PDA

View Full Version : new demo from Humus


Pages : [1] 2 3 4 5

{Sniping}Waste
07-01-04, 05:31 PM
http://www.beyond3d.com/forum/viewtopic.php?t=13716

Evildeus
07-01-04, 05:34 PM
Yeah, and some SM3.0 to compare would be cool. Anyone to test that on 6800s ;)

Ruined
07-01-04, 05:44 PM
Great, so when the Humus Game comes out we should expect great SM2.0 performance :)

DMA
07-01-04, 05:58 PM
Nvidia can consider themselves pwned

His new job at ATI? FUD slinger? Too bad..

Anyway, pretty cool demo i guess. No idea what i'm looking at though. :o

kahloq
07-01-04, 06:39 PM
True however its obvious that Humus hacked the demo together to make it look best for ATI. Not that I so much blame him really, since hes trying his best to make the "loser" this round(ATI) look better then what it is. Someone else in that thread sorta corrected humus and stated that for a demo, its all nice and fine, but real world gaming, it meant nothing.

I only reall say "loser" as far as ATI because of lack of PS 3.0 and generally the fact that the GT out performs the x800 pro. the XT PE and the Ultra are neck and neck tho.

Ahriman6
07-01-04, 06:40 PM
Great, so when the Humus Game comes out we should expect great SM2.0 performance :)

Wasn't the same sort of rhetoric thrown at 3DMark03, which actually did a good job of predicating FX boards' floating point performance? Dynamic branching is essentially the ability to jump between sections of code based on a variable or set of variables, and if ATI can somehow achieve this then more power to them.

kahloq
07-01-04, 06:41 PM
Thing is 6800 series already support dynamic branching

fivefeet8
07-01-04, 06:53 PM
Thing is 6800 series already support dynamic branching

And so will Ati's next video card. Heh.. Kinda makes you wonder if the technique will be used in real games at all because of it. Of course since he works for ATi now, we won't see a SM3 equivalent to the demo to compare. Atleast not until Ati has a card which also does SM3.

kahloq
07-01-04, 06:58 PM
Dunno, but ATI has no choice now but to include PS 3.0 on their next gen card. But they will have to competely make a new architecture to do it(hence why Nvidia did it........FIRST). But in the end, its actually good for everyone that ATI will be forced to do so, cuz that just keeps the tech advances coming and of course Nvidia's not gonna stand by idly either :)

Zeno
07-01-04, 06:59 PM
His new job at ATI? FUD slinger? Too bad..

Anyway, pretty cool demo i guess. No idea what i'm looking at though. :o

What he's doing is using relatively inexpensive stencil/depth/alpha buffer tricks to simulate "if" statements in the vertex/fragment programs. The "if" in this case is "if the light does not affect this pixel, don't calculate it". Kinda like the way stencil shadows are used to tell "if" a particular pixel is in shadow or not.

Just like with stencil shadows, though, he has to send the scene geometry through the pipeline (make another pass) to get this to work. I haven't completely thought through it yet, but I have a hunch that this technique may require an extra pass per conditional that you would otherwise put in your SM3.0 shader.

I made a couple posts in the B3D thread. I think it's dishonest for him to present it as he did.

My two cents.

jbirney
07-01-04, 07:08 PM
This is sad, Humus that has been making demos for many years does another one just to show off an effect. And he gets trashed just because he took a job with ATI. Good greif people.....

OWA
07-01-04, 07:19 PM
That's kind of what happens when you pick a side. The impartiality is no longer there so this is to be expected.

Ruined
07-01-04, 07:21 PM
This is sad, Humus that has been making demos for many years does another one just to show off an effect. And he gets trashed just because he took a job with ATI. Good greif people.....

Doesn't help much that his two recent demos haven't worked properly on Nvidia cards. In fact, his new "branching" demo runs 10% *slower* on Nvidia cards when its enabled compared to when its disabled, while ATI cards get 2-3x performance gain. Fishy to say the least.

pat777
07-01-04, 07:55 PM
Doesn't help much that his two recent demos haven't worked properly on Nvidia cards. In fact, his new "branching" demo runs 10% *slower* on Nvidia cards when its enabled compared to when its disabled, while ATI cards get 2-3x performance gain. Fishy to say the least.

Are you trying to say ATI is using an inefficient code in the "disabled" rendering path to purposely kill performance while using an efficient code in the "enabled" rendering path?

{Sniping}Waste
07-01-04, 08:35 PM
Doesn't help much that his two recent demos haven't worked properly on Nvidia cards. In fact, his new "branching" demo runs 10% *slower* on Nvidia cards when its enabled compared to when its disabled, while ATI cards get 2-3x performance gain. Fishy to say the least.

I think your problem is a user problem because others with 5XXX and 6XXX cards work with the last two demos. It must be the poor buggy Nvidia drivers fault :argh: .

Ruined
07-01-04, 08:43 PM
I think your problem is a user problem because others with 5XXX and 6XXX cards work with the last two demos. It must be the poor buggy Nvidia drivers fault :argh: .

Yeh, that makes plenty of sense. An ATI employee makes two graphic demos specifically designed to tout the features of ATI and make ATI cards look better and they just so happen to not work with Nvidia cards, so its Nvidias fault. And buggy drivers, lets not even go there. Come on now :)

PoorGuy
07-01-04, 08:54 PM
Lol the guy gets nothing but kudos and admiration for pushing out demos for the past two years and now he's the #1 ATi villain.

:lol2:

{Sniping}Waste
07-01-04, 09:07 PM
Yeh, that makes plenty of sense. An ATI employee makes two graphic demos specifically designed to tout the features of ATI and make ATI cards look better and they just so happen to not work with Nvidia cards, so its Nvidias fault. And buggy drivers, lets not even go there. Come on now :)


Not joking with you. Many ppl with 59XX and 68XX have no problems running the last 2 demos so you must be the problem for it not working on your system or its driver bugs. Try running released drivers from Nvidia and not betas.

ChrisRay
07-01-04, 09:08 PM
The last 2 humus demos do run fine. What are we discussing here?

Clay
07-01-04, 09:20 PM
Everyone please take a few breaths. Humus has contributed greatly to the community in the past. We all know that he is an ATI employee so just take that into consideration and let's spare everyone the drama. :) You might liken his demos now to demos such as Ruby or Chimpanzee. Not apples-to-apples of course but you get the idea.

Richteralan
07-01-04, 09:30 PM
is it all you guys or only myself.
I can't get the demo run correctly. The FPS jumping from 1 to 29593426 everywhere.

o yes
mine is FX5600 with 61.71.
Does it need DX9c? I only installed DX9b though.

Sgt_Pitt
07-01-04, 09:48 PM
yeah the demo runs fine at 112 fps.

Ruined
07-01-04, 09:57 PM
The last 2 humus demos do run fine. What are we discussing here?

No they don't you said yourself at Beyond3d:

"On my 6800 Non Ultra I get 37 FPS with The Dynamic Branching set to true. With Dynamic Branching To False its 40 FPS."

With Dynamic branching set to true you should get like 80+fps like ATI cards are getting.

In the 3DC demo, shadowing was broken.

ChrisRay
07-01-04, 10:02 PM
No they don't you said yourself at Beyond3d:

"On my 6800 Non Ultra I get 37 FPS with The Dynamic Branching set to true. With Dynamic Branching To False its 40 FPS."

With Dynamic branching set to true you should get like 80+fps like ATI cards are getting.

In the 3DC demo, shadowing was broken.


I was talking about the prior 2 humus demos, perhaps we are on different level. I still dont believe your criticism is justified until we get an answer from Humus,


His 3dc demo worked fine for me. It did have have a small quality issue. Considering his development platform currently I think thats expectable.


Have you ever heard the saying Ruined.

"Innocent until Proven Guilty" ?

991060
07-01-04, 10:03 PM
who knows where NV40 apply stencil kill? If it's the backend of the pixel pipeline, the whole technique is useless.