topic Re: Banded photo rendering again in Visualization

Banded photo rendering again

Scott Bulmer — Thu, 11 May 2023 10:34:57 GMT

I apologize if this is not the correct forum for this but here goes; I have tried everything I know to fix this problem (see attached banded image) and now need professional help determining if it is the graphics card. We have the latest driver for the card & I've set the hardware acceleration off in both the driver settings as well as Image Settings in AC. We have also messed around with the Virtual Memory settings but nothing seems to work. The model is only 16MB. The banded images do not happen the first rendering attempt after opening AC, but every time after that. The Photo Rendered image is fine every time using the built-in sun, however renders banded with the Sun & Sky objects, even set with low amounts of light sources. Although our card should be good, I will get a new card if required. Any other thoughts are sincerely appreciated as my back is against the wall. Our workaround for this is to restart AC9 each time we want to generate a LW rendering. Restarting AC resets something, who knows what.
Thanks,
Scott

Re: Banded photo rendering again

Scott Bulmer — Tue, 10 Jan 2006 02:14:54 GMT

Further information;

Using WinXP and a 16MB PLN, we are able to render in LW up to a point. Using the marquee for portions of the building, renderings are OK. But through experimentation, continuing to enlarge the marquee area eventually causes the resulting image to break down into horizontal banding. Further, if higher quality image selections are made, the banding becomes worse.

On a positive note however, we have been able to complete renderings from the same file using the LW engine on an old Power PC G4 867 MHz w/ 640 MB ram. The only downside is that it takes approximately 40 minutes, but comes through. Upgrading the ram would most likely help speed things up, I think or hope. Note: we can get about 90% of the same model with XP in about two minutes. More than 90% gets banding.

So the end result of all this is that, I guess there is some type of memory limitation to WinXP that OS X does not have. Maybe it’s the difference between 32 and 64 bit processing (which I don’t pretend to understand), but have been told is the main difference between the two architectures. Perhaps someone knowledgeable on this topic could share some “Light.” That’s all I have for now as I believe the problem's solved.

Re: Banded photo rendering again

Dwight — Tue, 10 Jan 2006 04:03:38 GMT

You seem to be the only guy in the universe with this problem.

There would be plenty of whining if a common operating system failed to allow a render.

Can it really be a software issue?

Re: Banded photo rendering again

Anonymous — Tue, 10 Jan 2006 04:21:29 GMT

Maybe not,

I have had similar experiences but the whole render disappears in XP while the mac continues on it's merry way, only in LW though. If I use the internal rendering engine it will render on XP but again it takes forever.

I do know that a 'problem/error' in the file can create blank renders but seems strange that the same file can produce 2 different results on 2 different systems, maybe that's just it though!

One product, 2 systems, many rendering solutions!

is it is in HK:

One Country, 2 government systems, many dining choices!

Re: Banded photo rendering again

Djordje — Tue, 10 Jan 2006 07:56:15 GMT

Dwight wrote:
You seem to be the only guy in the universe with this problem.

Nope, the same was reported by Kim some time ago ...

http://archicad-talk.graphisoft.com/viewtopic.php?p=48825#48825

... so this is not an isolated case.

Re: Banded photo rendering again

Dwight — Tue, 10 Jan 2006 08:03:00 GMT

Two, then. (avoiding obvious gender joke)

I can see that I am no match for your archive skills.

What is common between their systems?

Re: Banded photo rendering again

Gerald Hoffman — Tue, 10 Jan 2006 21:36:29 GMT

I have also had this happen recently on a big model. When I took off the Roof Tiles applied with the CadImage Accessories Add-on I was able to render the image.

Are you using roof tiles by any chance? I don't know if the add-on is the problem or just the large number of polygons of the roof tiles throwing the proverbial spanner into the works.

Re: Banded photo rendering again

Scott Bulmer — Wed, 11 Jan 2006 01:48:01 GMT

Gerald wrote:

Are you using roof tiles by any chance?

When this first became an issue several months ago, we were using the 'Roof Surfacer' that came with AC. After eliminating the RS objects, the model began to render fine again and we thought that was the end of it. Now however, the model has grown a little and the problem's back as I've described below. I guess for XP users, don't place too many curving shapes or objects with detail in your model as they will chew up memory as it relates to LW when using the Sun/Sky objects. OS X users, party on.

Scott

Re: Banded photo rendering again

Dwight — Wed, 11 Jan 2006 02:04:43 GMT

Just out of curiosity, now that we are getting closer to this being a RAMulon issue.....

Post your sun object and sky object settings - you can easily overdo it with these objects.....

Re: Banded photo rendering again

Scott Bulmer — Wed, 11 Jan 2006 04:48:01 GMT

Thanks Dwight. For your perusal:

Over sun sources = 6
DA = 45
Shadows = on

Under sun sources = 4
DA = 45
Shadows = off

Sky object
Lat = 3
Lon = 6
Shadow = on
Soft = off
S Quality = 2

Re: Banded photo rendering again

Dwight — Wed, 11 Jan 2006 04:54:43 GMT

Reasonable values, all. Duh.

Re: Banded photo rendering again

Scott Bulmer — Wed, 11 Jan 2006 05:14:51 GMT

Dwight,
Thank you agin for your valiant persistence, but I think this one is about to go nighty night. I'm not an expert but am told that Microsoft is beta testing their newest OS.... 'XP-x64.' I guess they're following Apple. For our complex building model, we've resolved to do all our renderings on the G4, which seems to work fine.

Again, you comments are appreciated.

Scott

Re: Banded photo rendering again

Scott Bulmer — Sun, 22 Jan 2006 20:01:59 GMT

Update:

Visited our new local Apple Store yesterday. They allowed me to load AC9, then update to the latest build onto a Quad 2.5 GHz PowerPC G5 with 4.5 GB ram. I loaded our 80 unit condo project PLA and attempted to render using LW.

I was prepared to purchase that machine believing it would solve our rendering problems. The identical banding issue however repeatedly occurred as with our dual PC. That leads me to think that for complex models (based on polygon count), LW has a problem functioning within dual processor systems.

I am waiting to hear back from GS for ftp upload instructions for their assessment. In the meantime, I seem to recall that there used to be a way to calculate a model's complexity relating to image creation. Could someone please refresh my memory on how to do this?

Thank you again for putting up with this.

Re: Banded photo rendering again

Dwight — Sun, 22 Jan 2006 21:06:40 GMT

I've simply GOT to have that model!
Figure out a way to send it to me!!!!

As for model size, the Special Menu can generate a report of the 3D model size.

Re: Banded photo rendering again

Dwight — Sun, 22 Jan 2006 23:48:07 GMT

Okay. 65 megatron PLA received through FileFront.

There is a problem with the file - as received, I couldn't render it at all. A marquee of the actual building, ignoring the outlying modules and suite layouts solved the problem.

Here's a perspective of the building rendered without banding on a Mac 2x2.0 machine with 3.5 Gigabigs.

Note error report on OpenGL..... whatever that means...

REPORT:
----------------------------------------------------------------------
Projection : Custom axonometry
Camera azimuth : 295.121
Start time : Sun, Jan 22, 2006 15:35:7
End time : Sun, Jan 22, 2006 15:35:10
Elapsed time : 3 seconds
----------------------------------------------------------------------
Internal error occured in the OpenGL 3D engine.
----------------------------------------------------------------------
LightWorks Rendering Engine
Initialization : 0.00 s
Model generation : 65.83 s
Rendering : 273.74 s
Termination : 0.00 s
...................................................................…
Elapsed time : 339.58 s
----------------------------------------------------------------------
###
ONBODIS 23720 BODIS 3225920 NTRANS 1431 TRANS 148824 UBODIS 1518080
NPGONS 307482 PGONS 9839424 UPGONS 25421536
NPEDGS 1287558 PEDGS 5150232
NEDGES 671586 EDGES 16118064 UEDGES 2058970 NHEDGES 0 HEDGES 0
NVERTS 482559 VERTS 11581416 UVERTS 0 NHVERTS 0 HVERTS 0
NVECTS 242260 VECTS 5814240
NTEVES 0 TEVES 0
NHOTSS 46666 HOTSS 1493312
NHPGONS 0 HPGONS 0
NHPVERS 0 HPVERS 0
NLGHTS 195 LGHTS 13260 ULGHTS 10140
NRTXTS 14 RTXTS 840
NUFILS 0 UFILS 0
NUMATS 151 UMATS 15100
NFIDES 0

--------------------------------------------------------------------------------------------------

Model 54295544
Projected 29008726
Cpols 2144856
Tpixs 346880
--------
TOTAL 85796006
###
- - - - - - - - - - - - - - - - - - - - - - - - -
Data structure sizes by element types:
all cut
WALLS 1003620 4294967295
BEAMS 34144 4294967295
- - - - - - - - - - - - - - - - - - - -

Re: Banded photo rendering again

Scott Bulmer — Wed, 08 Feb 2006 16:40:15 GMT

David Tauer of Graphisoft wrote:

From the sounds of your problem it does appear to be memory exhaustion. There are a few things you can do that may get you past the banding and not have a "drastic" change in the render you desire. (I would try them first individually in the order given)

1) change Antialiasing from "Best" to "Better"
2) reduce the size of the render from 17" x 10.19 to something smaller - perhaps cut off an inch or so from the overall size.
3) Change from soft shadows to hard shadows (this is a big memory user)

After looking at the project, I am now 100% convinced it definitely is memory. When you see "Internal error occurred in openGL" that also can indicate that the openGL is having a hard time with memory. That warning also came up. Scott clearly has a large monitor. Those picture sizes are using up memory. A machine with two processors can use up more memory. What happens is LightWorks divides its tasks amongst the processors. Each processor starts using up memory. If the memory is there then no problem, the time is more than halfed. The problem arises when one processor subscribes memory and cannot complete a task due to the memory subscribed by the other processor. That is, the amount of memory used over the period of the render is not significantly different between one processor and two processors machines, but HOW much memory used at one time is. And this could be significant. Which means that a two processor arrangement could exceed addressable memory (virtual and physical) due to the scheduling of subscribing and releasing memory whereas a single processor would not. By the time a single processor system gets to the next task it has already released the memory it needed to complete the task. Hope this helps.

Dave,
Aside from doing the things listed above, will I be able to run AC10 under the new XP-64 operating system? And will the new OS allow for less memory exhaustion?

David answered:

AC9 currently runs under the XP-64 bit OS so I don't see any reason why AC10 would not. That new OS will better utilize the memory - almost 4G instead of 2G limit on Win XP. The Win XP64 OS will utilize almost 4G of RAM.

On a Mac, Archicad can access a little shy of 4GB of memory (physical and virtual). So if the user has more it doesn't matter. On an XP using the 32-bit operating system version Archicad is limited to 2GB.

Here are some suggestions in conserving memory

1) ***** Window size in pixels **** very important
2) **** Window size in pixels ***** in case you forgot number 1
3) OpenGL options Turn off: Highlights, Emissions, Smooth Surfaces, Textures (play around with which ones are important)
4) Contours "draft"
5) Transparency in Shading set to off

Then limit model size, use marquee, use layers, leave out memory hogs like cars and 3d trees and bushes (or put them on layers that you turn on/off in context with a marquee) The Internal Engine was developed when memory was scarce and 256MB was a killer system. It no doubt is more conservative with its memory allocation algorithms, it also doesn't sport all the bells and whistles. The openGL method relies much more on technology supplied by external vendors, and is definitely not as stingy with its memory subscriptions as the Internal engine.

Thanks David for all your help and clarifications.
Sincerely,
Scott