PDA

View Full Version : slow performance - need some advice


tjnyc
09-22-2004, 01:17 AM
Hi,

I have been testing Modo with some models I made previously made, and Modo runs very slow and some tools are quite sluggish like the Edge Slice tool. Is there some option or thing I should do to improve performance?

I am running Modo on a P4 3.0GHz HT, 2GB RAM and a Quadro 4 980XL with the latest Nvidia driver with Dual Monitors.

Should I email the files to someone in support to check out?

http://www.telescript.com/images/tony/modo_dog.jpg

http://www.telescript.com/images/tony/modo_alien.jpg

Thanks

kursad_pileksuz
09-22-2004, 01:45 AM
you can try decreasing subd level, by clicking on layers with rightclick-properties. that should freshen quite alot actually. default value is 2 i guess.i think 2 is creating 16 subd per polygon or something like that. i found 1 to be quite efficient

let us know please if that helps

tjnyc
09-22-2004, 03:08 AM
Thanks,

Yes, it is on 2, and I knew about lowering it, but I was hoping for some way to improve the speed at the default 2 subD. Is this an issue with Nvidia cards BTW? I can work with both of these models fine in SILO and quite fast in XSI at subD at level 2, but Modo really slows for some reason. Are there any OpenGL settings to adjust with the Nvidia driver for performance that Luxology can recommend?


Cheers,

kursad_pileksuz
09-22-2004, 03:18 AM
well if the subd level "2" may not be same as xsi, because subd 2 is going to create alof ot polygons per subd polygon. i mean if you have 10000 subd polygon in level 2, i am guessing it means more than 150k polygons. i tested with 7400 subd polygon model on subd level2, when i froze the model it ended up with 120.000 quad polygons. i am not sure, if you think that is what xsi doing after you freeze it in xsi, then would be nice if we get some answer from luxology

i am seeing that you second model has some 40k , if that had subd level 2, and you think that it was sluggish, actually it sounds normal to me at that number. because that is like over 500k model actually

but i do have an ati 980pro here, and holds alright

tjnyc
09-22-2004, 03:31 AM
Not sure what the actual poly count is during subD for SILO or XSI, other than the fact that in SILO and XSI subD level 2 does produce 16 polys to 1 which seems to the same situation in Modo from what I have gathered. Could I turn anything off or freeze things like weights, lights, or whatever gets assigned by default that could help improve performance. Is there an Immediate option like in XSI so command history is not saved, to improve performance during modeling. I'll sacrifice the command history if I have to to get better performance.


Cheers,

kursad_pileksuz
09-22-2004, 03:37 AM
well then that would be a question for luxology. because i am not sure about limits of subd displaying, if i needed to go that high in numbers, possibly i would have splitted the model to different layers as much as possible and use combination of layer based subd level, like keeping some level one some 2. i will email Brad and lets wait an answer from them, if modo is supposed to be fast with 40.000 subd polygons or not.
also i do not think command history would change a nail there really.

tjnyc
09-22-2004, 03:42 AM
Yeah, the layer/part set approach is what I was thinking about using as well. If all else fails in improving things, I will most likely go this route.

Thanks,

kursad_pileksuz
09-22-2004, 03:45 AM
also different drivers may improve it too. but as a digital age rule, you do not mess with working system

Nando
09-22-2004, 05:40 AM
tjnyc,

Try using the drivers on pny's site those specific ones work better than some of the newer nvidia drivers. not that there different , but they have a recomendation as to what release(s) work.

also disable alll the interface gimicks in XP to speed up the sytem.

And heres a link to a pretty good XP setup optimizations for general appz

http://www.3dluvr.com/content/article/123

Pazur
09-22-2004, 07:24 AM
You might want to try locking geometry the You are not editing at the moment, there is a set of options for that in Edit menu. It gives a bit speed up. But better thing is to hide part of the model You are not working on (shift - h for hiding unselected geometry, u - for unhiding). This will give You faster refresh for sure.

tjnyc
09-22-2004, 02:43 PM
tjnyc,

Try using the drivers on pny's site those specific ones work better than some of the newer nvidia drivers. not that there different , but they have a recomendation as to what release(s) work.

also disable alll the interface gimicks in XP to speed up the sytem.

And heres a link to a pretty good XP setup optimizations for general appz

http://www.3dluvr.com/content/article/123
My winXP is already set to best performance with all the fancy nonsense from MS turned off. I'll try the PNY drivers though. Where is the recommended release info? I haven't found it nor can I find it.


Cheers,

ambient-whisper
09-22-2004, 06:12 PM
try going to prefs/opengl/and enable vertex arrays. ;). its funny because WITH the option turned on, modo seems more stable, even though luxology says theres a huge bug and recommends we dont use it. you should see a night and day difference. not exactly xsi/silo speed, but pretty damn close.:)

this is IF your using nvidia drivers that have the particular bug.

Nando
09-22-2004, 06:25 PM
http://www.pny.com/support/drivers/?prod=quadro


since your card is before the Fx series I say use a driver thats a bit older.
Im using the 30.82's for older Quadro DCC.
I think u might try 41.03, 44.03 , and last 53.03

Pazur
09-22-2004, 06:29 PM
42.51 are very good for GF4. When I installed them I can have vertex arrays on and it stopped crashing.

Dion Burgoyne
09-22-2004, 06:29 PM
nm, revising post :)

tjnyc
09-22-2004, 06:38 PM
Some good suggestion you guys. I'll try some of the things you guys were mentioning.

Dion, thanks for the heads up, I look forward to what you come back with. So, in Modo subD level 2 is really like 128 polys to 1? Why so many polys?

Cheers,

tjnyc
09-22-2004, 06:58 PM
Although I haven't tried it with XSI yet, I have noticed one thing.... modo's subdivision level of 2 very close to dividing a model 5 times in Silo, so a division level of 2 in XSI could be a different setting in modo... I'll do some research and post my findings...Here is a comparison of both Modo and SILO at subD lv 2, they both appear to look about the same, and comprise the same amount of polys. Can you explain a bit about your evaluation, so I understand what to look for, because it looks about the same at the same subd level to me.


http://www.telescript.com/images/tony/comp.jpg

Thanks,

MallenLane
09-22-2004, 06:58 PM
Um, subdividing a cube twice in Silo = 16 faces per original polygon.

Freezing a sub-d cube in Modo (which is level 2 set in the layer) = 16 faces per original polygon

So how do you figure its the same as dividing 5 times?

padawan171
09-22-2004, 07:34 PM
tjnyc,

Along with the other suggestions, you might wanna try using modo's faster SubD mode (hit TAB to turn off Luxology's SUbD algorithm, and then hit Shift-TAB to turn on the LW-compatible Subpatch mode). You'll get lower resolution and accuracy but faster performance, and then when you've finished editing you can turn back on Lux's SubDs. I can confirm this works faster on my system. Hope this helps. :) modo rocks, btw!! so much faster and better than LW!
p.s. also the vertex arrays problem shouldn't apply to you since you're using a Quadro, Lux says it only affects consumer Nvidia cards i.e. geforces. You might also wanna try switching to single monitor mode for modo (yeah i know dualies are great!) cause I can confirm that dualies does slow down OpenGL on nvidia with LW Modeler and Layout both open on 2 monitors. A lower res should also help a little.

tjnyc
09-22-2004, 11:41 PM
You are right about the Dual monitors. I got slightly better performance going single. You were also right about subpatch over Modo's subD, I got noticeably better performance out of the tools and Modo.

I can't say I am happy about these problems, and other troubles I have been having with Modo. I hope changing my Nvidia driver to an older one will help improve things, because things don't look good for me.

Thanks,

kursad_pileksuz
09-22-2004, 11:50 PM
i think this problems you are havign are available only when you get quite high in poly. in my standarts when i hit such high levels, i would start optimizing my work style and model. do you really need to work every detail visible?
but sure everyone has different take on different things. i personally expect optimum speed, a speed that would not keep me away from working.

tjnyc
09-23-2004, 12:40 AM
Not much anymore since I am using Zbrush 2, but my professional work is also in mechanical, arch. sculpture models and I generally need to model pieces that need to be detail. Like the following piece I did. One of the reason I bought Modo was that it was suppose to handle large models well. Very few app can handle large models, XSI being the only one that I have used that can handle large models like a champ. Also, I do manage my work, but you do have to understand I have only had Modo for about 2 days, it is a little early to be nip-picking on my workload management process, you sound a lot like my director.http://www.cgtalk.com/images/icons/icon12.gif

http://www.telescript.com/images/tony/lamp.jpg

Cheers,

tjnyc
09-23-2004, 12:43 AM
BTW, I reverted to an older version driver 44.03, and things seem to better performance wise, not a whole lot better, but things are far more workable now than before.

Cheers,

kursad_pileksuz
09-23-2004, 12:52 AM
wonderful models

i understand your needs, any 3d modeler would want such thing, i just feel similar speed problems from maya and lightwave, so i am not hard on modo in this case. except zbrush, since it is totally different thing in many mays. and i have not used xsi, but sounds like it si really faster than xsi, but question is this, if you like xsi more and if you think it is faster why use modo ? i understand you are testing it so far thou

tjnyc
09-23-2004, 01:11 AM
I like XSI for its speed, but I'm not too high on it's modeling workflow. It has plently of tools, but I generally only need about 6 anyway, so it comes down to workflow. I used to use Mirai, LW 6, Maya and now XSI and SILO but I use SILO most of the time, because it is simple and fast, I don't really have to think about how tools work or what tools to use, its the opposite way when I am working in XSI. Modo is more like a trial for me right now, it has some very appealing tools and I would eventually like to use it over XSI for modeling, I like the management system in Modo, it is exactly what I wanted. It is still too early to be sure what Modo will be for me, since there was no demo, this is my evaluation period and I will most likely throw a bunch of stuff and scenario to see what Modo can and cannot handle.


Cheers,

kursad_pileksuz
09-23-2004, 01:26 AM
i understand, i was just curious about xsi

thx

Levitateme
09-24-2004, 12:28 AM
Hey tj, i started my modeling today and am getting almost same errors as you. can you reccomend to me what worked best for you? ambient whisper said turn on vertex arrays even though its buggy with nvidia drivers. so ill try that, but anything else that helped you alot? cause both our pcs are not slow and this model in maya moves with no lag at all.

I am only showing this image from my maya viewport as you can see its by no means has insane poly count and in maya i can get no lag at all...this is really dissapointing...

http://www.chemicaldust.com/critiques/WB22.jpg

Griffon
09-24-2004, 12:37 AM
Are talking lag modeling, rotating the view or moving the object around?

kursad_pileksuz
09-24-2004, 12:38 AM
well if you are getting slow down as normal polygon object, then that is quite weird. but if you are getting the slow down when you turn subd on, then you can try decreasing subdivion level from layer>properties

tjnyc
09-24-2004, 01:42 AM
Hey tj, i started my modeling today and am getting almost same errors as you. can you reccomend to me what worked best for you? ambient whisper said turn on vertex arrays even though its buggy with nvidia drivers. so ill try that, but anything else that helped you alot? cause both our pcs are not slow and this model in maya moves with no lag at all.

I am only showing this image from my maya viewport as you can see its by no means has insane poly count and in maya i can get no lag at all...this is really dissapointing...


Nice model BTW, great imagination. There were a few things I did. One lower the default subD level to 1. It is in the preference menu up top. Also, depending on what tool you are using, Modo could go from good to bad performance really quick, mainly this happened when I used edge slice, the one way to deal with that is to clear history, and everything was good again. You can also opt to go into subpatch mode instead of Modo's subd model, it is shift+tab, Performance will be better in subpatch. However, I still had issue with crashing, so I eventually lost it and went to a lower driver version, 44.03 from pny's site. Performance is slightly better when navigating, but the tools don't cause major slow down problems like before and Modo DOESN'T crash. Woo Hoo! A bit long winded, but until some fixes are made with Nvidia's latest driver, it will save you some headaches by reverting to a lower driver version.

Man, I hope that helps.

Cheers,

Ric535
09-24-2004, 04:38 AM
Nice model BTW, great imagination. There were a few things I did. One lower the default subD level to 1. It is in the preference menu up top. Also, depending on what tool you are using, Modo could go from good to bad performance really quick, mainly this happened when I used edge slice, the one way to deal with that is to clear history, and everything was good again. You can also opt to go into subpatch mode instead of Modo's subd model, it is shift+tab, Performance will be better in subpatch. However, I still had issue with crashing, so I eventually lost it and went to a lower driver version, 44.03 from pny's site. Performance is slightly better when navigating, but the tools don't cause major slow down problems like before and Modo DOESN'T crash. Woo Hoo! A bit long winded, but until some fixes are made with Nvidia's latest driver, it will save you some headaches by reverting to a lower driver version.

Man, I hope that helps.

Cheers,
What's pny's site?, i need to get earlier drivers

Levitateme
09-24-2004, 04:38 AM
Thanks guys for all your help. im going to try what you guys said, hopefully it will work better now.

visualboo
09-24-2004, 05:59 AM
These are all unnecessary actions for a problem that shouldn't even exhist. This program is brand new so there is no reason (old dated code) it should have such bad viewport performance. IMO, this should be the #1 priority for the programmers (very nice guys btw, I met quite a few at sigg)

Toggling visibility on and off for parts of a model is unacceptable, especially when your only dealing with a couple thousand poly's. I would recommend purchasing a copy of xsi and using it to see what we mean. Speed is king on todays productions.

Btw,
well if the subd level "2" may not be same as xsi, because subd 2 is going to create alof ot polygons per subd polygon. i mean if you have 10000 subd polygon in level 2, i am guessing it means more than 150k polygons. i tested with 7400 subd polygon model on subd level2, when i froze the model it ended up with 120.000 quad polygons. i am not sure, if you think that is what xsi doing after you freeze it in xsi, then would be nice if we get some answer from luxology

i am seeing that you second model has some 40k , if that had subd level 2, and you think that it was sluggish, actually it sounds normal to me at that number. because that is like over 500k model actually

but i do have an ati 980pro here, and holds alright At the XSI user meeting at siggraph, Mark Schoennagel opened a model and cranked up the subD's to what was it? 37 MILLION!! It was still workable too. He was rotating the viewport selecting lots of verts... moving them. If modo is poopin out at 150k (haven't tested this on my system yet) that's an issue.

I'm not trying to bash anything btw, I'm just really dissapoited with this as my hope's were high.

Levitateme
09-24-2004, 06:54 AM
wow visualboo, i constantly constantly hear more praise for xsi among the people i talk to. its amazing how much i hear that... and i cant say anything truthfully bad about modo because my drivers are so dumb. so when they get "fixxed" and if modo still is as dead slow as it is, yeah then ill be really dissapointed. its actually slower than wings3d for me right now. in wings 3d i can move this model at real time just like maya, model in realtime "by that i mean no slow down at all" obviously wings is known for its slow sub d when you get a decent sized model in there. but your right for a brand new app as slow as its being, and its not just me apparently, yeah that is a pity.

kursad_pileksuz
09-24-2004, 07:32 AM
i do not think that you were bashing visualboo, and if you want to bash that si fine as well as long as they are all grounded arguments

numbers can be misleading, 37 million polygons does not mean anything about quality of tools, work flow, and creative options of the software, we heard alot of numbers from graphics cards amnufacturers , games etc they do not mean much. but we should be relying more on creative freedom software gives with optimum speed with today`s standarts. i wonder who is working with 37 million poly models ? working with 37 million poly model can be nothing but pain if you do not have tools to manipulate it. i have not used xsi, as far as i heard it is a great program, it is not just a modeler. and also i am sure nothing comes free, like in zbrush which is an awesome piece of software, but you cannot work in it with traditional ways of working with models ,like, to be able to select your model traditionally or having ability to merge 2 vertices. they are all different programs for different means.
the bottom line is, if you need to work on 37 million poly models (assuming that xsi works swith that number in any condition or without any condition) , modo is not for you currently. i do not think that any driver or card or options in ppreferences will give you such quantum leap.
on top of it, yes we need some speed improvement (s) :)

visualboo
09-24-2004, 01:04 PM
Actually numbers are pretty damn rock solid.

We have to compare apples with apples here. I said nothing about anything besides raw viewport speed, except about toggling visibility but that is directly related to viewport speed. There are some very cool tools in modo. Don't get me wrong, but that's not what this thread is about.

tjnyc
09-24-2004, 02:02 PM
visualboo,

I know where you are coming from, and I pretty much agree with your viewpoint, but like you said it is a brand new app, so there will be some bumps initially, XSI at 1.0 had terrible problems with stability and alots of bugs. I hope the Luxlogy guys do take this as a HIGH priority, as speed is generally the first thing people notice, before the tools, and that could hurt Modo. If I had no previous experience with Modo and I heard Modo had great tools but was slow, I would not have bought it and that is my honest feelings. Now that I have used it I can see the beauty of the program, I really like using it. I'm going to be patient and see what happens with this issue, and I will be watching VERY closely.


Cheers,

PaQ
09-24-2004, 02:55 PM
Wich "normal" cg card can handle a 37 million polygons model >??? I mean the faster software I know able to move/edit millions of polygones is zbrush right ???, and that's because they use the cpu/ram to render the model and not the cg card. I really didn't expect to be able to move 37 millions of poly on my little radeon 9700 128mb, whatever the software I will use !

tjnyc
09-24-2004, 03:14 PM
What's pny's site?, i need to get earlier drivers
Dude!? www.pny.com (http://www.pny.com)http://www.cgtalk.com/images/icons/icon12.gif

Cheers,

ambient-whisper
09-24-2004, 06:00 PM
Actually numbers are pretty damn rock solid.

We have to compare apples with apples here. I said nothing about anything besides raw viewport speed, except about toggling visibility but that is directly related to viewport speed. There are some very cool tools in modo. Don't get me wrong, but that's not what this thread is about.
it also has a lot of interesting "features" ;) where you are allowed to pick backpoints when tweaking, when you want to tweak what you see infront of you. its not bad on low, but on high models its sooo frustrating. and the piano setup for modelling is awesome too ;)

kursad_pileksuz
09-24-2004, 06:10 PM
i understand that numbers are damn rock solid, but we all have different standarts, that was my point. i would use xsi, not because it can throw 37 million, but if it had more falloffs beside screen falloff.

sure i understand that, as i said i think things need to be 2 times faster than right now, but if you are generally between 1k to 20k subd models speed is no problem. for higher modesl i agree about speed. i was trying to say that we cannot blame modo because it cannot throw 37 millon polygons.

Actually numbers are pretty damn rock solid.

We have to compare apples with apples here. I said nothing about anything besides raw viewport speed, except about toggling visibility but that is directly related to viewport speed. There are some very cool tools in modo. Don't get me wrong, but that's not what this thread is about.

visualboo
09-24-2004, 06:31 PM
@tjnyc
Yeah, I understand v1 of any app is not going to be perfect. But this (IMO) should not have gotten through development as far as it has. I agree with your post 100% btw. I still see great things ahead for modo and lux but I guess what I'm trying to get at is don't get so used to these "temporary" workarounds that you forget to keep on lux about this issue. It NEEDS to be addressed.

My stash of pennies thrown in is quickly rising, so I'll stop here :D

Levitateme
09-24-2004, 06:36 PM
oh, fyi about visualboo. no i had no ill comments toward his post if thats what you meant. yeah i honestly do hear that about xsi, and its just suprising to me how much i hear it. yeah no bashing him in the slightest. i was being serious, i was really tired wheni posted that, so when i read it now it is not sincere like i meant it to be.

CGTalk Moderation
01-19-2006, 07:00 AM
This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.