[QUOTE=Jawalt;21946991]Literally you could probably cut a fourth of the polys being rendered with some LOD maybe more in some cases.[/QUOTE]
I doubt it'll improve the framerate since the poly count of the spider model is already quite low ( around 288 )
[QUOTE=Xerios3;21947371]I doubt it'll improve the framerate since the poly count of the spider model is already quite low ( around 288 )[/QUOTE]
When rendering a few hundred you could probably chop the farthest's poly count by quite a bit so I think you'd be surprised.
Poly count doesn't really matter anymore, it's all about batching.
[QUOTE=nullsquared;21947421]Poly count doesn't really matter anymore, it's all about batching.[/QUOTE]
I don't think you know what you're talking about. On lower end/integrated cards which are the mainstream poly count is a huge factor.
[QUOTE=nullsquared;21947176][img]http://img248.imageshack.us/img248/3330/starwarsw.png[/img]
Equation from here: [url]http://www.leweyg.com/download/impview.html[/url][/QUOTE]
And that's exactly what I meant in a message I posted earlier.
How the fuck do you find out the equation of a tie fighter?
I mean, what's the way to accomplish that?
[QUOTE=s0ul0r;21947538]And that's exactly what I meant in a message I posted earlier.
How the fuck do you find out the equation of a tie fighter?
I mean, what's the way to accomplish that?[/QUOTE]
Insane mathematical shit
[QUOTE=Jawalt;21947487]I don't think you know what you're talking about. On lower end/integrated cards which are the mainstream poly count is a huge factor.[/QUOTE]
I have a few years of 3D programming experience, so I think I know what I'm talking about. Depending on your target machines, my statement is 100% correct. Your statement is only correct for integrated GPUs and insanely old non-integrated ones, like back in the NV FX5000 and ATI X800 range.
[QUOTE=s0ul0r;21947538]And that's exactly what I meant in a message I posted earlier.
How the fuck do you find out the equation of a tie fighter?
I mean, what's the way to accomplish that?[/QUOTE]
Probably just by combining the equations for the shapes that compose it.
-nvm, dumb reading-
[QUOTE=nullsquared;21947773]I have a few years of 3D programming experience, so I think I know what I'm talking about. Depending on your target machines, my statement is 100% correct. Your statement is only correct for integrated GPUs and insanely old non-integrated ones, like back in the NV FX5000 and ATI X800 range.[/QUOTE]
Not really. On standalone cards there's a noticeable difference when I turn down the 'model detail' and I see a noticeable improvement in frame time. The cards not a top of the line card, I forget it's exact model designation but it's an nVidia card that begins with 9. This is on quite a few games. So unless turning model detail down and LOD distance down does some other hippy magic for the mainstream user it will make a difference.
[QUOTE=Jawalt;21947960]Not really. On standalone cards there's a noticeable difference when I turn down the 'model detail' and I see a noticeable improvement in frame time. The cards not a top of the line card, I forget it's exact model designation but it's an nVidia card that begins with 9. This is on quite a few games. So unless turning model detail down and LOD distance down does some other hippy magic for the mainstream user it will make a difference.[/QUOTE]
The numbers after the 9 are very important. The other part is that current games may have "high quality" models at insane polycounts, which are then reduced to very low polycounts on lower settings. Vertex shaders running on each vertex will then impact the performance depending on how many calculations are done.
[editline]06:25PM[/editline]
For Xerios3, the batching is much more important and polycount is next to useless considering he only has about 300 triangles and it seems no complex vertex shader.
[QUOTE=nullsquared;21948024]The numbers after the 9 are very important. The other part is that current games may have "high quality" models at insane polycounts, which are then reduced to very low polycounts on lower settings. [B]Vertex shaders running on each vertex will then impact the performance depending on how many calculations are done.[/B]
[editline]06:25PM[/editline]
For Xerios3, the batching is much more important and polycount is next to useless considering he only has about 300 triangles and it seems no complex vertex shader.[/QUOTE] So wait you're telling me that the amount of polys rendered in a game matters? That's not what I said or anything is it?
[QUOTE=Jawalt;21948100]So wait you're telling me that the amount of polys rendered in a game matters? That's not what I said or anything is it?[/QUOTE]
The polycount alone itself does not mean anything. It's important what else is going on.
[QUOTE=nullsquared;21948219]The polycount alone itself does not mean anything. It's important what else is going on.[/QUOTE]
But obviously pushing polys is the performance drain. If I rendered no polys before displaying that buffer and clearing I'm sure my frametime would be much lower.
[editline]06:54PM[/editline]
And the other point is if he's not batching his polys in buffers or is batching too few polys it matters even more that his poly count be lower.
[img]http://i295.photobucket.com/albums/mm159/Siduron/sdnengine2-1.jpg[/img]
Scaling up my terrain and creating level of detail (not shown on picture).
Made a texture map and heightmap with L3DT so my terrain looks awesome by pressing a few buttons :v:.
So far i've completed 3 out of 8 indexbuffers for just one detail level and i'm getting bored of making them.
Meanwhile i'm struck by weird fps drops while turning the camera(even on low detail level on all patches).
No frustum culling is implemented so far so it could be that, but my expectation was that the fps would just be averagely bad instead of just when looking in a certain direction.
Maybe i'm just drawing too much triangles and i should get working on frustum culling, but if anyone has any idea about this please let me know.
Edit:
The problem seems to be my skysphere, it had so many triangles i could fold it out and make a second landscape with it in theory :v:.
[QUOTE=Jawalt;21947487]I don't think you know what you're talking about. On lower end/integrated cards which are the mainstream poly count is a huge factor.[/QUOTE]
You can draw a few hundred thousand polys in one draw call faster than 100 5 poly draw calls. Probably not them numbers but you get the point.
I'm sure you get some sick pleasure out of being called out on obvious bullshit :v:
[QUOTE=layla;21948862]You can draw a few hundred thousand polys in one draw call faster than 100 5 poly draw calls. Probably not them numbers but you get the point.[/QUOTE]
Yes I understand that. But the number of polys is still very important and a quite limiting factor. Saying poly count doesn't matter is bullshit.
I'm pretty sure nullsquared knows what he's talking about Jawalt.
[QUOTE=turb_;21949275]I'm pretty sure nullsquared knows what he's talking about Jawalt.[/QUOTE]
Yeah because nullsquared knows everything about everything and is pretty much god.
Poly count can effect FPS quite badly but is not the main drain in this situation. So they are both right.
[QUOTE=Jawalt;21949022]Yes I understand that. But the number of polys is still very important and a quite limiting factor. Saying poly count doesn't matter is bullshit.[/QUOTE]
Your argument is essentially, "if I draw nothing, then it will be faster than if I draw something."
Yes, that is obviously & trivially true.
However, like I said, verbatim: "Poly count doesn't really matter anymore, it's all about batching."
Rendering many meshes at once (usually by batching them into a single vertex buffer), aka instancing, will improve Xerios3's performance much more than reducing the polycount.
[editline]07:42PM[/editline]
[QUOTE=HTF;21949369]Yeah because nullsquared knows everything about everything and is pretty much god.[/QUOTE]
Nope. Because I've had my fair share of 3D programming
[img_thumb]http://dl.dropbox.com/u/2014606/desktop.png[/img_thumb]
made a very simple way of smoothing the image inside the program.
also this picture is right out of the program without any editing
Anyone doing [url=http://www.ludumdare.com/compo/]Ludum Dare 18[/url] this weekend? Or anyone done one before?
alright, i've come to the conclusion that a mach-o compatibility loader is too complex for me considering the time and effort required to get anything past proof of concept, which is unsurprising. i'm now going to shift my interest and effort towards a (m)dct-based audio codec that i have conceptualized.
[QUOTE=r4nk_;21950797]Anyone doing [url=http://www.ludumdare.com/compo/]Ludum Dare 18[/url] this weekend? Or anyone done one before?[/QUOTE]
pretty sure Zombuster has
Boolean plots :haw: (useful for inequalities)
[img]http://img101.imageshack.us/img101/1359/booleanplot1.png[/img]
Includes support for logical AND and OR in order to combine inequalities and what-not.
[editline]09:29PM[/editline]
I need to fix the info panel, it's not calculating the width right :raise:
Must be running out of features to implement by now null? :wink:
What about a feature where you can draw a graph with your mouse, and then the it generates a function for that? :smug:
Oh my God, that is the best idea yet.
[editline]02:08AM[/editline]
I'm not even joking.
[QUOTE=Chad Mobile;21952037]Oh my God, that is the best idea yet.
[editline]02:08AM[/editline]
I'm not even joking.[/QUOTE]
so amazing that it's incredibly impractical.
Why not add a table feature? Give it a x or y value or look at the table for them to see the points it plots.
Sorry, you need to Log In to post a reply to this thread.
