically different from the Geforce 6. I mean, that's a huge difference from 600 million transform to 800 million set-up. The G70 (RSX) has to be more revolutionary than we thought it was!

I don't think it's clear what numbers are transform and what numbers are setup, tbh..

The only thing I've ever heard about G70's setup is that, asides from improvements due to clockspeed and/or additional logic to accomodate more vertex shaders, it's also been made more efficient vs the Geforce6 series.

 

[Atsim]

So if that's true, then the Geforce 7 has to be radically different from the Geforce 6. I mean, that's a huge difference from 600 million transform to 800 million set-up. The G70 (RSX) has to be more revolutionary than we thought it was!

Didn't he just say that games don't come anywhere near that limit? Why is that revolutionary? Seems like a waste to me...

 

[jvd]

Okay. So is the 7800 GTX architechture radically different from the Geforce 6 series? The 6800 Ultra can peak out at 600 million vertices per second. So if the architectures are similar, then my overclocked GT can set up more polygons than the Xenos. I must say that if this is true, it's a bit disappointing.

I thought it was explained to you that peak and sustanied are two diffrent things . The tech with the lower peak can very well be the tech with the higher sustained.

Not only that but once again you fail to find numbers with lighted , transformed , shaded , texture polygons with advanced filtering on them


All these features are just as important or more so when talking about the poylgon performance because there is a vast world of diffrence between the polygons on the ps2 and the polygons on say the rsx . The diffrences are the effects and filtering along with texturing on the polygons themselves which all affect the set up engine and peak polygon preformance .

TIll you know how they perform with in game functions your not going to know enough to claim something is not impressive .

 

[Atsim]

Does Alpha_Spartan know what he's talking about? Because I don't know shit about technology and I still know that he is comparing apples to oranges...

 

[Alpha_Spartan]

Does Alpha_Spartan know what he's talking about? Because I don't know shit about technology and I still know that he is comparing apples to oranges...

All I did was ask questions and come to conclusions based on the information known. I welcome corrections. Please tell me where you disagree.