PDA

View Full Version : Performance issues with C4D in comparison with LW3D.



MauricioPC
04-11-2013, 08:44 AM
That's it has been lot's of talk about CineWare and the integration with After Effect we all know. I myself have been seeing the videos and enjoying. But I thought I could share a light
in some performance issues with C4D.

I've been playing around with LW trial and C4D Demo and both software are really nice. Yesterday Cinema4DTutorial.net released a website with some free tutorials. I went there and I saw a very nice tutorial of the new batmobile. So a just downloaded it (http://www.cinema4d.ws/?portfolio=tumbler-batmobile-modeling).

Happy to see the result, I pop up my C4D and open the scene file "tumbler". Well ... talk about getting nervous.

I do know my system spec is not the best there is, but I do think it's fast and capable. I'm running an AMD FX Octa-core 3.1GHz with 16Gigs of Corsair Dominator RAM and SLI Geforce GTX 550Ti (4Gigs of RAM and 384 cudas). But for my surprise I can't even move the viewport. Talk about frustration.

But before I kick my CPU, I decided to export the car as an OBJ and pop it up LW. Now ... for selecting the whole can with polys or points, it takes a while, I admit. But I can navigate smoothly in the viewport of Layout or Modeler. I can even open VPR and see realtime rendering.


Since I'm learning with my mind thinking that in the future I should by one for my freelance work, it's rather disappointing such a poor performance in C4D and very intriguing the good performance of LW.

Just to let you guys know before thinking that C4D with CineWare is the best in the world (performance wise). ;)

chikega
04-11-2013, 09:12 AM
Interesting. I had C4d for a while ... but finally let it go because I couldn't keep up with the exorbitant renewal fees. And if you missed a couple versions, you'd have to repurchase the software at full price. A page out of Autodesk.

mike_stening
04-11-2013, 09:20 AM
ive seen many issues between C4d and Lightwave, been using C4D for certain jobs (out of necessity rather than choice) and certain things it just can't match LW on, rendering glass and transparencies quickly is one of them. but on the flip side there are some things C4D does that i wish LW would

MauricioPC
04-11-2013, 09:26 AM
Interesting. I had C4d for a while ... but finally let it go because I couldn't keep up with the exorbitant renewal fees. And if you missed a couple versions, you'd have to repurchase the software at full price. A page out of Autodesk.

That's a real problem with C4D too. Since I purchased the student license, if I decide to go with the commercial version, I'll have a good discount on that. But for the performance, I haven't figure it out yet. Maybe the guy set the scene wrong, I don't know. I even downloaded the CINEBENCH, runned and thought I got some nice results. Go figure ...

MauricioPC
04-11-2013, 09:35 AM
ive seen many issues between C4d and Lightwave, been using C4D for certain jobs (out of necessity rather than choice) and certain things it just can't match LW on, rendering glass and transparencies quickly is one of them. but on the flip side there are some things C4D does that i wish LW would

Yes ... I know what you are saying. I think MoGraph it's the only reason for using C4D. But it's so cool. What I can't figure is why it's so slow in viewport performance. Of course today I make simple stuff, but the main goal is to improve and get more complex projects.

But from the apps I tested, I must say LW and C4D are the ones I feel more comfortable.

prometheus
04-11-2013, 09:42 AM
Arent you comparing wrong here, the main project file is what? cinema scene? and then you exported to obj?
In all fairness you should then use the exported obj format in cinema with exact polycount as the one used in lightwave, and this will only give you the obj...just questioning
the comparison if you had a cinema scene with all that comes with that.

So you are basicly comparing a cinema4d scene against just an obj, file in Lightwave..and that isnīt fair, If that is the case.

Michael

Surrealist.
04-11-2013, 09:47 AM
I would go to their website forum and post your issue. See if there is some system spec to change in C4D. For instance there are some settings in Modeler that will bring it to a crawl.

Just to be fair. I would make sure and check the situation out with the people most knowledgeable. It could be something pretty simple.

MauricioPC
04-11-2013, 10:03 AM
Arent you comparing wrong here, the main project file is what? cinema scene? and then you exported to obj?
In all fairness you should then use the exported obj format in cinema with exact polycount as the one used in lightwave, and this will only give you the obj...just questioning
the comparison if you had a cinema scene with all that comes with that.

So you are basicly comparing a cinema4d scene against just an obj, file in Lightwave..and that isnīt fair, If that is the case.

Michael

Yeah, I tried that. Even that it lagging a lot. I posted on CGSociety. Let's see if somebody answers it. All in all, thanks for the replies.

dee
04-11-2013, 10:52 AM
I've also downloaded the project file and don't have the problems you have. Tumbling and moving the car is fast. The I exported it as an obj an loaded into Modeler. Selected one poly on the front wheel and tryed to move - 2 seconds lag. Imported the obj in C4D and moved the same poly - instant and superfast.

Here's a video I've just recorded. You can see when mouse is clicked when the red/blue circles appear.
113498


System Information:
Operating System: Win 7 64Bit German
CPU Type: Intel Core I7-875, 3.2GHz
Number of CPUs: 1 QuadCore
RAM: 16GB
Graphics Card: NVIDIA GTX 660TI 2GB
Display Driver Version: 311.06
Lightwave 64bit

prometheus
04-11-2013, 10:55 AM
I've also downloaded the project file and don't have the problems you have. Tumbling and moving the car is fast. The I exported it as an obj an loaded into Modeler. Selected one poly on the front wheel and tryed to move - 2 seconds lag. Imported the obj in C4D and moved the same poly - instant and superfast.

Here's a video I've just recorded. You can see when mouse is clicked when the red/blue circles appear.
113498



System Information:
Operating System: Win 7 64Bit German
CPU Type: Intel Core I7-875, 3.2GHz
Number of CPUs: 1 QuadCore
RAM: 16GB
Graphics Card: NVIDIA GTX 660TI 2GB
Display Driver Version: 311.06
Lightwave 64bit

that is a more correct comparison, from what Ivé heard..cinema4d should be much better to handle huge polyamount data than lightwave.

MauricioPC
04-11-2013, 11:03 AM
I've also downloaded the project file and don't have the problems you have. Tumbling and moving the car is fast. The I exported it as an obj an loaded into Modeler. Selected one poly on the front wheel and tryed to move - 2 seconds lag. Imported the obj in C4D and moved the same poly - instant and superfast.

Here's a video I've just recorded. You can see when mouse is clicked when the red/blue circles appear.
113498


System Information:
Operating System: Win 7 64Bit German
CPU Type: Intel Core I7-875, 3.2GHz
Number of CPUs: 1 QuadCore
RAM: 16GB
Graphics Card: NVIDIA GTX 660TI 2GB
Display Driver Version: 311.06
Lightwave 64bit

I guess the differences in mine spec to your is mostly CPU and OS version. Your GPU is faster, but I do have 2 installed, so the differences should be minimal. Maybe I shall install Win7 and see how it handles.

jwiede
04-11-2013, 12:07 PM
I've also downloaded the project file and don't have the problems you have. Tumbling and moving the car is fast. The I exported it as an obj an loaded into Modeler. Selected one poly on the front wheel and tryed to move - 2 seconds lag. Imported the obj in C4D and moved the same poly - instant and superfast.

Here's a video I've just recorded. You can see when mouse is clicked when the red/blue circles appear.
113498


System Information:
Operating System: Win 7 64Bit German
CPU Type: Intel Core I7-875, 3.2GHz
Number of CPUs: 1 QuadCore
RAM: 16GB
Graphics Card: NVIDIA GTX 660TI 2GB
Display Driver Version: 311.06
Lightwave 64bit
Yep, that matches my experience as well, C4D is _much_ better at handling/manipulating megapoly+ scenes than LW here. If I run into problems working with a scene in LW or (pre-701) modo, my normal workflow is to switch to C4D precisely because it plows through high-density scenes where both LW and modo have issues (again, referring to pre-701 modo, with 701 modo's highpoly perf is pretty similar to C4D, IME).

The OP might want to also check their display viewport settings and/or the display preferences. It certainly is possible to configure C4D in such a way that it performs horribly (just turn off 3D hw acceleration, for example) as is true with nigh-all 3D apps. On Windows, also make sure to use the 64-bit C4D not the 32-bit build (both get installed automatically).

MauricioPC
04-11-2013, 12:23 PM
Yep, that matches my experience as well, C4D is _much_ better at handling/manipulating megapoly+ scenes than LW here. If I run into problems working with a scene in LW or (pre-701) modo, my normal workflow is to switch to C4D precisely because it plows through high-density scenes where both LW and modo have issues (again, referring to pre-701 modo, with 701 modo's highpoly perf is pretty similar to C4D, IME).

The OP might want to also check their display viewport settings and/or the display preferences. It certainly is possible to configure C4D in such a way that it performs horribly (just turn off 3D hw acceleration, for example) as is true with nigh-all 3D apps. On Windows, also make sure to use the 64-bit C4D not the 32-bit build (both get installed automatically).

Talk about looking like an idiot, right?

Just tried in my note and everything is fine. So there's something terribly wrong with my desktop. Damn.

Geforce GT 335M
i7 Q740 1.73GHz
6Gigs RAM
Win8 x64

Thanks for your help guys. I think it's time to format or throw it through the window.

MannaTheBerserk
04-11-2013, 05:43 PM
I've also downloaded the project file and don't have the problems you have. Tumbling and moving the car is fast. The I exported it as an obj an loaded into Modeler. Selected one poly on the front wheel and tryed to move - 2 seconds lag. Imported the obj in C4D and moved the same poly - instant and superfast.

Here's a video I've just recorded. You can see when mouse is clicked when the red/blue circles appear.
113498


System Information:
Operating System: Win 7 64Bit German
CPU Type: Intel Core I7-875, 3.2GHz
Number of CPUs: 1 QuadCore
RAM: 16GB
Graphics Card: NVIDIA GTX 660TI 2GB
Display Driver Version: 311.06
Lightwave 64bit

Why did you try the 32 Bit version of the modeler? I mean, in the video you made is clear that you are using Modeler 32 Bit.
You are basically using 3.5 GB RAM out of your 16 GB

dee
04-12-2013, 07:05 AM
Because it doesn't matter. Modeler32 occupies 500MB (vs 800 in 64bit) of RAM with this object loaded.

MannaTheBerserk
04-12-2013, 09:09 AM
Because it doesn't matter. Modeler32 occupies 500MB (vs 800 in 64bit) of RAM with this object loaded.

Yes, I see, but my point was about performance as well.

Unless Modeler 64Bit and Modeler 32Bit perform exactly the same, which I highly doubt, I can't help but see a pointless comparison between C4D and Modeler high-poly handling differences.

Of course I know that C4D is much better at doing this, but what if you try the same test with the 64 Bit version of Lightwave?

64 Bit is not all about using more RAM, but I am sure you know very well this, don't you?

I don't have C4D at home ( the company I work for has a license of Cinema 4D Studio R14, and it is amazing indeed ), so I cannot do the comparison by myself, and I don't have the obj file of the Tumbler.
This is why I am asking, I am curious about the difference in performance.

Cheers

M.

dee
04-12-2013, 09:37 AM
Trust me, there is no difference. I first loaded the model in Modeler64 but did the screencapture in Modeler32 because it's clean (no plugins) and was already setup for screencap rez. But just for you, here's the drama in Modeler64. :)

MannaTheBerserk
04-12-2013, 09:52 AM
Trust me, there is no difference. I first loaded the model in Modeler64 but did the screencapture in Modeler32 because it's clean (no plugins) and was already setup for screencap rez. But just for you, here's the drama in Modeler64. :)

Thank you for the time spent :)

So, basically NewTek is selling a 64 Bit version of a software that doesn't perform as a 64 Bit application at all, apart from the amount of RAM it is able to recognize... Curious, to say the least.

COBRASoft
04-12-2013, 09:58 AM
Can somebody share that obj file please? Would like to test it on my new machine :).

MannaTheBerserk
04-12-2013, 10:01 AM
Can somebody share that obj file please? Would like to test it on my new machine :).

Yep!
Agree.

Celshader
04-12-2013, 12:58 PM
Thank you for the time spent :)

So, basically NewTek is selling a 64 Bit version of a software that doesn't perform as a 64 Bit application at all, apart from the amount of RAM it is able to recognize... Curious, to say the least.

Now you are making me curious. With the same CPU specs, is there a software package that enjoys a significant performance increase between 32-bit and 64-bit?

I can work on larger projects with 64-bit LightWave, 64-bit RealFlow and 64-bit Photoshop, but I never noticed a dramatic performance difference between the 32-bit and 64-bit versions of those software packages.

hrgiger
04-12-2013, 02:07 PM
Yeah, I don't think performance is where RAM comes into play as much as the CPU and threading behind it. The 64 bit version simply allows more assets loaded into memory so larger scenes would be possible without turning to much slower virtual memory. But as far as the performance or speed of interaction or playback of a scene, that would remain for the most part unchanged between the 32 and 64 bit versions.

Nicolas Jordan
04-12-2013, 02:19 PM
I believe Lightwave was the first 64 bit 3D application. Rob Powers also demoed it back in the day. http://workstations.digitalmedianet.com/articles/viewarticle.jsp?id=32101

MannaTheBerserk
04-12-2013, 05:25 PM
Now you are making me curious. With the same CPU specs, is there a software package that enjoys a significant performance increase between 32-bit and 64-bit?

I can work on larger projects with 64-bit LightWave, 64-bit RealFlow and 64-bit Photoshop, but I never noticed a dramatic performance difference between the 32-bit and 64-bit versions of those software packages.


Well, the first releases of Mudbox, for example, were 32 Bit, like ZBrush is at the moment.

When it got 64 Bit not only it allowed you to work with more than double the polycount but it was much much faster in almost every task: subdividing the mesh, painting textures, exporting normal maps, and yes, even orbiting the camera and sculpting the geometry it self.

The 32 Bit version of Maya, especially the first Alias|Wavefront releases, is another example; when Maya started to take advantage of the 64 Bit architecture of the new processors and operating systems it literally transformed itself.
Much better overall performance along with visible greater efficiency in quite few areas: modeling, dynamics and even graphing bigger node networks. Not to speak of rendering of course ( but this is normal ).

The latest version of After Effects, now fully 64 Bit ( CS5 wasn't ), and I am talking about CS6, is a big step forward in performance: huge noticeable improvement in speed especially in presence of big amount of layers, where the previous version struggled.

Honestly, I have been working with Photoshop since 1995, and I have never noticed big performance improvements ( it has always been quite a powerful and fast software anyway ) until it became full 64 Bit. From 64 Bit on it received quite a boost, visible especially if you work with large CMYK high density ( up to 304 ppi, or 120 pp/cm if you wish ) and lots of layers and adjustment layers.

There are visible adavntages when a software is properly optimized to take advantage of your 64 Bit system, apart from the amount of RAM.

Now, I am not saying that Lightwave is bad, in fact I can tell exactly the opposite and I have recently purchased it ( but you know this... I suppose ).
I was just perplexed reading sentences like "It doesn't matter if you use 32 Bit or 64 Bit Lighwave because the performance is the same", when in my experience this shouldn't be.

That's really it :)

MannaTheBerserk
04-12-2013, 05:36 PM
Yeah, I don't think performance is where RAM comes into play as much as the CPU and threading behind it. The 64 bit version simply allows more assets loaded into memory so larger scenes would be possible without turning to much slower virtual memory. But as far as the performance or speed of interaction or playback of a scene, that would remain for the most part unchanged between the 32 and 64 bit versions.

64 Bit allows also the use of the 64 Bit instructions set of your 64 Bit processor, which obviously can address much more calculations and data per clock cycle.

Take a 32 Bit app on a 64 Bit architecture based computer and you are just wasting resources.

jboudreau
04-12-2013, 05:44 PM
That's it has been lot's of talk about CineWare and the integration with After Effect we all know. I myself have been seeing the videos and enjoying. But I thought I could share a light
in some performance issues with C4D.

I've been playing around with LW trial and C4D Demo and both software are really nice. Yesterday Cinema4DTutorial.net released a website with some free tutorials. I went there and I saw a very nice tutorial of the new batmobile. So a just downloaded it (http://www.cinema4d.ws/?portfolio=tumbler-batmobile-modeling).

Happy to see the result, I pop up my C4D and open the scene file "tumbler". Well ... talk about getting nervous.

I do know my system spec is not the best there is, but I do think it's fast and capable. I'm running an AMD FX Octa-core 3.1GHz with 16Gigs of Corsair Dominator RAM and SLI Geforce GTX 550Ti (4Gigs of RAM and 384 cudas). But for my surprise I can't even move the viewport. Talk about frustration.

But before I kick my CPU, I decided to export the car as an OBJ and pop it up LW. Now ... for selecting the whole can with polys or points, it takes a while, I admit. But I can navigate smoothly in the viewport of Layout or Modeler. I can even open VPR and see realtime rendering.


Since I'm learning with my mind thinking that in the future I should by one for my freelance work, it's rather disappointing such a poor performance in C4D and very intriguing the good performance of LW.

Just to let you guys know before thinking that C4D with CineWare is the best in the world (performance wise). ;)

Hi

How did you download the project files. The link on that site seems to be broken

Thanks,
Jason

jboudreau
04-12-2013, 05:54 PM
Can somebody share that obj file please? Would like to test it on my new machine :).

have any luck getting that obj file. I'd like to do the same.

Thanks,
Jason

COBRASoft
04-12-2013, 05:57 PM
Nope, haven't gotten it (yet) :(. Any1?

MannaTheBerserk
04-12-2013, 06:01 PM
Is this the same model perhaps?

LINK (http://www.3dcadbrowser.com/download.aspx?3dmodel=21393)

They just need one to upload a model in order to get the free ones.

Unfortunately I cannot give mine away, because they are work stuff.

dee
04-13-2013, 06:32 AM
Here you go: tumbler.zip (http://sdrv.ms/10SwACq)

MannaTheBerserk
04-13-2013, 06:36 AM
Here you go: tumbler.zip (http://sdrv.ms/10SwACq)

Thank you very much dee!

Very kind of you. :)

MannaTheBerserk
04-13-2013, 08:08 AM
Alright,

Before opening the obj in Lightwave I wanted to have a look at it inside Maya 2012, just to see what is going on with the model, the topology and the overall optimization of the mesh.

Well: almost a Million faces, and 1.7 million tris for just one car! :lol:

The result is: this is everything but a good job. I can see really bad booleans, which aren't the worst thing, and the main body of the car made of a decent amount of geometry ( meaning not too much and not too little ).

Then I have seen the wheels... Oh My Gosh! A freakin' minced of polygons.

One rear wheel counts "just" 140608 quad faces, and I mean just the tire, without taking into account the alloy, bolts and so on and so forth. This translates into 281216 tris faces for just one tire.

Well, Maya 2012 performance, with such a model, is pretty much on par with Lightwave Modeler one shown in your video.

Now, I wonder if inside C4D this model appears in a totally different way, like, for example, it would appear in Maya when you model properly and you use the "visual smooth" to make it pretty ( no need to raise the poly-count to the sky ), but when it gets exported in "obj" it also get subdivided in some way, giving us a high-poly model instead of a proper production ready optimized mesh.

That said: it still remains the fact that over a million polygons, actually towards almost 2 million polygons, both Maya and Lightwave drop in real time performance.

Next step: export from Maya and import in Lightwave and see what happens. I'll let you know! ;)

Lightwolf
04-13-2013, 08:23 AM
64 Bit allows also the use of the 64 Bit instructions set of your 64 Bit processor, which obviously can address much more calculations and data per clock cycle.
Nope, it can't. The 32-bit instruction set can already operate on 64-bit data types...
There are only two major differences: the amount of RAM that can be addressed and the amount of registers (think of these as temp storage for variables for fast access within the CPU) on the CPU.
That in turn implies a slight bloat in memory footprints (due to all address pointers using twice the amount of memory) and possible performance gains due to the registers.
However, it really depends... especially on how the data structures for the application are designed. And I suspect that LW doesn't do too well in that case.

Take a 32 Bit app on a 64 Bit architecture based computer and you are just wasting resources.
Not quite... for one, it can still be faster due to memory access issues (pointer size) ... and it can also profit from the OS itself possibly being a bit faster (as is the case on windows).

Cheers,
Mike

MannaTheBerserk
04-13-2013, 08:43 AM
Nope, it can't. The 32-bit instruction set can already operate on 64-bit data types...
There are only two major differences: the amount of RAM that can be addressed and the amount of registers (think of these as temp storage for variables for fast access within the CPU) on the CPU.
That in turn implies a slight bloat in memory footprints (due to all address pointers using twice the amount of memory) and possible performance gains due to the registers.
However, it really depends... especially on how the data structures for the application are designed. And I suspect that LW doesn't do too well in that case.

Not quite... for one, it can still be faster due to memory access issues (pointer size) ... and it can also profit from the OS itself possibly being a bit faster (as is the case on windows).

Cheers,
Mike

Well, that's in contrast with things like Mudbox 64 Bit being in facts much much much much faster than ZBrush, which is 32 Bit, tested in the same situation, same model but different subdivision level.
Actually, ZBrush cannot even reach the subd level that Mudbox easily can ( try to work with more than the 6th division level with both applications and see what I mean ).
In this scenario Mudbox destroys ZBrush with really no mercy at all.
Oh, not to speak of the presence of many "photoshop compatible" layers ( both sculpt layers and paint layers of course ) and all the viewport filters active ( Tonemapper, Depth of Field, Cavity AO, AO... ).

How do you explain this?

And what about After Effects CS6, just to take in consideration a "non 3D" software, for example? C6 literally outperforms CS5, hands down ( same system, same processors, same amount of RAM. In my case a Z800, dual Xeon, 24 GB RAM and a couple of Quadro 4000 ).

So, yeah, I am with you on the paper, but when you begin to dive into the actual facts things are quite different.

These, to be honest, are things beyond my comprehension.
Especially after reading what you just wrote, I don't get it.

Could you explain what actually happen in these cases?

Lightwolf
04-13-2013, 08:48 AM
Well, that's in contrast with things like Mudbox 64 Bit being in facts much much much much faster than ZBrush, which is 32 Bit, tested in the same situation, same model but different subdivision level.
The problem with this comparison is that you're comparing much more than 32-bit vs. 64-bit. The only valid comparison is the exact same code base on the same machine.
Anything else is guessing... and in this case even comparing two entirely different approaches to sculpting.


And what about After Effects CS6, just to take in consideration a "non 3D" software, for example? C6 literally outperforms CS5, hands down ( same system, same processors, same amount of RAM. In my case a Z800, dual Xeon, 24 GB RAM and a couple of Quadro 4000 ).
This is almost the same case though... have you compared 32-bit s. 64-bit of the same version application? Obviously making sure that the 32-bit memory limit doesn't come into play.
If I remember correctly (I don't use AE), CS6 had quite a few performance related tweaks.


Could you explain what actually happen in these cases?
Yes, you're comparing apples to oranges. ;)

Cheers,
Mike

MannaTheBerserk
04-13-2013, 08:59 AM
The problem with this comparison is that you're comparing much more than 32-bit vs. 64-bit. The only valid comparison is the exact same code base on the same machine.
Anything else is guessing... and in this case even comparing two entirely different approaches to sculpting.

This is almost the same case though... have you compared 32-bit s. 64-bit of the same version application? Obviously making sure that the 32-bit memory limit doesn't come into play.
If I remember correctly (I don't use AE), CS6 had quite a few performance related tweaks.

Yes, you're comparing apples to oranges. ;)

Cheers,
Mike


I could have compared Maya 2010 32 Bit with Maya 2010 64 Bit, the result is pretty much the same: you can almost forget any complex particle/dynamic simulation, and even simple ones perform slower than their counterpart on the 64 Bit app.
Not only simulation though, there also is the interactive speed when modeling with nurbs and subdivision surfaces: 32 bit, almost decent but limited, 64 Bit much better ( and I can even activate the fine setting for the nurbs surface precision calculations ).
Hair solver is faster in 64 Bit.
One "funny" thing: even popping in and out the viewports ( by tapping the space bar quickly, if you know what I mean ) is more responsive in the 64 Bit Maya.

I am not saying the 32 Bit is bad of course, far from it and it's actually fantastic, I am just saying that the other is a much better experience.

Now, this is apples Vs apples.

Again, I am quite ignorant in this technical field, so I judge from what I see and get, not from numbers.

MannaTheBerserk
04-13-2013, 09:18 AM
Ok,

I opened the Tumbler in Modeler after having saved it in FBX from Maya.

Modeler behaves with the same performance as Maya does, but there are a couple of differences I noticed:

1) Modeler is slightly faster when I move faces ( just a tad, but noticeable ).

2) Navigating in the viewport, orbiting around the object, zooming in and out is super-fast in both applications. No problems at all here.

3) Here comes one issue: as soon as I move one polyface in Modeler and then I unselect everything, suddenly the overall performance drops drastically, even when just moving the camera around the objects the framerate get much much slower.
If I close the application and open the model again, everything gets back to normal, navigation wise.

Now I am confused.

Lightwolf
04-13-2013, 09:45 AM
I could have compared Maya 2010 32 Bit with Maya 2010 64 Bit, the result is pretty much the same: you can almost forget any complex particle/dynamic simulation, and even simple ones perform slower than their counterpart on the 64 Bit app.
There you go then... it seem like Maya profits a lot from the extra register then.

Again, I am quite ignorant in this technical field, so I judge from what I see and get, not from numbers.
The problem with the previous examples was that you were drawing conclusions from your observations that weren't correct (in terms of what happens technically).

Other than that.. whatever works best for you, no problems with that.

Cheers,
Mike

MannaTheBerserk
04-13-2013, 09:49 AM
There you go then... it seem like Maya profits a lot from the extra register then.

The problem with the previous examples was that you were drawing conclusions from your observations that weren't correct (in terms of what happens technically).

Other than that.. whatever works best for you, no problems with that.

Cheers,
Mike

Yes, you're right.
Also, I usually do not install 32 Bit versions of software that I have the 64 Bit one ( don't want to waste disc space ).
As soon as I have some spare time I will definitely do a test between two identical applications ( apart from the 32/64Bit of course ).

Now, back to Lightwave tutorials ( I need to get back in shape ASAP :D ).

dee
04-14-2013, 04:07 AM
And here's the new Tweak tool in action. It takes about 4-5 seconds to cache the 916764 polys but after that you can really work with it. Mouse movement still feels core'ish (like operating in jelly) but a nice glimpse of Modeler's future. :)

MannaTheBerserk
04-14-2013, 04:47 AM
And here's the new Tweak tool in action. It takes about 4-5 seconds to cache the 916764 polys but after that you can really work with it. Mouse movement still feels core'ish (like operating in jelly) but a nice glimpse of Modeler's future. :)

This is actually impressive!

few seconds wait is definitely worth :)

Photogram
04-15-2013, 01:05 PM
This a great improvement!!!
This is a good thing for future big projects :)

I recently worked on an architectural visualisation. The file was huge poly count don't exactly remember but i know it was multi-millions polys.
The model was done with Revit so my guess was to convert it directly to FBX, the file never open in Lightwave, Modo or C4D... I've been able to open it in Maya and Max... but the file was very messy with so many unwanted layers and geometry.. so my guess was to find a way to do the cleanup with Max or Maya to delete those geometry and extra Layers.

First impressions Maya was a lot more fluid than Max but it was more easy to do the cleanup in Max as it present the model more clear in the layers and more easy to understand whats happen there..

So after the cleanup i exported again as fbx and this time Lightwave and C4D has open the model with every layers but Modo failed to open it.

The speed i was able to interact with the model in the Layout was so much faster than any other apps there.... don't know why but my idea is because the Lightwave layout is not a modeler and i think the layout dont have to manage many buffers relative to the manipulations of the components of the meshes (vertices, polygons, edges etc...)

By curiosity i exported my model from Lightwave as fbx and open it in other softwares and compare the speed and ease of manipulation and Layout was the faster for animation with huge model. The Modeler ran at a similar speed as Maya. Maya is faster than Max and C4D is the slower one... i then tried the bounding box threshold equivalent and that was more fast but i don't like the way C4D do the display.
Modo finally open it and was good at speed but has crashed after few manipulations...

So the work was done with Lightwave... The only speed problem i encountered was due to the huge amount of memory that model take.
When modeling only Modeler was running fine... The day i began to build my scene in the Layout things have become more difficult since my model was loaded twice in memory... For the Layout and for the modeler and everytime i had to go back and forth between the Modeler and Layout i had to be patient and pray to not crash even with 32 megs of RAM...

So i decided to run the Layout and Modeler in hubless mode because the hub had too much difficulty to update any small modifications..
This working method is more stable but i had to keep more versions of my modifications in modeler in case i wish to undo changes.

Now i think a unified Layout and Modeler may have the advantage of using less memory but may reduce speed for animating huge models.

Photogram
04-16-2013, 08:54 AM
i was meaning 32 gigs of RAM...

Photogram
04-19-2013, 11:14 AM
All above explanations where based on the experience i had while doing that project. Now months later i decide to do similar parformance test again with some of these software and this time Modo and C4D have evolve and their performance are better...

For the test i exported my Lightwave Objects as .lwo and .fbx and the open each with modo 701, c4d r14.

Modo performance is now faster than our modeler and C4D is the faster one now...

One thing i like a lot in C4D is their kind of navigation. So easy to go where you want because you simply have to click on any part of any geometry to zoom or orbit arround that location.. I think this is more intuitive and it is faster to navigate accross the model.

I wish now Newtek modernise the navigation in modeler. Core had these new way of navigating and i miss that...

prometheus
04-19-2013, 11:36 AM
All above explanations where based on the experience i had while doing that project. Now months later i decide to do similar parformance test again with some of these software and this time Modo and C4D have evolve and their performance are better...

For the test i exported my Lightwave Objects as .lwo and .fbx and the open each with modo 701, c4d r14.

Modo performance is now faster than our modeler and C4D is the faster one now...

One thing i like a lot in C4D is their kind of navigation. So easy to go where you want because you simply have to click on any part of any geometry to zoom or orbit arround that location.. I think this is more intuitive and it is faster to navigate accross the model.

I wish now Newtek modernise the navigation in modeler. Core had these new way of navigating and i miss that...

You could go to the edit shortcut menu and search for fit selected, assign it to whatever shortcut you want and when you select a poly and hit the shortcut it will zoom to that area and also
keep orbit rotation with that selection in center...not automaticly but works quite decent.
I prefer to set it to the z-key since I rarely use redo as shortcut and rarely do that anyway..so z refererences my brain to zoom....sort of:)

Damn.. my laptop has a poor response when I am writing, so apart from me being tired,dyslectic(not), screwing up the english...Im not responsible for my spelling which has
been inserting a lot of errors lately in my writings on some threads:)

Photogram
04-19-2013, 11:56 AM
You could go to the edit shortcut menu and search for fit selected

This is i ever do but i discovered that other navigation and it is very relaxing... less button to click.. you can do all this only with the mouse.
This wish is for layout too. while working if i want this behavior i need to use a camera and a target. The C4D navigation is like the target is dynamic and change to everywhere you click.. I like that!

prometheus
04-19-2013, 12:02 PM
sure...thereīs room for navigation enhancement, no argue about that.

Michael

jwiede
04-20-2013, 04:33 AM
One thing i like a lot in C4D is their kind of navigation. So easy to go where you want because you simply have to click on any part of any geometry to zoom or orbit arround that location.. I think this is more intuitive and it is faster to navigate accross the model.
The new navigation options in C4D are subtle, but I find they make a huge improvement in workflow. Having the ability to continually "recenter" viewport operations on the fly just using mouse clicks allows for much easier navigation, esp. when extremely zoomed into a complex model/scene. They allow you to move around in ways that require substantially more viewport manipulation to achieve in other apps. I'd really like to see modo & LW incorporate similar viewport navigation approaches as C4D.

lardbros
04-20-2013, 05:11 AM
Thank you for the time spent :)

So, basically NewTek is selling a 64 Bit version of a software that doesn't perform as a 64 Bit application at all, apart from the amount of RAM it is able to recognize... Curious, to say the least.

As far as I've always been aware, 64-bit versions of apps don't 'perform' better than 32-bit, just can access more than 3GB of RAM. That's a pretty massive performance increase if you ask me :D

Sorry... should have read the entire thread before responding :D seems Mike has given you the more precise answer.


To add to the discussion though... when Autodesk brought the first 64-bit version of 3dsmax out, it was much, much slower than the 32-bit. Mainly the material editor for some reason... it was PAAAAINFUL! Guess they figured that out in the end, I noticed no difference in performance now though.
3dsMax's Nitrous viewport is impressive though. We were throwing around 14million polys the other day, with Ambient Occlusion, and shadows on and it wasn't painful at all. There was a slight pause between moving the 3dConnexion device and the model moving, but that's kind of to be expected :D

Rayek
04-20-2013, 01:24 PM
aargh, double post??

Rayek
04-20-2013, 02:15 PM
Since everyone is comparing apples and pears, I am throwing Blender into the mix! ;-)

Performance of the tumbler object while editing and animating: absolutely silky smooth. Here's the reason why: in Blender this object is imported as a collection of sub-objects, so each part can be selected individually and edited in edit mode. There is no CPU overhead (or hardly any) for any of these objects in edit mode (including the wheels).

It means that when I duplicate the tumbler seven times, even with 8 tumblers the viewport remains extremely responsive, and editing any of the meshes is direct and real-time.

In Modeler the entire object must remain in "edit mode", which causes an obvious slow down. This is one of the caveats of having a separated modeler<->layout environment: to construct this tumbler, the entire thing must be editable as a whole in Modeler, and that taxes the cpu.

Here's a pic of me editing the wheels of one eleven tumblers in the scene. Still very responsive and realtime editing. :-) When I zoom in completely smooth. And let's not forget the great isolation mode in Blender (which allows us to work on a single isolated object, and blend out all the others in the scene - similar to the LW Modeler environment)

113757

Conclusions:
- this object is an optimum one for Blender. Excellent performance - could not be better.
- comparing apples and pears really does not explain anything. :-) Merely that they taste different.

Rayek
04-20-2013, 02:31 PM
One thing i like a lot in C4D is their kind of navigation. So easy to go where you want because you simply have to click on any part of any geometry to zoom or orbit arround that location.. I think this is more intuitive and it is faster to navigate accross the model.


Yes, Maxon took a good look at Blender's viewport navigation. :devil:

Sorry - just playing the devil's advocate :D

jwiede
04-22-2013, 07:22 PM
Yes, Maxon took a good look at Blender's viewport navigation. :devil:

Sorry - just playing the devil's advocate :D
As of when? You do realize most of this functionality has been in C4D for a while now, right?

Lightwolf
04-22-2013, 07:28 PM
One thing i like a lot in C4D is their kind of navigation. So easy to go where you want because you simply have to click on any part of any geometry to zoom or orbit arround that location.. I think this is more intuitive and it is faster to navigate accross the model.
"g" in Modeler. And if you don't have one hand on the keyboard when using LW, then it's not the right tool for you. ;)

Cheers,
Mike

Rayek
04-22-2013, 09:10 PM
As of when? You do realize most of this functionality has been in C4D for a while now, right?

Yes, I know - I use an older version of C4d myself. I was in a sarcastic mood - apologies. Just getting exhausted of all the app comparisons being discussed lately. Use whatever works best for you, I say.

MannaTheBerserk
04-23-2013, 04:04 AM
Since everyone is comparing apples and pears, I am throwing Blender into the mix! ;-)

Performance of the tumbler object while editing and animating: absolutely silky smooth. Here's the reason why: in Blender this object is imported as a collection of sub-objects, so each part can be selected individually and edited in edit mode. There is no CPU overhead (or hardly any) for any of these objects in edit mode (including the wheels).

It means that when I duplicate the tumbler seven times, even with 8 tumblers the viewport remains extremely responsive, and editing any of the meshes is direct and real-time.

In Modeler the entire object must remain in "edit mode", which causes an obvious slow down. This is one of the caveats of having a separated modeler<->layout environment: to construct this tumbler, the entire thing must be editable as a whole in Modeler, and that taxes the cpu.

Here's a pic of me editing the wheels of one eleven tumblers in the scene. Still very responsive and realtime editing. :-) When I zoom in completely smooth. And let's not forget the great isolation mode in Blender (which allows us to work on a single isolated object, and blend out all the others in the scene - similar to the LW Modeler environment)

113757

Conclusions:
- this object is an optimum one for Blender. Excellent performance - could not be better.
- comparing apples and pears really does not explain anything. :-) Merely that they taste different.

I don't think this has much to do with the object being imported as sub-object honestly.

In Maya the same thing happens: when I "Separate" the object and I remain with a bunch of "sub-objects" to deal with, performance is always the same as soon as I dive into "component mode".

imashination
04-23-2013, 08:50 AM
Just thought I'd pop my head in and say that the original file has some serious performance issues. Firstly, It has 40 character rigs hidden in the project, with over 100,000 keyframes attached to them, this, as I imagine it would in any 3D app, makes it crawl along. Secondly, the entire mesh shoudlnt be left in a symmetry generator, especially when there are that many components. The symmetry generator is really for making character models and such, flipping over thousands of components, textures, expressions and such will have quite an impact.

If I remove the 40 rigs and get rid of the symmetry, the file becomes very responsive, orbiting the object gives me 300 fps (i7 @3.5GHz, 12 gigs, GF680, win7, 2560x1600)

If the project is exported from c4d as an old obj file which doesnt support these items, this will strip all the rigs out and flatten down the symmetry generator. Even just re-importing this project back into c4d will give a huge speed boost.

*Edit Though there is also something else odd about the file, as simply copying and pasting the model into a new project boosts this to over 300 fps. I still havent managed to put my finger on it yet

*edit 2 - Found it, there were millions more keyframes hidden inside unused NLA motion clips.

Rayek
04-23-2013, 10:07 AM
I don't think this has much to do with the object being imported as sub-object honestly.

In Maya the same thing happens: when I "Separate" the object and I remain with a bunch of "sub-objects" to deal with, performance is always the same as soon as I dive into "component mode".

That was the point I was trying to make as well? In maya, max, c4d, blender and so on, this is something we would do naturally when building an object like this.

When I join all the Tumbler's components into one mesh, I get 3fps in edit mode, and editing becomes a bit of a pain.

With the object divided in components, I get a steady 70fps. Editing one of the wheels does impact the viewport speed, since the cpu is the limiting factor, not the gpu, but still very responsive for realtime editing.

I am sure Maya/max (and especially SI) will do better in this regard, as Blender's opengl viewport is generally slower and less sophisticated than AD's products.

In modeler v11 when I load this object the viewport navigation is good. But when I start pulling a couple of polygons it updates at less than 1 fps - which is understandable, since Modeller must deal with all the polys at once. Same happens in Blender with the joined object (although Blender seems to outperform Modeler by a factor of about two/three when in full object editing mode) for editing actions.

Again, I do not feel we can truly compare all the different apps, since each one has its own strengths and weaknesses as far as viewport performance and navigation goes, and the type of model. Though in this particular case the division between Modeler and Layout does force us to work around Modeler's slowness. Cutting the object up in individual objects in Layout is a bit inconvenient when building it - harder to fit all the pieces exactly. Luckily we can hide and show layers in Modeler, which allows us to hide the wheels for improved editing performance.

Tip: before loading the Tumbler into Modeler, open the general options and uncheck "OBJ One Layer". Otherwise the model loads up in one huge layer.
btw, the labeling is terrible in this file.