View Full Version : 64bit vs 32bit

06-28-2006, 01:54 PM
I'd just like to ask those who have tried it, what their experience has been with the 64bit version compared to the 32bit. I'm going to be setting up a new workstation and I'm debating which way to go.

For this application, my scenes are medium complexity so I don't need more than 2gig of ram to accomplish the task so the extra memory headroom is meaningless. I'm just wondering about performance.

Do medium complexity scenes render any faster? Is is significant?

Is OpenGL any faster, all else being equal? (remember, medium complexity, 300,000 to 1 million pollys).

How does stability compare?

Also, as a side note, how is compatibility with components like ultra 160 scsi controlers etc. Do you HAVE to have 64bit specific drivers?

Any insites would be helpful.

Captain Obvious
06-28-2006, 02:07 PM
First off, a disclaimer: I've only ever run Lightwave on a Mac, never on Windows. Thusly, I've only used the 32-bit version...

Anyway, I don't know how Lightwave reacts to being 64-bit, but some (Cinema 4D being a prime example) show a performance increase when running in 64-bit mode (64-bit CPU, 64-bit Windows, 64-bit application). When rendering, this can be as much as 20-30%, on the same exact setup. This is regardless of the amount of memory. Even the not very complex scene rendered in Cinebench shows a significant performance increase.

I would not be surprised at all if Lightwave showed a similar behaviour when rendering. Other than that, don't expect a world of difference.

06-29-2006, 01:06 AM
If you rely on 3rd party plug-ins, steer clear of 64bit for now. No FPrime, no Sasquatch (or even Sas-lite). It's getting better on that score but generally there are a dearth of 3rd party plugs so far.

You'll mainly benifit if you have large, complex scenes. If you aren't hitting memory issues, maybe you don't need it. Performance wise, it depends on the scene. Rarely scenes may render slower, but generally you will see an increase sometimes by a fair amount. Truth to tell I haven't done much comparison in the 2 versions of 9, I'll try to do some later.

Personally, I like it, I rarely use anything else. I drop into 32bit to use Pictrix plugins, (or rarely Sas-lite) but that is all.

06-29-2006, 03:58 AM
While running win x64 you can use both, LW 64bit and LW 32bit. So I use the 32bit version for modeling or scene setup and animating, and if I don't need any unsupported Plugs I render the scene in severall instances of LW64bit.

BTW, wouldn't it be cool to have a choice for the number of lw instances which render your scene? e.g. 4 threads on 4 (background) instances. That would be more elegant than starting 4 apps. I think I'll make a feature request.


06-29-2006, 12:43 PM
Speed wise, there is not really a big difference from my experience in 32 bit vs 64bit LW. However, there is a huge difference in the sheer amount of data 64bit LW can handle.

For example, if you go watch Microsoft's WinHEC from last year, Newtek demonstrated LW64bit. First, the demo began with 32bit Lightwave. The speaker brought up the Battlestar Galactica (one instance) in 32bit LW. The thing was so loaded with detail that he could only get a few clones (I dont remember the exact number) going before the renderer threw up. He then turns around and opens LW64bit, and immediately loads up several 100 clones of the Galactica, and renders it in a few minutes.

The difference here is that 64bit LW will not really (on a 1 to 1 comparison) render faster than 32bit LW. But, if your data set is larger than the 2GB limit of 32bit Windows, 64bit LW will be at a big advantage, both speed wise, and stability wise. This is because in 32bit LW, datasets that are too big for a 32bit system will either crash the computer, or page data to the hard disk, which reduces the render speed to something a 66MHz 486 could beat. 64bit LW and Windows have a much larger limit, and thus can process much larger data sets. And that is the advantage...

Of crouse you'll need a good amount of RAM....

06-30-2006, 02:45 AM
In fact most of the time the Battlestar had to be cut into pieces to fit into memory on the 32-bit version of LightWave, and not only could the whole model be loaded, but a bunch of clones of it could be made in 64-bit...


06-30-2006, 03:04 AM
I have found the 64bit version faster in scenes with 100 or more lights

06-30-2006, 10:50 AM
Complexity is where 64bit LW shines :) Much ilke the new render engine...

06-30-2006, 11:44 AM
My BIG question is will it help me to work with high poly objects in Modeler. Right now in the 32bit things start slowing won way before you get to a 100k poly's. And if you have a model with over a million your all but screwed. Will using the 64bit version allow me to work faster/better with high poly count objects?

06-30-2006, 09:40 PM
In my experience, this is limited by Lightwave's OGL implimentation and not hardware or memory limits. Just to give a hint though, the same 500k poly object I have is just as manipulatable in 32 and 64 bit lightwave. By this I mean not at all :P

07-01-2006, 03:16 AM
Yeah, my 1.5m poly chain-mail skirt (yes oDDity, I'm a cr*p modeler ;) ), rotates just as fast on both: a frame every other second :D

07-01-2006, 07:23 AM
OK so basically don;t risk partitioning my drive for a speed increase when working with high polyes in modeler. I'm not concerned about render times (well not today anyways) so I'll leave well enough alone for now. Thanks for the valuable info my friends!

07-01-2006, 07:30 AM
opp's, sorry this was supposed to be a post for a different thread????

07-02-2006, 10:06 AM
Thanks for all the replies. This helps.

09-06-2006, 05:22 AM
So will 64bit software run on a 32bit machine?

09-06-2006, 08:00 AM
No it will not, as the 64 bit version requieres a 64 bit OS, which in turn, requires a 64 bit processor. However, as mentioned above, you can (and I think most people here with a 64bit do this) install the 32 and 64 bit in the 64bit OS without crosstalk or weird behaviour between them.

09-06-2006, 08:41 AM
64 bit OS

LW 64 and 32 running on the 64 Bit OS have had no problems, it helps having the 8 gig of ram, only the 64 bit LW uses the ram more effiecentaly [bad speller]

I did the Galactica clone thingy on my machine using the 64 bit LW, and I was impressed, it handled it, a liitle slow, but it did it..


09-07-2006, 12:34 AM
The main reason I was originally going to jump to 64 bit was primarily because I could use more then 2gb of ram.

Many times I would crash Lightwave when I felt a scene just starting having the detail I liked, (1-2million polys + about 1 gb of image maps).

Now, if you are short on resources (which we all have learned Lightwave on) you'll know how to get around these things by careful planning and compositing/layering of elements in AE or Fusion.

But all that being said, when we made the Comet Collision animations for Nasa in early 2005 we had a hard time making some things work.

I had a Jupiter Map that would barely load in Lightwave. It is 16,000 by 8,000 pixels. It has great cloud detail and was custom built from hundreds of stiched Jupiter pics. Anyways, this one map would some times load and sometimes crash In Lightwave 7.5 -8.2 at the time. (This has been permanently solved by Infinimap btw)

Same with the Mars Undergroud. I made a Mars landscape and put a few vehicles in and it was already busting up.

Now on the 64 bit machine I don't hit that wall anymore.

Often the main wall I hit is an OpenGL wall in Modeler, which I can stave off by using wireframe mode.

These were big reasons for 64bit more so then say Rendering speed increases. Being able to create the details and complexity was first and foremost. Yet I still waited another year due to systems being wonky and going through the "compatibility" wars.

Now that the compatibility issue is all but over (Plug-ins still lagging) there are other incentives that make 64bit much more attractive. The first is that you can have Multi-processor computers where the processors are each multi-core. My new multi core machine can render a Maxwell image in about one hour. The same image on my P4 3.2ghz box took 12 hours to reach the same SL level.

This is really where you get the render speed increase. It was the final touch that allowed my trigger finger on "buy".

If you do decide on a 64bit system try to see what kind of deals you can get on a single proc/dual core systems then go up from there to see the range of pricing so you can get the best withing your budget.

I know that the Intel Core 2 Duos are really fast. If spending alot of money isn't an option you still have a lot of purchasing room due to 64bit single proc/dual core systems being out for a while.

You can also try the dual procs/single cores route which saves some money (single cores are better for Overclocking as well).

.... or finally if money isn't an issue then you can do a multi procs with dual cores system, (Bleeding edge technology) that will go the distance in render speed increases and, if you have the ram, polygon counts easily beyond 3million+.

We had some scenes with 12 million polys in our testing and they weren't bogged down. Although the poor OpenGL Geforce cards were crapping out around that high. Essentially this means that the OpenGL graphic cards that are for Video games like quake4 are not keeping up with the latest Pro 3d apps. They come close...

Anyways.. I hope my info helps you in some way.