Since the Nv30 Pic thread is closed...

[cellarboy]

nggalai: Please don't think my post was aimed directly at you, it wasn't. I was probably typing the thing while you posted yours!!

I dunno, I just don't get the whole ATI vs. Nvidia thing. I've owned both companys hardware and both have pros and cons. I know I won't be able to afford any of the new cards until we are probably on DirectX 15, so I couldn't give a hoot who's new card is the best.

What *does* annoy me though is that ATI have launched a new card, everyone who has had hands-on time with it seem pretty damn impressed, but within a week of this little tidbits of info start mysteriously leaking into peoples hands. In this case pretty dubious info (IMO) too. No real hardware has been seen as far as I know except for probably a very select few but enough FUD is being unleashed to for the NV30 to already being proclaimed the new king.

I seriously don't doubt for a minute that the NV30 will beat the 9700 for features - I think that's pretty much a given. What I wonder though is that are these extra features going to be applicable, fast enough or supported enough by developers targeting the DX9 and OGL2.0? If that is not the case then they will either end up being used by professional 3D artists for rendertests (I still doubt the likes of ILM and Weta Digital would do final renders on either a NV30 or 9700 when they worry about the differences in render precision between IRIX/LINUX and MIPS/Intel platforms). And in that case they are moot point for the average consumer (like the 8500's PS1.4 functionality??).

Objectivity seems to be lacking all around these days....

 

[darkman]

Why you would, I dunno, but take it from me that the images presented as 'NV30 screenshots' are nothing more than scaled version of the images found on the Exluna site. Why? Well, the images are pixel-perfect versions of the originals (within the constraints of reason 2 below). I would expect even the NV30 would produce some differences in texturing or shading. People keep citing the ATI natural light demo as some kind of confirmation of the validity of these images, but the ATI demo IS NOT identical to the original offline version and I don't think it ever could be. Neither is the Animusic demo.

Secondly, and most importantly, if anyone had bothered to check the pixel dimensions of these 'NV30' images, you would have noticed that they are 400 x 300 pixels in size. Now, download the originals from the Exluna site and whip 'em into Photoshop. Go to the Image Size dialog and look at the size - the bike is 700 x 525, the marker 640 x 480 and the teapot thingie 800 x 600. Now with constrain proportions checked, enter 400 in the width box and you see that the proportionally scaled images are, you guessed it, 400 x 300.

Prett co-incidental, huh?

It isn't a coincidence at all.

400/300 = 1.333

700/525 = 1.333
640/480 = 1.333
800/600 = 1.333
1024/768 = 1.333
1280/960 = 1.333
1600/1200 = 1.333

1.3333333 is the natural aspect ratio for NTSC and all the scaled-up derivative resolutions used on PCs from VGA on up. If you were to generate any of those three images at a normal resolution on NV30 or whatever other card(except the weird ones like 1280/1024) and scale them down to a width of 400, you guess it, you will get a height of 300 unless you want to screw up the aspect ratio.

If you really want to determine whether ben/NV's pictures are the same as the original ones on exluna's website, you must guess the scaling-down method used (sub-sampling, linear interpolation, bicubic, gaussian), try one of them on exluna's pics, do a bit-by-bit comparison of your scaled-down version and ben/NV's pics and show us that they are identical.

Cheers,
Darkman

 

[Dolemite]

Any way you slice it, its just unethical for a company to claim their products are doing something they aren't, which is not to say that it doesn't happen constantly, which it does. In this case, hopefully we can get a definitive answer on how these pics were rendered.

Hopefully someone with more time and interest than me will do the bit comparison, and I sense we will get some clarification on this from Ben's source at nVidia.

So, its not time to scream bloody murder yet, but I think this is worth following up on. Meanwhile, I'm sure the nVidia vs. ATI debate will rage on, just as thinly-veiled as ever...

 

[cellarboy]

Darkman: I didn't remember saying that my methods were foolproof Good point well made though - most of my 3D stuff is aimed towards print so that ratio of NTSC pixels is something that I wasn't aware of.

Doing a bit by bit comparison on these images isn't going to do a darn thing because both images have been jpeg compressed. The 'NV30' bike image looks like it's been JPEG'd at least twice with a quality setting of 5 or less to get the compression artifacts it has. Get me an uncompressed version of both these images and then we'll talk.

You do go to illustrate my original point though that some people will not accept that there is not even a doubt towards the origin of the images. What I am trying to illustrate is that these things should not be taken as undeniable fact. There is (or at least should be) enough doubt as towards the origins of these images for every single (unbiased) member of these boards to say, "Ok, so they may not be rendered by the NV30."

This whole thing reminds me of waaaay back in the days when software companies would stick the Amiga screenshot on the C64 box with the disclaimer in 1.5 point type, "images may not illustrate actual in-game graphics."

I hope they are, I really do. Anything that will cut my 3d package rendering time down would be great, even though I'll be purchasing my board when most of you guys are arguing the relative merits of the NV70 and R900!!

 

[nggalai]

edit: made obsolete by cellar's last post.rb

 

[cellarboy]

So, its not time to scream bloody murder yet, but I think this is worth following up on. Meanwhile, I'm sure the nVidia vs. ATI debate will rage on, just as thinly-veiled as ever...

I don't think there's any need to scream bloody murder either way. If the NV30 did render these images then sweet, more power too 'em. I would like to know that they were rendered on working hardware, not a hardware simulator or other 'approximation' of the architechture though.

At the worst case and these images are 'examples of the images possible with' an NV30 then as long as they can back it up come launch date then I will be suitably impressed.

I'd love to know what speed they were rendered at too. They are hardly the most complex scenes from a geometry or shader standpoint but any boost to rendering speeds in 3D rendering packages would be extremely welcome.

Actually, I have a question for the more knowledgeble of you out there. Assuming that the NV30 can render any complexity of Renderman Shader or comparable renderer (3D Max, Lightwave, Cinema 4D etc...), how useful as a *final* rendering tool would a standard off-the-shelf board actually be? We'd also have to assume that the chip would satisfy the rendering precision of animation houses (they couldn't have a frame rendered on Linux/P4 box under software and a NV30 frame that didn't match). Given that renderfarms typically have memory coming out of the ying-yang to hold massive textures and extremely dense geometry, how would a 128mb/256mb card manage? What would happen if the texture size and/or mesh (or even a useful portion of) was too large to fit into the card's memory? Wouldn't all the page swapping kill the performance, if it would work at all?

Just wondering.

 

[Mulciber]

Well, if anyone wants to see some funny misrepresentation of a product thats not nVidia, check out this link and go to videoshader example 2.
http://www.ati.com/vortal/r300/educational/main.html
Its worth a good laugh

 

[darkman]

Darkman: I didn't remember saying that my methods were foolproof Good point well made though - most of my 3D stuff is aimed towards print so that ratio of NTSC pixels is something that I wasn't aware of.

Sorry cellarboy, I just found it very amusing that anybody would use the universal screen aspect ratio of TVs/PCs to explain anything at all in this case.

Doing a bit by bit comparison on these images isn't going to do a darn thing because both images have been jpeg compressed. The 'NV30' bike image looks like it's been JPEG'd at least twice with a quality setting of 5 or less to get the compression artifacts it has. Get me an uncompressed version of both these images and then we'll talk.

Yes, I realised that after posting and checking that ben's pics were indeed jpegs. Now, where does the burden of proof lie from a theoretical point of view, I wonder? Does it lie with the graphics card manufacturers that have to give undeniable proof that the pictures they provide for previews/reviews were indeed generated by their cards as they claim (has ben actually claimed that in the first place at all?) ? Or does it lie with the people trying to refute these claims?

It just make things awfully complicated.

You do go to illustrate my original point though that some people will not accept that there is not even a doubt towards the origin of the images. What I am trying to illustrate is that these things should not be taken as undeniable fact. There is (or at least should be) enough doubt as towards the origins of these images for every single (unbiased) member of these boards to say, "Ok, so they may not be rendered by the NV30."

I don't think anybody claimed that there wasn't a doubt at all that the NV30 made these pictures. I just saw one semi-technical point trying to argue that they were fakes and it was using a universal constant that explains most screen-generated images on PCs to explain one particular case. It's like using the period of the female cycle or tidal waves to explain why the moon revolves around the earth in about 29.5 days. Universal constants explain many things, not the other way round.

Cheers,
Darkman

 

[Matt Burris]

A bit OT,
Any one here used to frequent the 3dfx news groups? I am sure I can remember seeing the name John Reynolds there, right? Also the Rev as well. Any of you guys remember a character going by the name Greg S. Truow (sp?) that used to frequent those forums? What happened to that bloke?

That was a very good place to be a few years back, i remember a few of these guys from the 3dfx groups.
I still have a .txt file with a few Truow-ism's it on a disc somewhere..

FWIW, Greg used to hang out at the 3DGPU forums a lot, and I believe he goes by the nick Feanor now. He hasn't been around much lately, possibly Real Lifeâ„¢ taking over a bit. I've been lambasted over roasting coals by Greg a few times.

 

[CMKRNL]

Its worth a good laugh


LOL, you're right! Demo 4 also looks bogus. Talk about exaggeration to drive home your point.

 

[LittlePenny]

If you don't like reading the truth head over to Hardocp.

I would atleast like to hear a solid apology from you to Brent.

I've been a member of Beyond3D for many years...don't tell me where I can post dig !

There is a large difference between being a member, and being a contributing member. Think of Beyond3d as being a priviledge for us less informed, and I hope that you will atleast stop the slander.

Sorry

 

[John Reynolds]

I closed the other thread because of all the unnecessary noise it was generating. Opening a new thread on a closed topic is not going to be tolerated, either.