Ogre VS IRRLICH VS PANDA 3D

[Anon]

Hey welcome to the forums!

Get the biased replies from the other communities. Hilarity ensues.

Obviously all three forums are biased, and all three think they are the only right one.

Not all members from all of the forums are biased. Rdb said

Get the biased replies

which doesnt mean all the replies posted.

Well, I also wrote ogre in lower case, and left the 3D off the end of both, so I was equally insulting them I guess.

You can't be serious...

From the link posted seems like the panda folks have a thing for real life pandas

ogres

[Gogg]

I've spent the last hours doing a common interface to Irrlicht and Panda3D in order to run exactly the same scenes and run some benchmarks, since OP asked me about it. Now, the first one I did is a little silly, just a thousand static balls in a scene, textured, with trilinear filtering; but I'm tired so I'll post this for now. I didn't include Ogre3D because it's a little more complicated to setup and my goal for today was just comparing Panda to one of the specialized engines, I'm pretty sure Irrlicht is approximately as fast as Ogre, and all I wanted to see was how would Panda3D perform at rendering since it does many more things. I plan to include Ogre it in the future, though.

Check the FPS and API's used in the window titles. (Irrlicht reports the FPS as an integer, I'm not rounding.)



The OpenGL results are surprising, but this is Windows, and the drivers are pretty bad, so it could mean anything. The DX results are much more like what I expected. For such a static scene I didn't feel the need to extract an average over time, the FPS don't vary at all. Also, the texturing is slightly different even though I'm specifying trilinear filtering in both engines, but that kind of differences are to be expected. Panda3D is rendering more balls than Irrlicht, because the aspect ratio handling code is also slightly different in both engines, so take that into account too, I'll try to adjust the fov better next time. All in all, I wouldn't recommend taking a decision based on these results, but I think it's good enough to conclude that Panda3D isn't bloated by all of its features (which is what I wanted to check). Consider that even though Disney does very low tech games for kids, these games still have to run in really crappy integrated VGAs, so it makes sense that it's highly optimized in despite of the low technical profile.

System: Nvidia 8600GT/Q6600/Windows 7

I'll share this common interface after I add Ogre3D to it, so that people can do new benchmarks.

[rdb]

Obviously all three forums are biased, and all three think they are the only right one.

Of course everyone is biased, that's why it makes little sense to ask which engine is better here. I just found it funny that on the Ogre3D forums, most people are somewhat more biased and a less serious than here (they just shout 'ogre' without arguments to support that), where you get serious replies from people that have tried other engines.

But what am I doing? Sorry for the offtopic post

[kojack]

Of course everyone is biased, that's why it makes little sense to ask which engine is better here. I just found it funny that on the Ogre3D forums, most people are somewhat more biased and a less serious than here (they just shout 'ogre' without arguments to support that), where you get serious replies from people that have tried other engines.

Well, we've had so many threads about ogre vs irrlicht, and none of them ever have a useful result except that whichever engine forum you ask on says to use theirs, so it's hard to keep serious.

Not all members from all of the forums are biased.

True, I didn't mean to imply that no one here (or on the other 2 forums) can give unbiased responses. It's just forums as a whole that tend to favour their own product.

You can't be serious...

I did put on the end.
I'm almost never serious.

From the link posted seems like the panda folks have a thing for real life pandas

ogres

Hmm, I didn't notice an over abundance of Pandas on here.
We once had a person from an ogre/orc fansite hang around trying to start conversations about how cool orcs and similar things were, until eventually they realised it was just the name and mascot mesh, none of us really wanted to pretend we were ogres.

Ok, sorry, going a bit off topic now.

[texugo]

You all right kojak. Lemme a better explain.

Ogre has a (good) architecture. But your game/graphic project must be placed inside it. When you planning your product you must planning inside this rationallity. Sometimes this is good, sometimes this is bad.

If you got a great project, this is good. But if you got a little one, this is not so good at all. Besides high customization of ogre system is not so simple alter the managers behaviors if you need.

Compare the work you have to make Robot Boxing in PANDA and the same work in OGRE (or PYOGRE). What's the diference?

In OGRE you got a lot of things to do. Initialize managers, resources, compositions, Listeners, etc. In PANDA just put the robots, define events and run! The keyword is flexibility!

PANDA's architecture is the most flexible i've ever seem. You planning your project as you like and PANDA connect with your project ... not my project must connect with PANDA. No matter your programming style, your planning style and if are a great project or home brew project. If you're master in design patterns or not.

PANDA changes for you. You must change to OGRE. This is my point of view and experience with theses systems.

Off course please ... the fact i dislike OGRE's architecture doesn't mean OGRE is worst. People must choose what they feel better. I choose PANDA ... if someone choose OGRE or IRRLICHT ... well ... the world is a great place i said



Regards

[Praios]

My reason for using panda instead of orge, irrlicht, crystalspace, ... is that for me installing and starting right away did only work with panda and only for panda i found some usable python-tutorial (was some time ago - so maybe things changed).
There are other important aspects of course - but being able to start without knowing much was most important to me.

[FenrirWolf]

Hmm, I didn't notice an over abundance of Pandas on here.
We once had a person from an ogre/orc fansite hang around trying to start conversations about how cool orcs and similar things were, until eventually they realised it was just the name and mascot mesh, none of us really wanted to pretend we were ogres.

Haha, that's funny. Takes all kinds, eh?

I've personally never noticed panda fanatics on here; Heck nobody has even mentioned ever trying to save the pandas as far as I can recall.

Gogg, good work, though I don't think it's going to really show much of a difference. Unless Ogre/Irrlicht/Panda3d does something *really* stupid when building/culling a list of renderable geometry (such as not using vertex buffers or failing to batch primitives), they should give about the same results on the same hardware -- you're going to be more limited by how fast the card can accept triangles than anything.

I never approached what engine I wanted to use based on speed; I felt they were all equivalent. I was more concerned about ease of use, and ease of asset import. I also wanted an engine based on modern 3D graphics ideas, and not have to screw around with things like BSP-based collision detection or hacked-on shader support. (I use Torque, and while the engine itself it's great, it's filled with a bunch of incredibly arcane rendering code, straight from the Voodoo1 days.)

[Gogg]

Gogg, good work, though I don't think it's going to really show much of a difference. Unless Ogre/Irrlicht/Panda3d does something *really* stupid when building/culling a list of renderable geometry (such as not using vertex buffers or failing to batch primitives), they should give about the same results on the same hardware -- you're going to be more limited by how fast the card can accept triangles than anything.

You are right, but that was exactly my intention, making sure that nothing weird is going on inside Panda. And maybe you didn't notice but there's a noticeable difference in OpenGL, though. So maybe it's not that silly after all. Sure, the OpenGL drivers on Windows suck, but the fact still remains that one scores double than the other on what is a pretty standard card with the drivers that everybody is going to have installed.

[emekapa]

Gogg many many thanks for you patience! I would like to see an head to head with ogre but i understand that it gives so many work!

You guys have been a light, i've already eliminated irrlicht, now i just have to decide between ogre and panda and the performance will make the diference! Thank you all

[FenrirWolf]

And maybe you didn't notice but there's a noticeable difference in OpenGL, though. So maybe it's not that silly after all. Sure, the OpenGL drivers on Windows suck, but the fact still remains that one scores double than the other on what is a pretty standard card with the drivers that everybody is going to have installed.

I see this all the time; OpenGL under Windows, as you said, is pretty goofy with all the driver issues. A different set of client states will result in wildly different kinds of OGL performance. That 70-ish FPS is suspiciously like a multiple of the 30-ish FPS -- maybe it's ignoring vertical retrace or triple/double buffering client state requests.

I wouldn't necessarily claim it's Panda3d being wildly more efficient in OGL than Irrlicht, especially since the DX3D tests are more even.

[Gogg]

That 70-ish FPS is suspiciously like a multiple of the 30-ish FPS -- maybe it's ignoring vertical retrace or triple/double buffering client state requests.

About the Y=2X relationship, I don't think that's relevant, because that would imply that geometry calculations are taking near to 0 time and I bet it's taking a good chunk of the loop iteration time. However, you get a point that something is probably buggy, but, isn't avoiding that kind of stuff part of being more efficient? Any engine's code is full of workarounds for buggy cards/setups.

I wouldn't necessarily claim it's Panda3d being wildly more efficient in OGL than Irrlicht, especially since the DX3D tests are more even.

Correct, but I don't agree that you should correlate the apis of each engine like that. Most of the inefficiency could be found in the renderers.
What it means, IMHO, is that either Irrlicht is doing something wrong at OpenGL or Panda is doing something wrong at D3D.

[bigfoot29]

Hey folks.

Just my 2 cents...

I tried all 3 of these engines - plus a couple of commercial other ones quite a few years ago. After my Java/C(++) Classes with good grades I ended up using Panda3D - learning Python. Why should I follow a fixed style on how to set up my program? Why should I twist my mind the way the devs want? Its the mind blowing simplicity that Panda3D and Python provide that convinced me in the end.

-> I wrote a simple chatserver (telnet-style) in "pure" C(++)-Code which took me a week (as first "bigger" project) and was/were good 3 pages of code that you need to scroll to the right for another page because of code complexity and usage of proper formatting (which I wanted to have). I was a bit disappointed that such a small task (binding to a socket, validating input/output, mirroring the input to the connected clients) created such intense headaches. - I stopped programming for quite some time. - Then I stumbled upon Panda3D and heard the first time of Python. I did some basic tutorials and wanted to write the same chat server again. And thats what I did. Results were a far more clean (readable) code in ~1.5 pages of code (in both cases the "tiny"-texteditor on a classic non-X linux screen).

Thats why I would tend to use Panda3D. Its awesome for Python'ers.

But of course we are all biased. We wouldn't answer HERE if we weren't no Panda3D-fans.

Gogg, nice work btw.

Regards, Bigfoot29

[astelix]

hey don't forget the upcoming p3d web plugin - does somebody knows if ogre or irrlicht may feature this?

[bigfoot29]

I would call it "currently not available there" - same as "here" in the official build.

But yea... I can't wait to see it "live and in action" once 1.7 is out. ^^

Regards, Bigfoot29

[croxis]

Hey Gogg, would it be possible to get a copy of this code so I can test it in linux? I am just interested in Panda vs Irrlicht. Thanks!

[deganderson]

I'll share this common interface after I add Ogre3D to it, so that people can do new benchmarks.

Did you manage to finish this Gogg? i have to say im very intrigued by this

[Mutos]

Hi all,


In fact Panda3D dis attract me because of its simplicity. In just 1-2h of work, while learning a new language, I was able to get a small but clean and satisfying result. And everything I see, find or read for now tell me I'll be able to expand it easily into the larger and more structured code I need...

In many other engines I tested, I never got this far because of all the stuff that had to be setup before you render your first mesh. That was what originally attracted me in TV3D for my first stellar system editor project, and believe me, Panda3D is even simpler and Python is a breeze !

Maybe it has drawbacks on other areas. But for what I want to do, it is a fine resource...

[emekapa]

Have anyone managed to finish the tests like goggs did? Or someone have done tests between ogre and panda 3D to make a comparison in terms of results ? It would help me a lot : )


Regards