Hi,

I am noticing that rendering in Poser on medium settings is just mind numbingly slow.

In the time it takes to render on medium settings, I could save the pz3 file, open it in Cinema 4D, render on comparable settings, and have oodles of time to spare. Its just nuts.

I have a 8GB  6 CPU machine with a Radeon 6870, not exactly a dinosaur...it almost feels like Poser isnt using all the processors or...something..?

Is there a fix for this, or should I basically just resign myself to doing all the rendering in CD4?

Thanks

have you set the number of threads in preferences , its not automatic .

With 6 cpu I would try 5 threads, and then maybe try 10 and compare the difference in times.

The video card is not used for rendering.

What version of Poser?

Pro 2010.

 

Ah, great, I was hoping I had missed something about the number of threads. Great tip, will give that a shot.

 

Thanks.

6 cores - I7? with hyperthreading? If so, set threads to 12.

Lots of people keep telling users to use one less thread than cores, so as to leave some free so you can do something else. That's a laudable goal, but not the best way to go about it.

The best way is to lower Poser's priority to below normal. Then let it have 100% - all cores. The moment you touch the mouse in another app, it will suspend Poser instantly - no waiting, and yet Poser will run as fast as possible if you're not doing anything.

There's a way to set this on the command line as well as via task manager.

Or are you on a Mac?

PC.

OK I tested it both ways. 5 threads, then 12. Suprisingly, 12 threads was slower! AND it utilized all 6 cores at 100% till I cancelled, where with 5 threads it was around 83% on 5 cores. Weird.

Anyway, neither change really solved the issue. I guess C4D just has a faster engine. Not too big a deal for me, but I feel for others in this boat that have to render everything in Poser.

Another thing I noticed - Poser seems to take forever to load texture/jpg files on the first render. InterPoser in C4D is faster there too, so odd.

Please fon't get me wrong, I am enjoying the heck out of Poser, and I think they have done a great job putting a lot of power into an interface that is on the surface at least, simple and very user friendly.

Thanks again for the inputs.

If you were at 83% on 5 cores, then it's not hyperthreading. So I'd use 6.

The first step was to get using 100% CPU. That's out of the way. So now we could look at render settings and materials.

Yes Poser is not the fastest by far. But there are quite a few things that increase the render time non-linearly. I never use the automated slider. I select settings myself. Also, when IDL is involved, I don't even use Poser's advanced render settings dialog. It's still to simplified. I use the add-on one included with it, found in menu Scripts/Partners/Dimension3D/Render Firefly.

There is also the question of final render versus test render. Among test renders there are lighting tests versus material tests. For lighting tests, I really jack up speed in exchange for quality. There was a thread when Poser 8 came out where a challenge was thrown down - to make a good render in less than 8 minutes with IDL. I did it easily.

Have a look:

Breathtaking indirect lighting renders in under 8 minutes?

 

Quote - ...

Another thing I noticed - Poser seems to take forever to load texture/jpg files on the first render. InterPoser in C4D is faster there too, so odd.

Please fon't get me wrong, I am enjoying the heck out of Poser, and I think they have done a great job putting a lot of power into an interface that is on the surface at least, simple and very user friendly.

Thanks again for the inputs.

Poser's rendering algorithm caches textures on disk before it starts putting colorful pixels on you screen. When rendering those blocky parts of an image, one by one, it loads the textures again into RAM. This can be both a good and a bad thing. If you have much RAM installed, Poser likely will not make good use of it - the result is that rendering can actually more depend on the speed of your hard disk than on the speed of your processor. A typical raytracer for example would load all textures at once, and if they fit it can be much faster (i do not know what C4D does, but it might do that). So, if you have a slow disk, it might be worthwhile to try a larger bucket-size in the render settings, resulting in fewer texture-loads.

Attached Link: elvish apartment flets in mallorn tree, with IDL emitter lamps and auras

I've noticed a rendering issue which I would like to see reduced/eliminated in P9/Poser Pro ____.  As the IDL precalc nears the end (the progress bar is at the end, though the displayed buckets often don't keep up), the cores begin to drop to idle.  On the current image, it has been this way for over a day.  Eventually, it suddenly dumps all the precalc buckets to the screen and then begins the render.  Near the end of the render, it again drops cores one by one to idle, then sits there with minimal processor activity for several hours, and then eventually it will suddenly dump the remaining buckets to the screen and finish.

P8 does this also, but lesser in magnitude, simply because being 32bit I am forced to limit myself to smaller/simpler renders.

I would really like for SM to try to reduce/eliminate this idle time in the next versions of Poser.  I'd be willing to run test renders on my workstations.

It isn't an issue of running low on memory and resorting to hard drive page filing;  There is plenty of RAM available.  The scene is fairly complex.  Render settings use 5 raytrace bounces at 90% irradiance cache, IDL quality 7, pixel samples 3, min shade rate 0.25, bucket size 64, min displc bounds 1.0

Galadriel is running Poser Pro 2010 on 64bit Vista Ultimate;  she has a HyperThreaded quad (core i7) and 12GB RAM.  She renders 8 threads.

Cameron is running Poser Pro 2010 on 64bit Win7;  she has dual HyperThreaded hex-core procs and 96GB of RAM.  She renders 24 threads.

The link shows the last render from Cameron - it's a big tree, and the spiral stairway is fairly complex.  The lamps glow from internally concealed IDL emitter props, and the surrounding aura uses the double edge-blend node method discussed in another thread.

Attached Link: acroflet atop the gnarly mulberry tree, with IDL emitter lamps and auras

*TinkerBell* runs P8 on 32bit WinXP with a quad core proc and 4GB RAM.  Tink managed to finish this closeup shot of the flet at the top of the gnarly mulberry tree, but then there isn't a lot in the picture.  Same render settings -less gamma correction- and same IDL emitters inside the lamps and auras surrounding.  The three default Poser infinite lights are dialed down to 8% for these shots.

The idle core issue is noticeable in P8, but the delay is less than what I see with bigger RAM-hogging scenes done in Poser Pro 2010.  Tink renders 4 threads (her quad proc does not seem to be H/T).

They are working on it. As each thread finishes, it does a memory cleanup. Basically, it releases all the millions of tiny memory chunks it accumulated during the render.

The method by which it does this is being improved.

The traditional approach of releasing each chunk causes an analysis to see if the newly released chunk can be coalesced with a previously released chunk, or must be kept as a new free chunk. This is a lot of work. If you know you're giving it all back, there's no point in analyzing each piece one by one.

That's good news!  :D