Putting NVIDIA Radiosity rumors to rest [Archive]

View Full Version : Putting NVIDIA Radiosity rumors to rest

Shadowmage

10-17-2005, 01:54 PM

I asked, and they delivered. :toast:

http://www.beyond3d.com/forum/showthread.php?t=24589

Quick summary:

I asked this question:

Someone I know claims that the ATI X1800 won't be able to do Radiosity, while NVIDIA's 7xxx series can, in upcoming games. Looking at NVIDIA's website, they list Radiosity as one of the features of 7800GTX, while it's not listed for ATI's X1xxx series.

Can someone elaborate on this? How come the X1xxx can't perform Radiosity in upcoming games?

Thanks!

-----

So how does this work out?

Precomputed Radiance Transfer (aka "Realtime Radiosity") support, allowing for Real-Time Subsurface Scattering and Soft Shadowing.

http://www.artificialstudios.com/features.php

Also, Unreal 3 is said to use Radiosity.

If it's not real radiosity, then what is it?

PRT is not a total solution however, since it only works on rigid objects.

Well, there is Local, Deformable PRT for deformable objects (siggraph 2005)
http://msdn.microsoft.com/library/de...ormableprt.asp

ATI personell provide an answer:

There are many situations where PRT techniques can be used successfully in dynamic scenes. The thing to keep in mind is that PRT is an approximation, but that's ok because all real time lighting techniques are approximations

We used PRT and Irradiance Volume techniques for diffuse lighting on all the characters in the Ruby2 demo (Ruby: Dangerous Curves).

http://www.ati.com/developer/demos/rx850.html

If you run the demo, there are modes for viewing the various lighting terms. Unfortunately these additional modes aren't in the Mpg or Quicktime movies.

If you're interested, there's presentation on some of these methods here:

http://www.ati.com/developer/gdc/GDC...acticalPRT.pdf

--Chris

A fellow Radiosity noob asks:

Im lost, did anyone answer the question in first post ?
Some say NV has that as an extension and it would speed up "radiosity" big time and other cards except GF7 just dont have the power to do it.

Gets answered very directly:

it got answered if you read the thread: there is no way radiosity (the algorithm) can be done in a useful-for-games manner on any todays (and, the way the algorithm works, future) hardware.

the fact that nvidia has it on their page is just some marketing fuzz. they possibly put that in because of some PRECOMPUTED radiance transfer demos, that run in realtime (they, in realtime, show precomputed data.. wow ).

and no, there is no reason the ati hardware can not handle any solution or fake solution nvidia has to show anything. quite the other way round..

the ati card is definitely capable of hardware accelerate radiosity calculators, thus could speed up the lightmap generator in a q3 level editor (or similar stuff). it's gpgpu capabilities are still waiting to really show up.

simple said: the dude who told the OP, that nvidia can handle radiosity while ati can't, simply said something wrong.

The GameDev article is brought up:

http://www.gamedev.net/reference/pro...y2/default.asp

Would this work?

Answered:

yeah, but afai remember, this is simply some approach similar to cubemapping. but i'm unsure, just looked at the link, haven't read (and it's long time ago). nothing to do with real radiosity (but has some of the effects of it). and definitely not an "automatic all purpose" way. means a solution. only a hack.

Asks about the rumor with NVIDIA and Radiosity:

The rumour is that the guy that wrote that article in link posted above sold hes idea to Nvidia and it is going to be used in UE3.

as expected, it's BS!

and that rumor is just bs afaik. i mean, you can always rumour around, but that doesn't mean anything.

LOL SO TRUE

Except fooling people that dont question things to choose NV over ATI.

Turok

10-17-2005, 02:24 PM

So ATi's x1xxx line will be able to run the Unreal 3, Offset, and Reality engine as well as nVidia's with Radiosity? same with VTF (Vertex Texture Fetching)?

I dont think ATi would be stupid enough to release the x1xxx series without being able to play UT2k7 and Gears of War. Performances may be different, and hopefully image quality isnt lost, but it would kill them if they cant run UT2k7 head to head with the 7800 series and would be even worst if their cards cant run the game at all.

Shadowmage

10-17-2005, 02:31 PM

They won't be using "radiosity". They'll be using PRT, which is similar to Radiosity, meaning it's precomputed.

X1xxx doesn't support VTF, but it supports R2VB, which performs better than NVIDIA's VTF implementation.

DilTech

10-17-2005, 02:41 PM

No, shadow is wrong, as always...

yeah, but afai remember, this is simply some approach similar to cubemapping. but i'm unsure, just looked at the link, haven't read (and it's long time ago). nothing to do with real radiosity (but has some of the effects of it). and definitely not an "automatic all purpose" way. means a solution. only a hack.

His "source" didn't even LOOK at the algorithm NVidia used, they just brushed it off! Read it again shadow, and again, and again. Read the explanition of PRT then the one of the algorithm I've posted several times. They're DEFINITELY different.

I'm tired of explaining this, if what NVidia was boasting was PRT, it'd WORK ON THE 6 series, yet notice it's only mentioned for the 7 series.

Next, PRT doesn't handle MOVING light-sources, only still ones.. Fire a rocket thru a room using prt and it won't effect the scene. Fire a rocket thru one using nvidia's and it DOES effect it!.

You see shadow, there's a BIG difference between PRT and nvidia's RealTimeRadiosity. It's NOT PRT!

Therefore, it's not laid to rest, you're still wrong...

Also, Shadow, Epic IS using VTF, they stated it with the 6800 ultra showing of the UE3.0 engine! They explained it when they showed VDM on it. You're going by guesses from lesser quality developers as to how it's done. Ones who obviously never heard of optimized code. Considering they said 33m displaced vertices per second via a VTF is too slow, when most vertices in a scene don't need to be displaced, and you can easily double up the rendering by simply using dynamic branching in the pixel and vertex shaders, as alot of the vertices end up looking identical anyway....
:slap: :slapass: :woot:

Care to try again?

[XC] leviathan18

10-17-2005, 03:09 PM

di tech should be named NVIDIA ULTRA NAZI FANBOY lol....... j/k pal :toast:

DilTech

10-17-2005, 03:28 PM

Eh, leviathan, I'm just tired of PR bs...Ironically, it's all done by ATi's side.

I know you're kidding either way, but for the record, people will see things the other way when the R600 comes out, as I'll DEFINITELY be buying that card as things currently stand. Any card that has 3 years of work known of publically, plus however many done before that HAS to be a monster. It's current implementation in the XBOX360 is astonishing, and thus I cannot wait to see what another year of work does to it for the pc.

BTW, my nickname on other boards is conspiracy brotha.. can you guess why?

Shadowmage

10-17-2005, 09:45 PM

Dude, if you can prove that NVIDIA uses this implementation, then I'll 100% believe you. However, you don't have a shred of evidence.

Until then, it would make more sense to believe the graphics professionals over a graphics "nobody" (no offense, but we (myself included) are all "nobodies" compared to them)

Tim

10-18-2005, 02:51 AM

Great Comeback Shadow :rolleyes: :stick:

perkam

10-18-2005, 06:18 AM

It does seem that everyone is quite well informed at Beyond3d on the issue, but the only source we can go by is "ChrisATI"...as he's the only official source who answered, Dil is right in countering "DavePerman"s opinion being taken as fact. However, if ATI believes PRT can counter VTF, we have no evidence from either company to refute (or believe) this claim and unless a movie/image can be shown comparing the two technologies, there is NO way for any one side to have a definite advantage or disadvantage over the other, especially since we're talking about one game engine by one developer, and one that wont be out until well into 2K6.

Interestingly enough though, games like gears of war (if I'm not mistaken) use this engine and work perfectly fine on the Xbox 360's graphics tech (ATI Xenos), so its not like its a PR cover-up or anything. Its a feature that they've worked with in the past and will most likely have support for in the future.

Perkam

Tim

10-18-2005, 08:03 AM

If the Unreal 3 engine can do it on every hardware, both nVidia and ATI...fine with me...it would be extremely sucky if only nVidia could do it.....I don't think Epic would make such an enormous mistake...

DilTech

10-18-2005, 10:28 AM

perkam

10-18-2005, 11:17 AM

they claim it's too slow on non-unified designs

And the x360 costs a lot less than the X1800 ;) I just wanna buy one to tear it apart and get a pic of the chip running :D (i know they gave us a huge pic of the gpu, but not actually running :D )

Perkam

Shadowmage

10-18-2005, 06:50 PM

No, ATI said that VTF requires too many transistors to create a fast implementation. It wasn't worth it.

slider99

10-19-2005, 06:57 AM

Dil said

The R500 has VTF, as ATi said, they use it on their unified shader designs, they claim it's too slow on non-unified designs. Therefore, of course it works on the x360.
Shadow said

No, ATI said that VTF requires too many transistors to create a fast implementation. It wasn't worth it.
While what both of you say could be correct at the same time (concerning the R500 ofc), does anyone of you actually have any solid source or "proof" for these statements? It'd be very interesting to hear, since imo way too much is said loosely without a being based on a solid foundation from a rigid source.

Moreover, a little question mark actually arises upon hearing that the R500 (and R600 too) has VTF, but still VTF is not important? I mean, if ATI had a better solution than VTF, then why use VTF on their better piece of hardware (R500)? If VTF isn't good for anything at all, then why waste the "too many transistors [required] to create a fast implementation" on doing it in the end? This seems a fair bit too inconsistent for me to buy, rather it smells coverup. However, I'm not the expert and I'm open for other possibilities.

ahmad

10-19-2005, 07:43 AM

perkam

10-19-2005, 08:05 AM

To be honest with you, I just want to see your post get ripped to shreds
Quite honest of you ahmad, but the same can be said of Shadowmage and nzone :p:

Two sides to every story I'm afraid. Let's keep on topic here ;)

Perkam

P_1

10-19-2005, 08:08 AM

Dil said

Shadow said

While what both of you say could be correct at the same time (concerning the R500 ofc), does anyone of you actually have any solid source or "proof" for these statements? It'd be very interesting to hear, since imo way too much is said loosely without a being based on a solid foundation from a rigid source.

Moreover, a little question mark actually arises upon hearing that the R500 (and R600 too) has VTF, but still VTF is not important? I mean, if ATI had a better solution than VTF, then why use VTF on their better piece of hardware (R500)? If VTF isn't good for anything at all, then why waste the "too many transistors [required] to create a fast implementation" on doing it in the end? This seems a fair bit too inconsistent for me to buy, rather it smells coverup. However, I'm not the expert and I'm open for other possibilities.
http://news.teamxbox.com/xbox/9487/ATI-Releases-Xbox-360-GPU-Technology-Papers/
Finally some sensibility in this thread.

slider99

10-19-2005, 09:01 AM

Thanks P_1!

Seems like ATI considers VTF to be one of the most important features of the R500 Xenos if this is how their overview of the chips looks like:
http://media.teamxbox.com/dailyposts/xbox360/specs/xenos/02.jpg
http://media.teamxbox.com/dailyposts/xbox360/specs/xenos/06.jpg
http://media.teamxbox.com/dailyposts/xbox360/specs/xenos/11.jpg

and perhaps these? at least they're about displacement something...

http://media.teamxbox.com/dailyposts/xbox360/specs/xenos/17.jpg
http://media.teamxbox.com/dailyposts/xbox360/specs/xenos/18.jpg

Looks like indeed ATI chose to implement VTF on the R500 and that it's even one of the major features! Seems like DilTech indeed was right on this one too.

Sorry for the literally long post!

DilTech

10-19-2005, 09:29 AM

Called it, didn't I slider...

DilTech

10-19-2005, 09:34 AM

I looked over the rest of the thread at Beyond3D... knowledgable folks.

DilTech, why don't you head on over there and make a post. If you say you don't want to register, I can or ShadowMage can quote you. That is, if you feel you are sufficiently versed in this and think you can successfully defend your point in this debate.

To be honest with you, I just want to see your post get ripped to shreds :slapass:

Why bother with beyond3d when I can just take it up with my boss, of who makes the tools those guys use in the first place, and completely cut out the middleman? Compared to him, they're all noobs. No matter what you know, there's always someone who knows more than you on that subject, and in this case I'm getting my answers directly from him.

Something about beyond3d, most of their members think their opinions are golden, and half the time they won't even look into what a newer member has to say(hence, them completely skipping the posted algorithm). Another thing, ATi has quite a bit of staff members there, and we ALL know what happens to a site once ati gets staff there, biasedness occurs... Not sayin they don't know a few things, but they're not the be-all end-all like shadow makes them out to be. They're just small fish in a very big pond, that's all.

Just to give you an example of beyond3d's biasedness...
They stated that VTF is too slow, and that at max it could only handle 33 million vertices per second(notice, that was for the 6800NON-ULTRA, the 7800gtx can do DOUBLE to TRIPLE this!), which is completely not-true. It can DISPLACE 33million vertices per second, however, when using instancing and dynamic branching, you can effectively cut down majorly on the gpu power by simply replicating all vertices that are identical...

You figure, in a single image, there's thousands vertices that are identical, and less than 1/6th of the total vertices would need to be displaced. Now, you can handle displacing 33million different vertices a second using VTF, you only need to displace 1,000,000 and about 200,000 all end up identical. That wouldn't hurt NEARLY as much as ATi claim it would, but it can all be done exactly as I just stated it to be. It'd be doing 4/5th of the work it originally would've had to do. However, the guys at beyond3d didn't even mention this method, they simply say "it's to slow"... Gee, REAL PRO'S, ain't they shadow?

The 6800nu can do it fine, as can the 6800gt and ultra. The 7800gtx can do double to triple that with ease.... So why didn't your "pro's" bring any of this up when asked about it shadow?

ahmad

10-19-2005, 09:42 AM

Something about beyond3d, most of their members think their opinions are golden, and half the time they won't even look into what a newer member has to say(hence, them completely skipping the posted algorithm).

Just to let you know before hand, the radiosity argument is out of my scope, so I am still learning. Someone in that thread did comment on the algorithm FYI.

DilTech

10-19-2005, 09:50 AM

He said he looked at it, but didn't read it at all. ahmad.

perkam

10-19-2005, 09:52 AM

If the R500 has it, most likely will be implemented on the R600...definitely not the R580 if its using a faster or more enabled version of the current R520/RV530 architecture.

Perkam

Shadowmage

10-19-2005, 09:55 AM

Dil said

Shadow said

While what both of you say could be correct at the same time (concerning the R500 ofc), does anyone of you actually have any solid source or "proof" for these statements? It'd be very interesting to hear, since imo way too much is said loosely without a being based on a solid foundation from a rigid source.

Moreover, a little question mark actually arises upon hearing that the R500 (and R600 too) has VTF, but still VTF is not important? I mean, if ATI had a better solution than VTF, then why use VTF on their better piece of hardware (R500)? If VTF isn't good for anything at all, then why waste the "too many transistors [required] to create a fast implementation" on doing it in the end? This seems a fair bit too inconsistent for me to buy, rather it smells coverup. However, I'm not the expert and I'm open for other possibilities.

An ATI representative stated so on B3D forums (it was Humus)

---

BTW, we have no idea who your boss is, so it doesn't help quoting him.

---

BTW, I already provided sufficient proof in the VTF thread. Just go look that up if you didn't already see it.

EDIT: Read this if u still need info:

http://www.beyond3d.com/forum/showthread.php?t=24352

twiggy

10-19-2005, 11:32 AM

Think autodesk shadow...

Doesnt really matter does it? I could say my boss was whoever and quote alot of things but without proof its still shiat IMO. :stick:

DilTech

10-19-2005, 11:34 AM

I was simply pointing out that he stated that "we don't know who you work for"... Also, notice I deleted the post as I figured it'd end up with a response just like yours twiggy.

Ubermann

10-26-2005, 08:15 AM

I downloaded the SDK 9.5 from Nvidia to find any info about the radiosity.
Could not find any document in it with radiosity title.
So i did a "search for text" on the 800meg documents and other stuff in it to even find the word "radiosity".
It found 0 matches.
Wazzup ? Is there some new SDK not on their site ?
This files was dated 2005-09-23.

slider99

10-26-2005, 10:58 AM

Yes, let's put all our trade-secrets on the front page!

Ubermann

10-26-2005, 11:02 AM

Huh ? I mean its a hardware thing thats not in any other GPU according to what i have been reading.
So if i want to make whatever demo/game stuff program with use of radiosity i cant ?
Im totaly lost with that answer slider but please explain.

Shadowmage

10-26-2005, 12:21 PM

Duh, it means that the Radiosity thing is FALSE, as I had proved in the first post.

Thanks for giving me more evidence in case certain people claim otherwise :D

perkam

10-26-2005, 12:43 PM

Duh, it means that the Radiosity thing is FALSE, as I had proved in the first post.

Thanks for giving me more evidence in case certain people claim otherwise I'm afraid this doesnt help your case Shadow. There are different terms used for radiosity across the industry, so that may be the reason ubermann may not have found it.

I'll see what i can find on their website.

Perkam

Shadowmage

10-26-2005, 12:57 PM

OK you're right, Radiosity is a general term describing a certain concept.

However, it still doesn't make sense in that real real-time Radiosity is literally a pipe dream, so if some special algorithm worked, and worked WELL (aka better than PRT), it would be quite famous, and NVIDIA would do their usual hyping of this feature (remember NVIDIA's hype for SM3.0 and NVIDIA's current hype for VTF?)

NVIDIA isn't hyping. In fact, there is no proof that it exists. So where did the rumor start from?

On a side note, this would make a very good psychological study (the spread of rumors with no proof).

DilTech

10-26-2005, 03:06 PM

NVidia hasn't hyped VTF in the slightest. On no website, and in no slides or anything is this mentioned.

Website made the statements themselves, NVidia had nothing to do with it... Infact, NVidia's PR team is usually pretty quiet, something ATi should learn from.

Also, Shadow, it all springs from the "featured technology" page on www.nvidia.com for the 7800. Either way, what should likely be done is this, wait until it's used and then resurrect said thread. NVidia likely won't make a statement, ATi is likely to just blabber saying "NVidia can do it, so can we"(like always), and we have no way of knowing the actual name of the technique.

Therefore, this is just going to turn into he-said, she-said battle, and it's doing nothing but making us both look bad shadow. You know as well as I do it's getting tiresome, and wasting time for both of us.

Ubermann

10-26-2005, 03:26 PM

VTF should be a good thing in games.
But i cant find any example of radiosity nor can i find 1 pic to see what it look like in real time compared to a non radiosity pic.
I mean they have it on their website as "featured technology" but they dont want to show it ?
Maybe i posted this in wrong radiosity thread cuz i have no idea if it real or not yet, i just find it odd that theres no info at all about it.

DilTech

10-26-2005, 04:19 PM

They show it in the madmodmike demo uber.

Ubermann

10-26-2005, 10:45 PM

Thank you.

Ubermann

10-27-2005, 03:04 AM

Please delete that.

saaya

10-28-2005, 04:44 AM

Eh, leviathan, I'm just tired of PR bs...Ironically, it's all done by ATi's sidenow heres something i can disagree on :D

about the rest, no idea, beyond my level of understanding graphics :D
so what exactly is this thread about? diltech and shadow are in some kind of a battle of minds, diltech wont post on beyond3d so he posts half a thread from there on xs to win this battle of minds?

shadow, please dont post half a thread like you did but only a link and maybe point out what in that thread you want to discuss.
fuerhermore this is all just rumors, from all sides as i havent seen even the slightest bit of proof for either side whatsoever, so this is not news!
please post this in the computer related open discussion section, where i just moved this thread.

isnt it funny how both of you, diltech and shadow admit you dont really know a lot about this topic, yet you try to win this argument by quoting people you claim know everything about this? :P

you remind me of two french people without any english sklills fighting over how to correctly pronounce "butter" :D

as i see it there is no proof, so just sit back and wait for nvidia and ati giving us some material to play with... and i dont mean some whitepaper but a techdemo or patch for a game enabling radiosity.

slider99

10-28-2005, 05:01 AM

:with: ^ Dito

DilTech

10-28-2005, 11:08 AM

now heres something i can disagree on :D

about the rest, no idea, beyond my level of understanding graphics :D
so what exactly is this thread about? diltech and shadow are in some kind of a battle of minds, diltech wont post on beyond3d so he posts half a thread from there on xs to win this battle of minds?

shadow, please dont post half a thread like you did but only a link and maybe point out what in that thread you want to discuss.
fuerhermore this is all just rumors, from all sides as i havent seen even the slightest bit of proof for either side whatsoever, so this is not news!
please post this in the computer related open discussion section, where i just moved this thread.

isnt it funny how both of you, diltech and shadow admit you dont really know a lot about this topic, yet you try to win this argument by quoting people you claim know everything about this? :P

you remind me of two french people without any english sklills fighting over how to correctly pronounce "butter" :D

as i see it there is no proof, so just sit back and wait for nvidia and ati giving us some material to play with... and i dont mean some whitepaper but a techdemo or patch for a game enabling radiosity.

Also, Shadow, it all springs from the "featured technology" page on www.nvidia.com for the 7800. Either way, what should likely be done is this, wait until it's used and then resurrect said thread.

The proof is in the tech specs and madmodmike demo, but the point is that until it's used in games it's useless to all of us(much like truform on ati cards, and ultrashadow on NVidia cards before the Doom 3 engine). We don't play technical demos, we play games.(and apparently some of you "play benchmarks")

If you cannot tell saaya, I whole-heartidly agree with you on this. I don't post on B3D simply because they're not as knowledgable as shadow makes them out to be, and think their opinion is the be-all end-all in every case, which it's usually not. Ontop of that, due to the 3+ ati reps they have on their site, their opinion tends to sway towards ati's side. I don't deal with ego-centrical people just to make a point, it's not worth my time or arguing. If you're going to ask about hardware, you should always get info from non-biased locations.

Also, thanks for moving this, it didn't belong in the news section at all, I asked for it to be moved pretty shortly after it was posted in the first place. :D Oddly enough it wasn't.

etherealrhythm

10-29-2005, 12:59 AM

all that im gonna say is ive worked with radiosity in relation to the mental ray rendering engine with Softimage XSI (3D App). i dont know much about the science behind precomputed and real-time radiosity solutions, but if they could ever get something vaguely resembling radiosity running in real-time in a game engine, id cream in my pants with graphical delight.