DX10 demo

I'd say it's quite impressive, 21 soft shadows at 244 fps. For an 8800gts.



Looks like DX10 IS all that and a bag of chips.

Now all we need is a Oblivion DX10 patch and get 30 fps back...

 

doesn't look all that great

 

It's not supposed to, it's soft shadows.

We aren't yet supposed to have the technology to render soft shadows in real time...

Most effects were faked.

This is true soft shadows. If they could make a full game with soft shadows, OMG.

http://www.gamedev.net/reference/articles/article2193.asp

Soft shadows allows a perfect shadow of a character.

A big game with huge amounts of soft shadowing? DX10's future.

Soft Shadows would still murder a game like Oblivion, but any smaller game could use Soft Shadows, dynamic lighting, etc, to create a truly impressive game.

 

mmmm. soft shadows + HDR = incredible.

 

....it's just shadows....i dont see the big deal.

 

Soft Shadows by themselves aren't a big deal, but if DX10 can run 21 soft shadows at 244 fps, it could make any game run far faster...

Or have alot better effects.

This kind of makes me wonder how Crysis on DX9 will look... Maybe it runs so well because of DX10?

 

Not necessarily. When dealing with computer graphics, there's no such thing as "real" soft shadows, or any other effect. Everything is faked one way or another. The soft shadows you've seen in games so far have been faked, and the ones in that pic are faked too. They might be faked in different ways (and I think I'll poke around the DX SDK to check it out one of these days because it does look interesting), but it's still fake. And for such a simple scene, the framerate doesn't mean much. Especially since we don't have a DX9 comparison.

 

First off, I only count 15 shaddows, and it is a still picture, so I would hope it had good FPS. Also, the graphics do not look that great.... And I have this card.... So, I really think that my X1400 could do 15 soft shaddows.

 

Soldiers are also giving soft shadows. That makes 21.

And no, your card really couldn't do it at that high a framerate.

A still picture wouldn't make the framrates higher or lower, because it is still loading the models and textures.

 

YOU CANT SEE THE SOLDIERS!!!! AND IT IS A STILL PICTURE SO YES IT COULD. Wow its not like its a video...

 

Um, the soldiers are right there, and if you look closely, you can see their shadows.

A still picture changes nothing, you still hafta load the models and textures.

 

a still photo absolutely makes a difference on framerate, zellio.

Still photo only has to load the textures ONCE.
Video has to redraw everything for every frame. If you somehow think that same scene would be 250 FPS as video, then you don't understand the laws of physics.

 

I think they are not that realistic, really. Shadows should look sharper when closer to its object (and blurrier when further from it's object), here that is not the case. In Crysis' demo videos, you can distinguish the leaves of a tree's shadow while the leaves are 6 meters above the ground, that's not realistic.

 

It is a screenshot of a running DX10 demo. So yes, it is running at that framerate. But as have been said several times above, it is a tiny simple scene, and there is no actual *game* running in the background. These two things always boost framerate *a lot*. They also make it impossible to compare to what we're used to from games.

But as I've said, soft shadows are nothing new, and have been used in more than a few games so far. So this is really nothing new. *unless* of course, it is a noticeably better implementation, which is impossible to tell from this screenshot...

 

Dirxess, I dont want to sound rude, but doesnt that look silly? i mean, if you want to have a detailed leaf, go to the park! I think we are talking about stupid details, since, when you go shooting in a game, at least me, you dont stop to see if all the leaves proyect their shadows.

See what i mean?

 

Want to go back to this, then?

People back then screamed they did not need any more detail or realism.

 

Those shadows doesn't seem to be realistic shadows compared to the form of the soldiers and cars

 

1. "Stationary" vs "In Motion" is irrelevant. For argument's sake, let's say that one frame takes 0.1s. Whether that was one frame rendered as a "snapshot" or a sequence of ten - it still took 0.1s. The debate I think you're trying to have is "Live action" vs "Prerendered".

2. This demo is showing soft shadows from ambient light. Those will be bloblike and only vaguely resemble the original model. If you need proof, step outside on a sunny morning or evening.

 

Try standing still in a game. Try an older game, where the models don't move.

Does the fps improve?

Thought so.

Obviously these people have never done game programming, or even edited models.

Let's give a lesson.

When a level is loaded, everything is loaded into memory at start of level, rather standing still or moving. Why do you think games with wide open spaces have long ass loading times?

Moving makes hardly a difference at all.

Are you gonna tell me that by standing still, it totally negates the fact that there are tons of 5,000-10,000 poly models on that screen?

You telling me that it's NATURAL for it to run faster while standing still? You don't understand the simple process of a pc. It isn't natural. It does things in a way that's not natural.

 

*sigh*
Zellio, first you are firing up again for no-good reason.
Secondly, in the games, hidden surface removal techniques are used.Meaning that the geometry that is out of viewing volume is discarded.See this for more info:
http://en.wikipedia.org/wiki/Frustum_culling
(and you can see some screenshots indicating the viewing volume)
http://www.cs.umbc.edu/~rheingan/435/pages/res/gen-8.Viewing-single-page-0.html

It means that in Oblivion when there is a HUGE landscape in front of you.Much of the geometry that is out of the viewing volume is discarded by the CPU before being sent to the GPU.If they don't do this the framerate will be reduced to something like 1/10.This part was about the geometry submitted to the GPU by the CPU.
The GPU itself discards the pixels that cannot be seen so they will never make their way into the pixel shader.
-Conclusion : In a game [the frustum culling is always done nowadays] the framerate might differ greatly if the scene is changing or is stationary.

 

It's true what you say, it IS a tiny detail! It still annoys me though! It should be very easy to fix this little incompletion (right word? doesn't matter)

 

Differ greatly huh?

So basically it will cull when it's stationary, but not when it's moving?

Light culling, model culling, if it's done while moving it's done while stationary, fool.

Stationary or not it would have NOTHING to do with movement. What it would have to do with is objects in the map it must load.

http://en.wikipedia.org/wiki/Buffer_%28computer_science)

As long as the area is loaded, and you have enough ram, the fps doesn't change standing still or moving. It's the same because the engine has already buffered the area. When it needs to buffer another area it loads it.

It's not hard.

 

I'm just gonna hafta put you in your place, eh?

What better way then a game that SHOWS when you run!



Look at us stand still!



We lost... one fps for running. WOW.



Well, gained 4 fps that time.

You guys were saying...?

It all has to do with buffering. You need enough ram to buffer the area. If you don't, the ram ends up loading faster, and you get delays.

And huge decrease of ram ends up with major buffering problems, and a game engine that stalls every 5 seconds.

Simply put, not enough horsepower, or not enough ram to buffer, will reult in a huge decrease when moving. But that's not due to stationary objects, it's due to not having enough ram to buffer and enough cpu to load/gpu to run.

Now that we have substaintial proof instead of people running their mouths, I'd like to move back to the original topic.

 

My master is programing.... And no standing still does little in a game. With a still picture the crapyest card in the world could run that at infinate framerate. So for your obvious stupidity thinking that no one understands, I confirm that you just do this type of thing to get attention.

Also, you dont even have this card, I do, I know a bit more about its capablities than you would, AND that picture is just a small part of the game, and has the game cut off in the back, so it really makes the 8800GTX look bad.

 

Sorry, to go offtopic with the above ^^*hot* discussion, but i am pretty dissappointed with the particular DX 10 demo..

Soft shadows?? Great!! So wat?? Dx 9 brought HDR with it which made the graphics look real nice.. I dont load up the latest game to see soft or hard shadows!! I rather see better realistic graphics..

EDIT: If itsnt obvious, this post is to divert the topic from the flame war..

 

Alright man, no more arguments from me, I give up.

All I'll say is, good luck on the programmer biz, but lose the 'holier than thou' attitute, NOW, because you have ALOT to learn.

Just try 3dmark01, you'll find the same stationary objects being used to test power. Test it on different machines, you'll find faster machines give more fps.

Every object still has to load. If it did as you say, we'd play games that would be 200 fps and drop to 40 when we move.

All animations take is a little bit of power. Take all the animations away, you still have mapping, shading, modeling, texturing, and the game is still programmed to run at how many fps you set it to. They all hafta load, and they load at the beginning. Animations are a small equation.

You may end up far better than I did, esp. since all I did was indie, but lose the attitude and learn.

 

Wait... I finally get what you're saying. You think it's a static, test map, not running at 60 fps. A placeholder.

I hope you realize it's physicaly IMPOSSIBLE for it to be a test map? Test maps are made on a 2d operating system, this was a 3d test. The only way to show lighting is to do it in 3d.

Soft shadowing requires pixel shaders. No way this ran without 3d acceleration.

PHAIL

 

Download and run this demo and use the Space key to toggle frustum culling.Look around while the frustum culling is activated.
http://www.gametutorials.com/gtstore/pc-80-1-frustum-culling.aspx
The reason that you didn't notice a difference in the framerate in your little test of game is because that the ground you were walking around in was pretty much similar.There were no biggie graphical effects running around.Don't try to judge a very important aspect of graphics engine design just by a few samples.

 

Jesus, you don't give up, do you?

Fps dropping would have NOTHING to do with culling or movement, as you'd get the same fps standing still or moving! It has to do with what is loaded and is loading.

There is no difference in a moving scene, what matters is how much is in the scene. You could stand still and get the same fps as moving, if it's the exact same scene, because the scene must load.

Once again, do you wish me to turn on fraps and try your demo, try culling standing still and moving?

Must I make a VIDEO so you'd shutup?

http://www.tweakguides.com/Graphics_5.html

Interesting, no speaking of fast movement slowing down games, just that you NEED higher fps for fast movement.

 

Did you download and run that program or not ?
You are totally missing the point.

 

http://www.tweakguides.com/Graphics_5.html

Perhaps your perceptions are off.

 

Let's keep in mind one thing:

You can argue with me, you don't know me.

But start arguing with a site known for modding games, and you'll only make yourself look like an idiot.

Take my advice, end this argument now.

 

I don't know who is right or wrong, but we clearly have irreconcilable differences of opinion.

I'm terminating this thread and please don't pick up the same argument elsewhere.

John