Hi DesuDeus, do you mean the same models are noticeably slower than on your previous machine?

One thing that I'd recommend testing is whether it's edge display or shaded surface display that is slow - you can do that by going to the scene browser and first hide all edges and see what performance is like, and then hide all faces and see what performance is like.

If things are very fast with edges hidden and only shaded surfaces being displayed, and only slow when edges are being shown, then it might be possible having so many cores is actually causing a problem with too much contention between them and setting a display thread limit might help.

Also is it possible that each individual core is actually slower on this new machine than on your old one?



> I'd like to know if changing the gpu from an ati 5870 with 1gb to a GTX660 with 3gb will
> improve the framerate of MoI on large models on osx.

It's possible that it could, but if it did make an improvement it would be due to any GPU processing speed advantage, not to any larger memory available. MoI does not actually use all that much video memory and so just an increase in video memory alone will not really do anything.



> Or is there a limitation in Wine on the allocated graphic memory ?

I don't think there is any special limit to it, it just basically passes requests for resources through to the operating system. But since MoI does not particularly use a lot of video memory it doesn't really make much difference in this area anyway.


So anyway one thing you might experiment with is turning the display thread limit down, to do that go to Options > General and push the "Edit .ini file" button. Under the [View] section go to the DisplayThreadLimit= entry and experiment with a few different settings like setting it to 1, 4 or 8 and see if a lower thread limit actually improves the performance. To set the thread limit, edit the line and put the number directly after the = , like this:

DisplayThreadLimit=1

- Michael

Hi, Thanks for the reply.

That macpro is slower than my old pc for MoI (i7 quad core at 2.6ghz with 24gb ram and a gpu with 3gb of ram)

I will try the Displaythreadlimit

It's slow with edges, a bit less slower with edges off

The model is a complete building (no interior) with around 2200 windows (the file is around 800mb which is huge)

Is there any performance differences between moi on osx vs windows ?

Hi DesuDeus, so I guess I remember now hearing some general grumbling that the MacPro line hasn't really been updated recently so the current crop of stuff is based on things maybe 3 to 4 years old...

So it sounds like you may have gone down a step in processing power, that's mostly probably what you're seeing.

A video card upgrade could possibly help, now that I look into it, the Radeon 5870 that you currently have is from the Radeon HD 5000 series. I looked that up and that general chipset wave was launched in 2009, and there has been the HD 6000 sand HD 7000 series since then, so your card is actually a couple of generations old at this point.

Each new generation usually brings increased processing speed and more efficient operation, and it's these speed improvements that will have the most impact for use with MoI with heavy scenes, not really just having more video memory. Any video memory increase itself doesn't really do anything for MoI in particular, but having the latest generation and most efficient GPU can help with heavy scenes.


> It's slow with edges, a bit less slower with edges off

Usually if it's not pretty dramatically faster with edges off (when you're only showing just shaded surfaces alone and no curve drawing at all), then that probably means a GPU upgrade would help.


> Is there any performance differences between moi on osx vs windows ?

I have not done any heavily detailed comparison benchmarks but from what I have seen there should not be any really big difference.

It seems like you are probably seeing the difference from going from a current generation GPU to a 2 generations older GPU I guess.

But you've also got to be careful to get a GPU that Apple supports properly, it's not generally as easy to just stick in any internal hardware because the video drivers for the Mac are coming only from Apple I think.


- Michael

Hi Michael,

Yes indeed the 5870 is crap and old (wondering why apple still put these in macpro...)

We buyed a windows pc at work with a gtx680 with 4gb of ram (for heavy scenes in cinema 4d as well)

In our current macpro we can stick gtx560ti (use only 1 6pin connector) if osx is updated to 10.8.3 because Apple released drivers for gtx6XX cards in 10.8.3

I tried the displaythread thing, doesn't improve things much (tried with 1,4,8 and 12)

Will compare with the pc then maybe buy a gtx560ti :)

Thanks for your answers !

Hi Desudeus - I have used the GTX-660-Ti on 10.8.3 and it worked great on some of my larger models (130mb+) It starts of a little sluggish but once it has done a few rotations it becomes smoother but you certainly don't get too much of a speed boost. Sadly I am back on Windows at the moment so can't give you a frame rate, but If i remember rightly on default MoI interface settings I got 15.?? fps on my Lamborghini model that can be found on the forum here http://moi3d.com/forum/index.php?webtag=MOI&msg=5101.1 ... Once i have mastered some poly modelling application then i will be returning to Mountain Lion - It's so much smoother on rendering and working on other things.

Other specs that might have some influence ...

CPU i3570k @ 3.8ghz quad core but no HT
32gb 1600Mhz
GTX660ti
SATA-II Drives @ 1TB and 500 Gbytes

To be honest, I am a little disappointed with the bang per buck of the 660ti, but I use Keyshot for rendering which is CPU based and not GPU based, however Blender Cycles gets a massive boost over the old 560 I was using. I have gotr round to setting things in Modo to use GPU, or indeed if Modo can use it, but the render engine in Modo is pretty damn rapid too

Hope that helps you a little ...

Martin Spencer-Ford

Sorry but I would hardly call the 5870 crap.

I'm running windows 7 and upgraded from the AMD 5870 1 GB to a NVIDIA 670 4GB card. It's nice to have the extra video ram with modo and games,
but I tested the frame rate in MoI with the same file, settings and monitor resolution as in this thread:

http://moi3d.com/forum/messages.php?webtag=MOI&msg=5252.72

I got almost exactly the same results, 17 to 20 ms.

I would say that your problems have more to do with ram, you said your working with a 800 MB file. MoI is a 32 bit application so it doesn't matter if you have
28 GB of ram. I'm not sure how the Macs manage memory but on windows the best you could expect would be 4 GB of ram for MoI.

Hi Dan - running out of RAM with 32-bit MoI won't really have much to do with things going slower - if you run out of RAM things just won't work at all, like chunks of the model won't end up getting loaded at all or things like that. But basically things just won't work at all, it's generally not that they will work but more slowly.

The main time that running out of memory is related to running more slowly is when you have a machine that has less than 4GB of ram and it's used up all available physical ram and is using "virtual memory" instead where it's swapping chunks of existing memory out to disk in order to free up more of the limited physical RAM. That does cause a really major slowdown when that happens, but that will happen if you're trying to run a scene that uses say 1GB of RAM on a machine that only has 512MB of RAM. This "excessive swapping" type of slowdown won't happen at all on high memory machines since MoI can't even fill up the entire physical RAM to start with.

That does remind me though, there is a difference in max memory with MoI between OSX and Windows - on Windows MoI can use up to 4GB (in practice more like around 3.6GB), while on OSX it will max out at around 2GB instead. That is something that I may be able to fix, it is on my list to investigate in more detail.

I guess it is possible that if you were right almost exactly at the limit it could make for some strange operation with a lot of things ending up sporadically failing. But you'll probably end up just crashing instead of just running more slowly.

- Michael

Hi Michael,

The last paragraph referring to ram was more of a caution about working with 800 MB files in MoI and not
directly related to navigation, sorry.

MoI will do it but a user should not expect to be in the 'comfort zone' for navigation even with 24 GB of ram.

This is one of my Torus experiments, the saved file size is almost 300 MB.


Here is the Task Manager with this file loaded into MoI, and ram is still needed to export out polys.

Hi Dan, the surfaces look pretty smooth there - are you running with the default display mesh density?

By default MoI makes pretty dense display meshes, it can help reduce memory consumption by quite a bit to turn those down and make a rougher display mesh. It can also help to avoid bogging down the display as well since there's less data being shoved over to the video card.

To turn down the density, go to Options > View > Meshing parameters, and set "Mesh angle" to 30 degrees and uncheck "Add detail to inflections".

Then shut down MoI and try loading your heavy torus file into a new instance, does that reduce your memory consumption?

- Michael

Hi Michael,

I usually run the display mesh at 6-12 deg. With a setting of 30 deg. it dropped my mem. usage to about 1 GB.

Thanks!

Is the add detail to inflections option basically AA?

> I usually run the display mesh at 6-12 deg. With a setting of 30 deg.
> it dropped my mem. usage to about 1 GB.
>
> Thanks!

You're welcome - yes with heavier models it really helps reduce memory use.


> Is the add detail to inflections option basically AA?

Not quite exactly - when it is enabled it forces additional refinement of the display mesh to put more finely diced triangles in places that have inflections, like in the middle of an "S" type shape where curvature is switching from one side to the other. Those areas tend to be prone to having display artifacts where the shading can look bad if there are not many polygons there. But when you have a heavy model it's better to just not worry about display artifacts and instead have speed and reduced memory footprint to have more priority.


- Michael

Hallo Michael,
my MAC Pro 12-core runs out of memory if I try to export a 130 MB .step file into an .obj file (ca. 1.000.000 polys).
Can I change some settings in MOI to solve this problem?

Thanks for your answer
moritz

Hi Moritz,

> my MAC Pro 12-core runs out of memory if I try to export a 130 MB .step file into an
> .obj file (ca. 1.000.000 polys).
> Can I change some settings in MOI to solve this problem?

Check out here for some things you can do:
http://moi3d.com/forum/index.php?webtag=MOI&msg=5322.2

There are some settings you can adjust to reduce memory load and also if you export the file in a couple of pieces rather than all at once that can help a lot too.

- Michael

Hi Michael,

Thank you - your advice to turn down the display thread limit also helped fix the display refresh problems I was having with MoI3D v3 on a Lenovo Thinkpad T410 with "Intel HD Graphics" and Win7 64 bit.

I was experiencing very slow refresh and complete white-out blanking in either single or multi-view display windows - i.e. all or most drawn entities were disappearing, including the background grid - after zooming, selecting a drawn entity or deselecting all by clicking on a blank area of the display.

MoI3D display performance on my T410 now appears to be as good as my desktop PC by using:
"DisplayThreadLimit=1"

As it now works fine, I see little point in experimenting further, but I'm kind of curious to know what advantages if any, might be seen from using: 1 < DisplayThreadLimit < 'moi.ini (default) unknown number causing my reported symptoms'?

Hi hairykiwi,

> As it now works fine, I see little point in experimenting further, but I'm kind of curious to know
> what advantages if any, might be seen from using: 1 < DisplayThreadLimit < 'moi.ini (default)> unknown number causing my reported symptoms'?

The default in v3 is to use multiple threads to work on filling buffers that are then sent to the graphics card.

The reason behind the multiple thread use is to get more use out of multiple CPU cores during a screen draw operation, speeding up the display on heavier models.

But certain drivers seem to be sensitive to it.

- Michael

Thanks Michael.
Cheers,
hairykiwi