PDA

View Full Version : Why is Lightwave OpenGL that slow on the Mac ?



Radamanthys
07-27-2005, 11:19 AM
Hi,

I have the two following systems:

PC : Athlon 64 3500+, 2 Go RAM, windows xp sp2, Geforce 6800 GT, 2x20" display (at 1600x1200)
Mac: PowerMac G5 dual 2 GHz, 2,5 Go RAM, OS X 10.4.2, Geforce 6800 GT, 2x20" display (at 1600x1200) (just received)

I just compared the display speed on these two machines, and to my big disapointment, the Mac is alot slower in displaying OpenGL in Lightwave :( (when i say alot its realy alot, i would say 3 to 5 times slower)

has someone idea if i could to something to improve this or its it a known problem on the Mac ?

eblu
07-27-2005, 01:01 PM
Hi,

has someone idea if i could to something to improve this or its it a known problem on the Mac ?


Known problem?
lets make a metaphor!

openGl on the mac is like shoe laces in a new pair of shoes.
wait for it, wait for it...

new shoe laces need to be tightened up, and then a little later when the shoe gets broken in, the shoe laces need to be adjusted again and again till they are just right. OpenGl is relatively new on the mac. To you and I and the speed of business, its been a Darned long time since OpenGl shipped on the mac. But to the trends that were mapped out in the way drivers mature, its been a very short period of time. Windows has been maturing for a very long time, and its going to take a bit of time for OS X to catch up.

the good news is Apple is acutely aware of this, and are getting behind the things that will shorten that wait. OpenGl is the Only hardware acceleration Apple uses, Apple features OpenGl in multiple parts of the OS least of which is Quartz Extreme. Apple does these things in order to speed up the development of OpenGl, drivers, and associated software products on the mac. They freely give away all of their openGl research.

Apple did some spelunking in their history, and determined that it took 10 to 15 years to develop Mature drivers for Quickdraw. They are determined, irritated, and aggressively trying to shorten that time period, and I have been impressed with the pace they have been keeping up for the last 5 years.

Now having said that, I am sure there are many many many places where Lightwave is Processor dependent, or the opengl code is not necessarily the fastest it could be. Your 2 20 inch monitors, may very well be completely sucking down the availible OpenGl resources, or you may not have enough ram. Also Lightwave is not one of the fastest 3d application available for the mac. Its one of the most useful.

Radamanthys
07-27-2005, 02:03 PM
Known problem?
lets make a metaphor!

openGl on the mac is like shoe laces in a new pair of shoes.
wait for it, wait for it...

new shoe laces need to be tightened up, and then a little later when the shoe gets broken in, the shoe laces need to be adjusted again and again till they are just right. OpenGl is relatively new on the mac. To you and I and the speed of business, its been a Darned long time since OpenGl shipped on the mac. But to the trends that were mapped out in the way drivers mature, its been a very short period of time. Windows has been maturing for a very long time, and its going to take a bit of time for OS X to catch up.

the good news is Apple is acutely aware of this, and are getting behind the things that will shorten that wait. OpenGl is the Only hardware acceleration Apple uses, Apple features OpenGl in multiple parts of the OS least of which is Quartz Extreme. Apple does these things in order to speed up the development of OpenGl, drivers, and associated software products on the mac. They freely give away all of their openGl research.

Apple did some spelunking in their history, and determined that it took 10 to 15 years to develop Mature drivers for Quickdraw. They are determined, irritated, and aggressively trying to shorten that time period, and I have been impressed with the pace they have been keeping up for the last 5 years.


I understand that tweaking the driver and the openGL takes time, but the difference I see is realy huge




Now having said that, I am sure there are many many many places where Lightwave is Processor dependent, or the opengl code is not necessarily the fastest it could be. Your 2 20 inch monitors, may very well be completely sucking down the availible OpenGl resources, or you may not have enough ram. Also Lightwave is not one of the fastest 3d application available for the mac. Its one of the most useful.

for rendering Lightwave is processor dependent, and there the dual 2 GHz is faster than the AMD 3500+ (about 50%)

as for the monitors, both system have dual 20" setup, and both have at least 2 Gb of ram (and when i check there is at least 1 go free on the Mac) and both system have Geforce 6800 GT with 256 Mo RAM


what is strange to me is that i think i saw OpenGL benchmark somewhere and the difference between the Mac and the pc was 20 to 30% (from what i remember) while here i would say 300 to 500% difference EEK!, could it be that the problem is more related to lightwave than to Mac OS ?

Ade
07-27-2005, 05:37 PM
I wouldnt say its all Apples fault, yet alot is if you look at frame rates in games vs pc's... But using apps like c4d and m odo, you can see how much better their Open Gl is compared to lw's. I believe alot of it is and always was, the fact macs arent seen as important as to pc's except if you are maxon.

pantone
07-28-2005, 07:44 AM
In my experience the Mac OpenGL slowness in LW is a Newtek issue.

I've taken full scenes that slow the Layout interface to a crawl and loaded them up in Cinema4D and Maya and had zero display slowdown. That for the exact same scene.

I hope this real-time OpenGL rendering in the next update also means they've worked out OpenGL display issues as well.

Lightwolf
07-28-2005, 08:06 AM
I wouldnt say its all Apples fault, yet alot is if you look at frame rates in games vs pc's... But using apps like c4d and modo, you can see how much better their Open Gl is compared to lw's. I believe alot of it is and always was, the fact macs arent seen as important as to pc's except if you are maxon.
Well, Mac C4D users complain about openGL being slower on a mac as well. I guess we have two issues here:
1) openGL on LW is generally slower than on any other 3D app (or, put it this way: the display speed is - I assume most of the slowness is related to the data not getting to the gfx board fast enough, not to openGL per se).
2) Macs seem to have lower openGL performance compared to PCs with identical gfx boards and apps.

cheers,
Mike

Ade
07-28-2005, 08:13 AM
I think we need to get off newteks back and accept mac isnt worth the trouble, they do great work in the updates for pc lw and should be thankful.
Either buy a pc or move to maya or c4d i say if you dont like newteks mac support...We have asked for ogl update in speed from day one of carbonisation, they must be sick of this request and so r we.

pantone
07-29-2005, 10:28 AM
Lightwolf, you nailed it.

Yes, in general, OpenGL acceleration is slower on a Mac than a PC.

On the lightwave side it is probably more accurate to say that interface performance degrades with larger scenes.

When I load a large scene in Layout not only do the animation viewports slow down...but the interface slows down as well. If I slide the time thumb there is a huge delay between when I drag and when it actually starts to move. It's this type of behavior I don't see in the other Mac 3D programs I've used. (Although viewport manipulation is also faster on the same scene as well.)

And I suppose it wouldn't hurt to offer Radamanthys an answer to his original question....

Yes, what you're seeing seems to be normal between Mac and PC in the Lightwave world. I'm sure you know this, but just in case, make sure and set all the Sub D objects you can to display at Zero in layout. This should help some.

Radamanthys
07-29-2005, 11:47 AM
thank you for all your answers.

I was asking the question because I cannot comapre with other 3D applications, and I'm new to the Mac. I think I'll wait till LightWave 8.5 to decide if I use the Mac for the 3D or if I better upgrade a bit my PC processor (for rendering my mac (dual 2 GHz) is faster than my pc (Athlon 3500+)) and keep the Mac for Video editing

Ade
07-29-2005, 06:10 PM
Rada i dont think this 8.5 applies for us mac people...
Judging what ive read and the fact theres been no response from NT about it, I think in the 8.5 update all we will get is the new multiple smoothshift and new i/o features.

If the mac isnt supported for these new shaders to gpu, i think i wont jump to pc, ill just jump apps.

Lightwolf
07-30-2005, 04:07 AM
If the mac isnt supported for these new shaders to gpu, i think i wont jump to pc, ill just jump apps.
Why shouldn't they work on the Mac? openGL is extremely platform agnostic and this includes shaders as well... if the gfx board and the drivers support it that is.

Cheers,
Mike

rpsolo
08-01-2005, 06:58 AM
you think Mac switching to intel is going to "speed up" the situation any?

powersolo

Lightwolf
08-01-2005, 07:03 AM
you think Mac switching to intel is going to "speed up" the situation any?

Not really. I don't think this is a processor issue in the first place, the G5s seem to perform comparable to P4s, so the culprit must be somewhere else. Maybe Apples GUI takes away to much ressources from the openGL board or whatever...

cheers,
Mike

rpsolo
08-01-2005, 07:10 AM
I looked at some of the graphics boards lately... and you can get more ram on a PC graphics card than you can a Mac. I'm guessing that Macs can use the same boards as PC's once the change is made...

Powersolo

Lightwolf
08-01-2005, 07:15 AM
I'm guessing that Macs can use the same boards as PC's once the change is made...

Not really, they will still need OSX drivers... After all, only because Linux runs on intel CPUs as well doesn't mean it supports all the hardware that Windows supports...

Cheers,
Mike

gerry_g
08-02-2005, 12:33 PM
Not too sure about Open GL being 'platform agnostic', I remember reading on ArsTechnica a while back about the pros and cons of the Apple Intel switch in which one of the things mooted was how Mac Open GL performance would benefit from Intel's algorithms........also not too sure Newtek are blameless either, last night I opened a sub patch model I'd been working on in modeler, in Modo and it ran noticeably faster, same thing when I took it into ZBrush...... this is on a Dual 2.7 with a 250Mb graphics card and 6GB's of installed ram.

Lightwolf
08-02-2005, 12:56 PM
Not too sure about Open GL being 'platform agnostic', I remember reading on ArsTechnica a while back about the pros and cons of the Apple Intel switch in which one of the things mooted was how Mac Open GL performance would benefit from Intel's algorithms........also not too sure Newtek are blameless either, last night I opened a sub patch model I'd been working on in modeler, in Modo and it ran noticeably faster, same thing when I took it into ZBrush...... this is on a Dual 2.7 with a 250Mb graphics card and 6GB's of installed ram.
Have you tried the same object on the same openGL hardware on a compareble processor in the windows world?
I don't know how Apple would be helped by intel in that area, since intel doesn't write the drivers except for their own (low end) integrated gfx hardware targetted at the business market.
The only thing that will be easier for hardware vendors will be to port over optimized code written in assembler/machine language. Then again, it would supsrise me if Apple didn't optimize to that level in Tiger already (after all, having limited hardware support is supposed to be an advantage of the Mac, so the OS should be fully optimized considering how few choices there are to optimize to).
That's why I said what I said, NT is to blame for the relatively poor performance vs. other apps, Apple for the relatively bad performance vs. other platforms.
I mean, I'd never blame NT for the fact that Quicktime is crappy slow on even the fastest PC, even though LW supports it...
Cheers,
Mike

archiea
08-02-2005, 07:12 PM
Not really. I don't think this is a processor issue in the first place, the G5s seem to perform comparable to P4s, so the culprit must be somewhere else. Maybe Apples GUI takes away to much ressources from the openGL board or whatever...

cheers,
Mike

While not related but equally intruiging, if you goto barefeats.com, you can see some benchmarks where the G5 is actually slower than the G4 in the area of Firewire HD's. Its really strange. It was thought a firmware would solve it but it hasn't. There seems to be some bottlenecks in apple's stuff, but in odd places...

Lightwolf
08-03-2005, 02:48 AM
While not related but equally intruiging, if you goto barefeats.com, you can see some benchmarks where the G5 is actually slower than the G4 in the area of Firewire HD's.
That might just as well be a chipset problem or so. I guess that is one good thing abut going to intel, their supporting chipsets are excellent.

Cheers,
Mike

eblu
08-04-2005, 11:13 AM
Light,
its long been known that LW display performance suffers on the mac due, in part, to processor bound processes in the display pipeline, that are just slower than their Intel counterparts.
Theres no question that the display performance will be enhanced by the intel switch, but there is a big question mark over how much.
I do agree that the intel switch won't benefit display technology in general on the mac, but Lightwave does alot of stuff during display thats got nothing to do with the card, and even some of the stuff that the card 'could" be doing, is done in processor. This is one of the things that Luxology claims makes them faster.

Lightwolf
08-04-2005, 02:58 PM
Light,
its long been known that LW display performance suffers on the mac due, in part, to processor bound processes in the display pipeline, that are just slower than their Intel counterparts.
Which is weird, I was under the impression that a G5 was at least as fast as a 2.8 - 3.0 GHz P4... (well, according to Apple as fast as a 10GHz P4 ;) ).


I do agree that the intel switch won't benefit display technology in general on the mac, but Lightwave does alot of stuff during display thats got nothing to do with the card, and even some of the stuff that the card 'could" be doing, is done in processor. This is one of the things that Luxology claims makes them faster.
Yeah, but that is the "other" issue, the part that NT is responsible for and they are supposedly working on it (in 8.5 and 9.0 if I read Jay Roths interview correctly).
Then again, those changes will profit both platforms, so the realtion is likely to be the same.

Cheers,
Mike

jasonwestmas
08-08-2005, 12:25 PM
Not too sure about Open GL being 'platform agnostic', I remember reading on ArsTechnica a while back about the pros and cons of the Apple Intel switch in which one of the things mooted was how Mac Open GL performance would benefit from Intel's algorithms........also not too sure Newtek are blameless either, last night I opened a sub patch model I'd been working on in modeler, in Modo and it ran noticeably faster, same thing when I took it into ZBrush...... this is on a Dual 2.7 with a 250Mb graphics card and 6GB's of installed ram.

I think you may be comparing apples with bananas here. Programs like Zbrush are really stripped down and optimized so it can import and edit 8 million poly meshes. You would have a hard time importing that kind of mesh in any 3D animation program reguardless of the machine you are using. Truth be told, competing programs like Lightwave and Maya use up a lot more Ram mem than Zbrush when modeling. The multiple viewports in Lightwave and maya are a part of that. As for modo, I've never used it so you may be right about the subpatch modeling.

gerry_g
08-08-2005, 03:58 PM
I think you may be comparing apples with bananas here.

Donít have Maya so I canít make the comparison, so hereís Modo in quad-view and Lightwave in tri-view, guess which is faster ..............and for all this I still think Lightwave the better product on balance, (I don't hold the the development team bashing that goes on on this site, some of it's downright nasty). Yet I still say between the open GL implemented in Lightwave and that in OSX Tiger 4.2 thereís room for improvement

http://homepage.ntlworld.com/gerry_g/modo.png
http://homepage.ntlworld.com/gerry_g/lightwave.png

jasonwestmas
08-08-2005, 05:31 PM
Yes, I really love the subpatches in LW modeler. It's the raw polygon counts that seem to really bog down lightwave. Don't have that problem in Zbrush, it handles huge numbers of polys great. You see I like to import high poly meshes into lightwave from Zbrush to create normal maps. Unfortunately I can't really import very high poly meshes. 300K triangle meshes are really pushing my 1GB of Ram when I render a normal map. Maybe 4GB of ram would give me the ability to rend out my nmaps in higher detail.

DragonFist
08-08-2005, 09:15 PM
You all have to understand that LW's OpenGL implementation is software based (proir to 8.5) and receives little to no help from the GPU. So, it IS going to be slower than other apps that have hardware accelaration.

jasonwestmas
08-08-2005, 09:56 PM
Good Point DragonFist! That probably does explain a lot of the slowup. This has probably been addressed by newtek in the past but are they planning on making better use of the GPU I wonder? Sorry I haven't been keeping up do date with the technology.

DragonFist
08-08-2005, 10:13 PM
Hardware Accelerated OpenGL Options come with the 8.5 update, available soon.

Tesselator
08-08-2005, 10:19 PM
You all have to understand that LW's OpenGL implementation is software based (proir to 8.5) and receives little to no help from the GPU. So, it IS going to be slower than other apps that have hardware accelaration.


You do realize that this is a completely
wrongfull stantment? I hope! :D

If you have a display card that supports
accelaration and if there is an OpenGL
driver then the driver will translate the
the OpenGL functions used in the program
into accelerator functions the card can
process. As LW uses any OpenGL functions
to draw it's display those functions are
then routed to and executed on that
card's GPU via driver "translation" if you
will.

If I write a simple OpenGL program I don't
have to worry about if there is a display
adaptor or not. If there is one and it has
an OpenGL driver... it get's used. If not
then the MS software based library handles
it.

:cool:

DragonFist
08-08-2005, 10:31 PM
Nope, in fact, I am fair certain that I am very correct on this, as I have read about it here on the forums and when Newtek first implemented it. It was originally to get around various difficulties with OpenGL on win9x systems back in the 4.0-5.0 days.

Hardware Accelleration is now being implemented in 8.5 as seen here this quote:

"New OpenGL Hardware shader option using the hardware renderer in OpenGL video cards to display many of the render functions in real-time. Lights, light falloffs, blending images on surfaces, per-pixel lighting will all be available in real-time, with an option to turn this feature on and off."

Gotten here:


http://www.newtek.com/lightwave/newsletters/v2-n6/index.html#85

Anyhow, since I am not a programmer or developer, I suppose, I could be wrong but it definitely my understanding as per the above and quite a few earlier articles on Lightwave's OpenGL implementation. Since, some are from years ago, I might have a hard time finding them but I'll take at look. But I do believe the above says a fair amount.

Cheers,

Shawn

Nu Visual Scien
08-08-2005, 11:15 PM
Open GL is a standard used so that if any compatible hardware is available it will be used to acelarate the graphics, As it is a standard i would say for Newtek to stray from this standard would be strange indeed and probably more time consuming, I would also say that if you where to remove an Open GL graphics card from a system with Lightwave and replace it with a non Open GL card and then try it, You would find it to have considerably slower Open GL performance because the hardware wasn't available, The new enhancements to the Open GL are just that and not a new hardware Open GL implementation, I guess you are being mistaken by the implementation of hardware shaders, This does not mean that all the previous features in 8.3 and before are not hardware based it means that the previous versions didn't have hardware shading that is all.

Tesselator
08-08-2005, 11:17 PM
Anyhow, since I am not a programmer or developer, I suppose, I could be wrong ...


You are :D

You're confusing "hardware shaders" with
other forms of hardware acceleration.

Just trust me. If you have a card that
supports acceleration and there is an
OpenGL driver for that card then ANY
and ALL programs the use the OpenGL
language WILL be accelerated. Same
with DX... It's the entire purpose of
these graphic languages or "wrappers"
so that we programmers don't have to
write code for each and every graphic
chip that exists.

:cool:

Tesselator
08-08-2005, 11:24 PM
Here, try this:

Type: "define:opengl"
into the google search engine (without
the quotes) and read up for yourself.

:D

jasonwestmas
08-08-2005, 11:43 PM
Wow, realtime perpixel shading! That's really cool stuff. Not sure if any of those enhancements will speed up OGL but Thanks:)

Lightwolf
08-09-2005, 02:51 AM
You do realize that this is a completely
wrongfull stantment? I hope! :D


Thanks Tess, now I don't have to jump in ;)

Cheers,
Mike

Lightwolf
08-09-2005, 03:02 AM
Nope, in fact, I am fair certain that I am very correct on this, as I have read about it here on the forums and when Newtek first implemented it. It was originally to get around various difficulties with OpenGL on win9x systems back in the 4.0-5.0 days.

Hardware Accelleration is now being implemented in 8.5 as seen here this quote:..

As Tess mentioned, this is per pixel shading and openGL shader support... which is new and wasn't in LW before anyhow...
LW has been supporting openGL since the early days on the PC (I believe 4.0 was still software based, for many reasons*).
As soons as an openGL video board driver kicks in and installs an openGL driver (which all do nowadays), LW uses that.
If openGL was still software based, you'd never even get 1024x1024 texture maps with mipmaps and smooth interpolation in a viewport even on current machines.
The issue is that LWs pipeline to push out data to openGL is a bit antiquated, using openGL features that were en vogue when it openGL first appeared but have since then been superceeded by newer calls that make much more sense in high poly scenes.
So the problem is software based (i.e. how is openGL being used by LW) ... but openGL is still hardware accelerated.

Cheers,
Mike
* As a bit of trivia: One of the reasons was Modeller. Previous to 5.0 Modeller was using wireframe modes only, drawing them via the CPU, not hardware accelerated. Due to the lack of acceleration by boards back then this was in many cases faster. It also had one advantage: The drawing of the display could be interrupted (if, for example, you decide to pan to a different part of the mesh, the viewport drawing would just stop, pan, and redraw the mesh). Using openGL this isn't possible anymore. In those early days using openGL was actually a set back for people using high poly models, because it did make work slower in many cases.
Another "feature" that went when Modeller supported openGL was the fact that before that background images could be any size, now they are limited to the size of an openGL texture, which is (again) limited to 1024x1024 in LightWave.
Oh, and clipping problems didn't occur half as often in modeler back then either ;)
Then again, _all_ openGL (or D3D or whatever) based apps have to fight with those problems.

Radamanthys
08-09-2005, 04:16 AM
here are few arguments that tend to make me think like DragonFist


1) I have on my Mac a Radeon 9600 (not pro or xt) and a 6800 GT DDL, and guess what ? I cannot see any difference in the OpenGL speed of Lightwave.

2) Software like Combustion and Vue 5 have functions to let you choose between a software or hardware accelerated openGL. And in Vue, they recommend you to switch to the software OpenGL if your card is not validated.

http://mail.materianova.be/vue_ogl.jpg

3) I remeber when i was developping an OpenGL application (on PC) that the visual result could vary alot from one board to another, and this could explain why one whould choose the software OpenGL (the same on all board, less time to validate the cards, drivers version, etc...)

4) As for the speed, the software OpenGL is not that bad nowadays, I used Lightwave during my hollydays on a centrino laptop that jsut have a shared memeory graphic board (intel chipset) and it was running pretty good (at 1024x768), i felt it even alot more responsive than the 6800 GT on the Mac :(

Darth Mole
08-16-2005, 04:14 PM
This has bugged me for years: I've had LW since the first Mac versions and every couple of years I update the Mac and/or the graphics card. For as long as I can remember the playback of the Hummer scene has been within half a second or so, no matter what card I have. I'm currently running an ATI X800 XT - it's no slouch. You should see Halo with everything turned on!

Interestingly, when I first got this card I ran several tests including Maxon's Cinebench and it scored almost identically to my old ATI 9800. (Which I have sitting in a drawer if anyone feels like upgrading for a few quid?)