Blender 2.63 fastest build with CUDA support, now with all CUDA kernels (except version 3.0)....
After having moved my HD to another board, I've had a hard time setting everything up again, but the wait is over....
Here is Blender Fastest build, now compiled with all CUDA Kernels, for complete GPU support.It doesn't matter whether you just got a brand-new Nvidia Card, or (like me) you're happy with your 9800 GT: this build should work with any CUDA-Capable card (except those that use CUDA kernel version 3.0).

About old GPUs and CUDA problems:
If CUDA doesn't work with your GPU, go to Render panel, and change Feature Set, from "Supported" to "Experimental".
About the Contrib Add-Ons
The Contrib Add-Ons are now in a separated category called "Testing"
http://wiki.blender.org/index.php/User:Ideasman42/ProposalAddonsContrib
I've carried out render tests of different scenes using Blender's Internal Rederer, and comparing this to many other x64 (and even x86 optimized) builds, and renders were always faster using this optimized build, that's compiled right from the trunk, therefore without any cut-outs.
Tested on Windows 7 x64, Core 2 Duo E7500 running at default clock. (2.93 GHZ)
Please, test it in your machine and give feedback about its performance.
Enabled features:
Built-ins:
Bundled Python
Game Engine
Blender Player
Cycles Render Engine (CPU and CUDA)
Codecs:
FFMPeg
libsndfile (not tested)
FFTW3 support (Used for smoke and audio effects)
Imaging:
CINEON and DPX Image Support
DDS
image FrameServer Support for rendering
HDR Image Support
OpenEXR
OpenJpeg
RedCode
TIFF
Other features:
Physics Engine
3D Mouse
Best LZMA compression, (used for pointcache)
Fast LZO compression (used for pointcache)
OpenCollada Support
OpenMP support, for Parallel code execution
SSE optimizations for the ray Tracer (not tested if it's actually effective)
Samplerate conversion
SDL support (sound and joystick)
International fonts and text (I18N)
Features that aren't available:
Jack support, QuickTime
LATEST CHANGES:
|
Revision 46708 by Joshua Leung
2 hours 15 min ago
|
|
Nitpicky stuff
* Ensure "Base" is fully nullified before anyone uses it * Force channel flush when changing Action Editor modes |
|
Revision 46707 by Joshua Leung
2 hours 15 min ago
|
|
KeyingSets UI - Improving button layout for legibility of buttons
* Use custom names for the keying options enum. The old automated layout dumped out lengthy names which didn't work well with the limited UI widths we were using, leading to all the interesting stuff getting cropped out * Moved the "array target" options to be in line with the rest of the target specification stuff. I've ended up flattening that set of options into a single row, which seems to work quite well. * Removed label from grouping method enum. There was a perfectly good section header above, and with narrow UI's, the name of this and the one below ended up looking the same/confusing. |
|
Revision 46704 by Campbell Barton
5 hours 27 min ago
|
|
style cleanup: block comments
|
|
Revision 46703 by Nicholas Bishop
6 hours 37 min ago
|
|
Fix for clay strips brush not showing Add/Subtract buttons.
|
|
Revision 46702 by Bastien Montagne
8 hours 29 min ago
|
|
Fix [#31432] Angles values stuck at 60 degrees.
|
|
Revision 46699 by Campbell Barton
12 hours 28 min ago
|
|
fix warning in last commit
|
|
Revision 46698 by Campbell Barton
12 hours 36 min ago
|
|
evil mango feature, alt transform sequence strips bumps other strips out of the way
|
|
Revision 46697 by Bastien Montagne
14 hours 3 min ago
|
|
Nodes i18n: don't call gettext stuff twice for categories' names, when building Add menu.
Now ui-names in foreach_nodeclass functions are only marked as translatable, it?s up to the callback to actually translate them or not! |
|
Revision 46696 by Brecht Van Lommel
14 hours 28 min ago
|
|
Fix issue with linked dupli objects not being displayed in the viewport, if they
have no object bounding box. Better fix will be to figure out why they are the bounding boxes are not made. Thanks to Campbell to tracking down the commit that caused this. |
|
Revision 46695 by Antony Riakiotakis
14 hours 34 min ago
|
|
modal operation for bevel and inset:
This commit adds some first building blocks for the two operators to work modally based on mouse input. To make their function easier, two hotkeys are introduced, Ctrl+B for bevel and I for inset. TODO: After discussion with Campbell, we would like to add scale-style line indicators for the operators. This is already done for transform operators but a new interface for mesh operations may have to be written using pieces from that code since, strictly speaking bevel and inset are not exactly "transform" operators. Also, a better input method for inset is needed and more options exposed. The method implemented right now uses mouse move for thickness and ctrl-mouse move for depth. These are calculated using the distance of the selection center in screen space and the mouse position. While that may work and prevents abrupt changes in values when switching from thickness tweak mode to depth tweak mode, it limits the magnitude of values that can be put into the tool especially in small or large scale. Alternatives until a better method is written include: * use relative offset (works but may give strange results) * tweak manually after the operation. |
|
Revision 46694 by Bastien Montagne
14 hours 39 min ago
|
|
Spell check fix (lenght -> length)
Note: two remains, in comments in extern/bullet2/src/LinearMath/btVector3.h and extern/libmv/libmv/image/tuple.h. |
|
Revision 46693 by Daniel Genrich
16 hours 46 min ago
|
|
Fix compile error with cmake + windows (reported by MiikaH)
|
|
Revision 46692 by Campbell Barton
16 hours 50 min ago
|
|
style cleanup: indentation
|
|
Revision 46691 by Gaia Clary
17 hours 44 min ago
|
|
fix [#31320] Collada now supports import/export of loose edges (edges not attached to faces)
|
|
Revision 46690 by Nathan Letwory
18 hours 40 min ago
|
|
The first build was apparently incorrect. A small fix and updated libs now have mingw64 builds working correctly on export.
Thanks to Gaia-Clary for assisting in the bughunt and fix. |
|
Revision 46689 by Lukas Toenne
18 hours 54 min ago
|
|
Fix #31369 (discussed in bug report comments). Setting the path/layer name strings of the file/layer slots of the file output node would crash, due to missing id.data pointer in the PointerRNA. Solved by using the iter->parent.id.data pointer for the RNA collection iterator.
|
|
Revision 46687 by Campbell Barton
19 hours 38 min ago
|
|
style cleanup: imbuf
|
|
Revision 46686 by Lukas Toenne
19 hours 42 min ago
|
|
Fix #31472: displace node wrong y derivative (simple typo). Credit to Ervin Weber (lusque) for the fix.
|
|
Revision 46684 by Campbell Barton
20 hours 22 min ago
|
|
fix [#31474] Crash when object has parent and dupli_type == 'FRAMES' and show_x_ray, show_transparent are set to True
dupli objects should never be added to after-draw, added an assert to ensure this (and make fixing such bugs easier). |
|
Revision 46681 by Campbell Barton
21 hours 26 min ago
|
|
style cleanup: imbuf
|
Note 1: It requires Microsoft Visual C++ 2008 SP1 Redistributable Package (x64) to be installed
Note 2: A small add-on is you can launch Blender from AN_Priority.bat (located in the same folder than the Blender executable).
This simply runs Blender with priority set to "ABOVENORMAL", and is useful if you have heavy tasks running in background, and want Blender to have precedence to use processor's resources.As stated above, the bat is just a small add-on: It isn't, in any way, the reason why this build is faster than the other ones.
Still, the render tests I've done were performed WITHOUT launching Blender with high priority.
I first tried a BI rendering of a complex scene with many shadows and reflections :
1) r46561 by alb3530 : 16.80s
2) r46461 official 2.63a : 18.60s
3) r46534 by bat3a : 19.15s
Then tested the BMW1M Benchmark
1) r46534 by bat3a : 29.81s
2) r46561 by alb3530 : 35.28s
3) r46461 official 2.63a : 40.82s
Interesting results... and many thanks to all builders !
http://wiki.blender.org/index.php/Dev:Ref/Release_Notes/2.63a
http://wiki.blender.org/index.php/Dev:Ref/Release_Notes/2.63a/log
hmm not sure about that, me no computing engineer lol!
but since now haven't have the support for Cuda 3.0 kernel,
it's good time to save up the money if u intend on getting either GTX 680 or 690
it's kinda meh if u buy either of the cards and found out it doesn't have the suppport for ur fav. apps/programs
like, u buy a Bluray player but there's no Bluray disc for ur fav movie yet.
Thanks for the reply.
What i meant, in fact, was about the Cuda version : does 3.0 cards support other Cuda versions ?
(i have now just a GT 330m... but i envisage to buy another pc, just for blender... so 680, or why not 690, could be a solution... but only if blender can work with them !!)
it just mean support for cards using 3.0 kernels (currently GeForce GTX 680 and 690 only) doesn't have much support yet,
since these 2 cards are expensive, most here in release builds/releases haven't got their hands on them
just wait for those developers who can afford these card to do their magic
For small scene, all is ok and fast (3 times faster than CPU). But with a bigger one, windows crashes : blue screen.
Just a question concerning Cuda support : "all but 3.0"... does it mean that 3.0 kernels could cause problem in the future (or for the official releases) ?
roll back to previous driver version..
for my case, 270.61
now real-time viewport rendering with GPU works fine (both Supported & Experimental)
all the time :(
im using it on my laptop (which sadly is the only machine i have)
updated to the latest driver
selected GPU for Blender via both NVIDIA Control Panel and Blender User Preferance
Oh, the problem is already fixed on 46443 and probably some builds below. Sorry for the trouble
Yup, I'm using Windows 7 Home Edition but I don't have that problem in the older version which I also have been using.
Best regards
Best Regards
I mean, when I select an object, the default is "Single Select" but when I press shift and select an object, "Multiple Select" would forever activate even though I not holding Shift anymore.
CUDA error: Unknown error in cuCtxsynchronize()
here's a screenshot that show the problem
http://www.hotimg.com/direct/Rbb2j3B.jpg
Blender will crash everytime i select GPU..
Problem Event Name: APPCRASH
Application Name: blender.exe
Application Version: 2.6.3.0
Application Timestamp: 4fa468fc
Fault Module Name: nvcuda.dll
Fault Module Version: 8.17.12.9610
Fault Module Timestamp: 4f4e56a8
Exception Code: c0000005
Exception Offset: 000000000004a4d5
OS Version: 6.1.7600.2.0.0.768.3
Locale ID: 1033
Additional Information 1: 70f7
Additional Information 2: 70f7ea30317dc4c839d5242da8037c48
Additional Information 3: 99b0
Additional Information 4: 99b0eefefc776abdfc0c9b91eb942ce1
Best regards
Can you confirm?
Best regards
Cuda error : Invalid value in cutexRefSetAddress(NUL,texref,cuda_device_ptr(mem.device.pointer),size)
???
:-)
(sorry for my english xD)
My graphic card is gtx 570 and i'm using windows 8 consumer preview, drivers 296.17 . Is there any way to solve this?
its a very long error message but the bottom says this: 3 errors detected in the compilation of C:/users/robert/appdata/local/temp/OCLA/96c.tmp.cl/ Internal error: clc complier invocation failed.
What error exactly do you get in the console?
Best regards
Maybe there's some way to do compile using different toolkits i.e. 3.0 with Cuda 4.1 and other kernels with Cuda 4.0....
Best regards
You could try DingTo's build, he compiled the 3.0 kernel for the 680
http://graphicall.org/888
Cheers,
~Tung
Best Regards
Please respond - anyone experienced such annoying bug:
After switching to another window type eg. from 3D to UV -> abnormal termination and...out.
OR (and thats interesting) after splitting the display horiz. and/or vert. and assigning 3D to the first window and UV to the other - crashes too, ....BUT: only when UV had been assigned first and stays as Upper or Right window, otherwise seems ok.
My desktop: Win7 x64 Premium 8GB RAM, Nvidia GTX260 latest drivers.
Podpis problemu:
Nazwa zdarzenia problemu: APPCRASH
Nazwa aplikacji: blender.exe
Wersja aplikacji: 2.6.2.0
Sygnatura czasowa aplikacji: 4f58aa23
Nazwa modułu z błędem: python32.dll
Wersja modułu z błędem: 3.2.150.1013
Sygnatura czasowa modułu z błędem: 4d6e5b4a
Kod wyjątku: 40000015
Przesunięcie wyjątku: 000000000012c1fa
Wersja systemu operacyjnego: 6.1.7601.2.1.0.768.3
Identyfikator ustawień regionalnych: 1045
Dodatkowe informacje 1: 838b
Dodatkowe informacje 2: 838b9268a341d4219ae2b63ba1356f2a
Dodatkowe informacje 3: 4bfa
Dodatkowe informacje 4: 4bfa31904dfc023146fde5e727b60bd5
I have been unable to find a single build of Blender 2.6+ that can load Collada without crashing.
Thanks
I have a GTS 250, I checked experimental GPU, but it is "out of memory" or the texture is black, is it still called Cuda 1.3 (my cuda is 1.1). This is in the window, or rendering. Yet it works with a simple cube to boot windows ...
For the stage of the BMW I can enable rendering to the window but I can not make the scene ...
In the render if I put on the display image editor, in rendering if I press the escape key blender crash ... (all versions since BMesh)
Windows 7 professional 64bit, GeForce GTS 250 512Mo, 2 screens, AMD Quad core and 8 Go RAM
It seems to me that it's not actually using the GPU unless I select "experimental" feature set, just in case anyone has the same issue.
I have the very latest Nvidia drivers, released on Feb 21 2012
Best Regards
9800 GT is my main board here... :D
Best Regards
Thankyouthankyouhankyouhankyouhankyouhankyouhankyouhankyouhankyouhankyouhankyouhankyouhankyou.
You ROCK man! :D
Here everything's ok
Best regards
I got this to see if I could get Cycles to run on my GPU (Nvidia NVS 4200m, CUDA 2.1) but it only shows OpenCL in User Prefs (and that fails because version is only 1.0).
Does anyone have any suggestions? Is there a problem because my system has Optimus? (Intel integrated graphics+Nvidia NVS 4200m) I have set Blender to always use Nvidia...
David
Sorry that I've missed the latest changes in the Blender builds about my problem the last time I've posted, I missed the new options on User Prefs from
User Preferences > System > Compute Device > CUDA > select your GPU, For users too that missed out this option this was just the problem...
Could someone set up a scene with three cubes, each on three different layers, and try to render that? Maybe post the file? From what I can tell, this should work and you should be able to render separate layers out just fine.
Could someone set up a scene with three cubes, each on three different layers, and try to render that? Maybe post the file? From what I can tell, this should work and you should be able to render separate layers out just fine.
But for what you say, it seems like some limitation of Blender...maybe it will be available in future revisions
Best Regards
The preferences allow the use of one of the available GPUs, or all the GPUs, but it is not possible to use Two of Three available GPUs for example. I have one GTX 260 for display, and Two GTX 580 3GB for rendering, and I can't use the Two GTX 580. Except if I select all GPUs, but the rendering time is equal or worse than with only one GTX 580, and I think that the availlable memory amount will be the amount of the weakest card, so I would like to use the two big cards and exclude the GTX 260 (like I do with Octane). I tested with Cuda. OpenCL gives only a black screen.
My config : Win7x64Pro. IntelQuadXtreme9650.3GHz. 8GB RAM. One GTX 260 for display. Two GTX 580-3GB for rendering, in a Cubix-Xpander Pro2.
use blender 2.61 official for the city gen.
I would expect it to be updated for blender 2.62 offical.
Traceback (most recent call last):
File "C:\Program Files (x86)\BlenderGraphicAll\2.61\scripts\addons\suicidator_city_generator_0_5_3_Full\__init__.py", line 360, in execute
SCG_output = imp.load_module('SCG_output', file, pathname, description)
File "C:\Users\ABSPQI~1\AppData\Local\Temp\SCG_output.py", line 21, in
SCG_build_terrain()
File "C:\Users\ABSPQI~1\AppData\Local\Temp\SCG_output.py", line 11, in SCG_build_terrain
clearMeshOfObject('SCG_terrain')
File "C:\Program Files (x86)\BlenderGraphicAll\2.61\scripts\addons\suicidator_city_generator_0_5_3_Full\__init__.py", line 100, in clearMeshOfObject
bpy.ops.object.select_name(name=objectName)
File "C:\Program Files (x86)\BlenderGraphicAll\2.61\scripts\modules\bpy\ops.py", line 180, in __call__
ret = op_call(self.idname_py(), None, kw)
AttributeError: Calling operator "bpy.ops.object.select_name" error, could not be found
location::-1
error DOES NO T occur in the latest official release
on experimental 01:18.45
not bad, this beats bat3a's build...
thanks ! i think this will be going faster as
the devs might introduce new ways of
optimization...
more power on your builds :)
Best regards
http://www.pasteall.org/pic/show.php?id=25901
I just noticed that in the newer builds the old selection CPU/GPU combo box is changed to Supported/Experimental... sorry for that I was using the old builds (Rev 42885) which was more faster when rendering cycles in preview mode using GPU.
and this is my test scene I uploaded it here - http://www.pasteall.org/blend/11434
I've got reports CUDA is working nicely....
In my machine that's also fine...also using Experimental, by the way....
Is it performing slowly in all scenes?
Best regards
Best regards
problem seems to be solved!!!
I've uploaded a test szene here: https://rapidshare.com/files/1225521933/untitled_TEST.blend
Rendertimes with your build 5889: 1:14:39
with build 5895: 0: 33:44
greetings, ponc
A .blend file, and if not asking too much, a render time comparison between mine and other x64 build...
I've tested here with some scenes, and render times aren't slower in this very build actually...But I'm on a 9800 GT (old GPU)
Best regards
@alb3530 - it occured to me that maybe I'd see if passes render with CPU and guess what, it worked. Still, it refuses to deliver them with GPU on my GTX460SE and CPU rendering is dead slow compared to GPU. ;]
My blender works again. All is well with the world. Thanks to the devs and all you guys who are "just putting together builds". As far as I'm concerned, you could all be saying "I'm just a brain surgeon- I didn't actually create the brain"!
I have a question about cycles multi gpu support: I have a geforce 590 (dual gpu) + 8800 gtx, i use the 8800 generally for physics and other quick calcs, i would like to set the render to use the 2 gpus of the 590 and not the 8800 (low memeory causes problems) i found out i can only use the whole 3 or just 1, how can i set it to use the 2 gpus of the 590 and not the third of 8800?
Thank you!
Reference:
http://projects.blender.org/tracker/index.php?func=detail&aid=30016&group_id=9&atid=498
Best Regards
I could reproduce (and fix) this problem, but in a odd way....
1 - Started my machine from a virtual disk (Containing Windows 7 x64)
2 - Started Blender, set Geforce 9800 GT as the compute device in "User Preferences"; and saved settings;
3 - Turned the machine off and started a virtual machine with the very same virtual disk (Given that virtual machine can't see PCI devices, so the GPU is unavailable to it).That simulates the removal of the card.Maybe changing the GPU model would cause the same issue;
4 - Went to Blender "User Preferences", which caused blender to crash instantly (exception code: C0000005)
5 - Started Blender again, and loaded factory settings.After that, I could open all tabs of "User Preferences" screen.
Best regards
Best regards
Conz
Will rebuild and upload the working build within just a few minutes...
matisso: please upload a .blend file you're having problems with (after making sure the problem is in this specific build, of course....)
Delta: I wasn't able to reproduce it here....I can actually go to System tab in user preferences...Try to reset Blender settings, and tell me if it works for you
Best regards
In the meantime, have you tried another build to see if you get the same issues?
If possible, can you share the .blend file you're having problems with?
Best regards
CUDA error: Unknown error in cuCtxSynchronize()
CUDA error: Unknown error in cuMemFree(cuda_device_ptr(mem.device_pointer))
The second line is being repeated about ten times. After these errors, GPU and memory clocks are severely set back, like it was messing with a driver in some way, and they won't go up upon next render, or even launching stable build. I have to reboot my PC, maybe reloading video driver would help, but didn't check it.
cheers,
Matt
- Scene with default cube
- switch render engine to "Blender Game"
- Press the button "render" -> "standalone player" -> "Start"
Than there should be the standalone player coming up. In the build bot version it's ok. But in your version the player doesn't show up, but is running without any window.
You have to kill the proces "blenderplayer.exe" to reactivate blender. Until this blender seems frozen (is waiting only to the ending of the player)
Will try your r43736 now.
I have installed:
Microsoft Visual C++ 2005 Redistributable
Microsoft Visual C++ 2005 Redistributable (x64)
Microsoft Visual C++ 2008 ATL Update kb973924 - x64 9.0.30729.4148
Microsoft Visual C++ 2008 ATL Update kb973924 - x86 9.0.30729.4148
Microsoft Visual C++ 2008 Redistributable - KB2467174 - x64 9.0.30729.5570
Microsoft Visual C++ 2008 Redistributable - KB2467174 - x86 9.0.30729.5570
Microsoft Visual C++ 2008 Redistributable - x64 9.0.21022
Microsoft Visual C++ 2008 Redistributable - x64 9.0.30729.17
Microsoft Visual C++ 2008 Redistributable - x64 9.0.30729.4148
Microsoft Visual C++ 2008 Redistributable - x64 9.0.30729.6161
Microsoft Visual C++ 2008 Redistributable - x86 9.0.21022
Microsoft Visual C++ 2008 Redistributable - x86 9.0.30729.17
Microsoft Visual C++ 2008 Redistributable - x86 9.0.30729.4148
Microsoft Visual C++ 2008 Redistributable - x86 9.0.30729.6161
Microsoft Visual C++ 2010 x64 Redistributable - 10.0.40219
Microsoft Visual C++ 2010 x86 Redistributable - 10.0.30319
I've tested Cycles viewport rendering before uploading the build, and it worked just fine...Doesn't it work for you as well?
Best regards
@alb3530: I've posted recently on bat3a x64 thread same issue: from 2~3 weeks till todays builds I cannot fully use Blender and Cycles. Blender crashes when: with Cycles activated I click on 'Render' tab in standard right panel, press F12, press Ctrl+Alt+U and click 'System' tab ->all crashes related to CPU/GPU/Cycles. I have all MSVC++2008 installed, latest nvidia drivers, Cuda 4.1 toolkit. What can cause the problem in Your opinion? Older Cycles builds works.
http://naforum.zapodaj.net/b33f10caf6e7.png.html
When it would be fixed or can I do it on my own
Best Regards
There is a problem with the new "render" -> "standalone player" -> "start".
The player will not be visible, it is running in background and blender looks like frozen. (it simply wait that you end the not visible player)
So you have to kill the process with task manager.
I have tested this version: http://builder.blender.org/download/
And there the player is working fine. So it's a problem with your build.
Slower rendering with recent builds when using GPU (CUDA) should be adressed soon.It's up to Blender developers though, not me.
Best Regards
Unfortunately, with certain GPUs, it won't work until I have everything set up again.
Best regards
Anyway thanks a lot for your great builts!
2. I get the same cuda error as mentioned before CUDA error: Invalid value in cuMemcpyHtoD(mem, host, size)
my gpu is a GTX260
@alb3530:
Thanks a lot! :D no Cuda Problems anymore. works well with 2x GTX 570.
Best regards
Do do you have in plane to add multy Gpu support, it is possible ??
Revision number is now updated
Best Regards
great, CUDA works again (for me).
Thank you for your effort!
Best regards
P.S.: In the build overview list this build still shows "r43575" under the windows icon on the left side, although inside it is already r43660.
Uploading right now...
Best regards
First: Thanks for your builds, they're always good and fast :)
Second: I got the same error with cuMemcpyHtoD
CUDA error: Invalid value in cuMemcpyHtoD(mem, host, size)
Running GTX460 w/latest driver 285.62, Win7-64bit
Thanks!
Best regards
Cycles GPU rendering (CUDA) on my GeForce GTX 560 Ti always worked, however in build 43575 I now get
CUDA error: Invalid value in cuMemcpyHtoD(mem, host, size)
It shows up in the viewport after the "Building BVH..." phase and on the console. Happens also when rendering the scene with F12. CPU rendering however is fine. Seems to be independent from the scene (happens with default cube also).
NVIDIA driver version is 290.53, but I didn't change the driver since testing the build before this one, where CUDA rendering still worked fine. I am running Windows 7 Home Premium SP1 64bit.
Is anybody experiencing the same problem?
Best regards
...btw what the DS in Your username means?
i have GeForce GTS 250 1.1 Compute Capability
in system console is says " Cuda device supported only with compute compatibility 1.3 or up, found 1.1 "
" it is recommended to use a GTX 4xx or GTX 5xx card (shader model 2.x), since only those are likely to give good speedup, earlier cards are often slower than just using the CPU." http://wiki.blender.org/index.php/Do.../GPU_Rendering
with my GTS 250 1.1 computing i rendered at twice-speed faster (when it worked) than with my i5 720 4-core CPU...
does it means that form now on i won't be having gpu rendering on my machine ?
here's the screenshots of rendering:
1. experimental http://www.pasteall.org/pic/show.php?id=24359
2. supported http://www.pasteall.org/pic/show.php?id=24361
http://wiki.blender.org/index.php/Dev:2.6/Source/Render/Cycles/OpenCL
tells you what you need to get shaders to work.
Many thanks anyways... HAPPY NEW YEAR TO YOU... in advance
Check this one: http://graphicall.org/767
Best regards
Best Regards
dbn: Maybe you should report it as a bug using Blender's bug tracker then.... Best regards
Hi, i did what you suggested and yes, it happend to be a bug so Sergey Sharybin (nazgul) made a patch for it. Is it possible for you to make a build with it? If so you can grab the patch here in the attachments section:
http://projects.blender.org/tracker/index.php?func=detail&aid=29671&group_id=9&atid=498
Thanks in advance
Best regards
2.61.0 r42615 official release.
2.61.0 r42667 tomato branch
2.60.7 r42494
2.60.5 r42025
dbn: is it working in other builds?
Best regards
http://www.microsoft.com/download/en/details.aspx?id=2092
http://www.pasteall.org/pic/show.php?id=22396
As for the display driver issue, it's probably geometry-related. When I work with the same file, switch to another layer containing some simple meshes and then turn on the Rendered View, it's OK. When I change layers back, it usually works (reluctantly, though - it's a massive architectural file, roughly 3×3 km, with probably a few hundreds of buildings), but does crash sometimes. I'll post a screenshot of error and the console.
matisso: the error you mentioned happens only with a specific .blend file?If yes, can you please share it?As for the runtime error, I haven't had any so far, considering I always use my own build
Best regards
Sorry for my crappy english.
No one else had a C++ Runtime issues I was mentioning?
It should be working now.If you open a .blend file (saved with this build) using an older build (i.e. official download from blender.org), the objects should keep their size
Best regards
Also, for particularly geometry intensive scenes, it crashes and returns an error that Nvidia OpenGL driver timed out. Not sure if it's driver, or implementation related. I'm on GTX 460 SE, recently updated drivers.
yeh so basically I was having this problem of when solving a track, the solve would give an error value of 500-2000, and it was unuseable. Submitted a bug report and nazgul replied saying it was a problem with c-make and that it would be fixed in revision 42309, as stated here:
"Fix #29469: Latest motion tracking not working.
It was error in CMakeLists.txt caused by automatic bundling script which
expanded variables instead of substituting them as-is.
Fixed both of bundling script and CMakeLists.txt"
He also refered me to the official build bot build which i downloaded, which works. So making sure that it was past revision 42309, i downloaded this build, and it seems the fix has not been effective . If you can't find the solution, ill submit another ticket for nazgul and notify him that it was not effective.
I did not know that, I stand corrected. Thank you a lot for your explanations, Alb.
Take a look at nvidia support page:
http://nvidia.custhelp.com/app/answers/detail/a_id/3007/session/L3RpbWUvMTMyMjgzNzc0NC9zaWQvQkozWkRBS2s%3D
ArekBulski and evan3585: The objects you model using older builds are meant to be forward-compatible, but the opposite is not true.
If you model something using recent builds, you probably will have problems related to dimension and distance of objects if you open the blend files using 2.60.0, for example....
This isn't a specific issue of my build: it's related to the development cycle.
Best regards
tek5828: I would tell you about that fix but couldn't post it earlier....but what is the official build you mentioned?I'll take a look at the blend file you share....
ArekBulski: I suppose you're getting "Display Driver Nvidia Windows Kernel Mode Driver Version 285.62 stopped responding and has successfully recovered."
It might be related to the nvidia driver/card, as I've already heard a lot of people with this problem, also in nvidia forums.....I even found a Blender user speaking about this problem....
But, I remember when the first Cycles build was released, I had that problem when connecting/disconnecting nodes in a given setup....I'll take a look at your .blend file to see if I can reproduce it here...
boby: by "all of my CPU works together computing the subdivision surface of my cubes (not only 1 cpu 100%)" you mean the problem doesn't exist anymore?
Best regards
16 hours 5 min ago
Fix #29469: Latest motion tracking not working.
It was error in CMakeLists.txt caused by automatic bundling script which
expanded variables instead of substituting them as-is.
Fixed both of bundling script and CMakeLists.txt"
I submitted it to the bug tracker and nazgul replied as above. Directed me to the official build which works, although even though this is build 42330, I still have the same problem with this build. Just to let you know.
Here I uploaded the problematic .blend and screenshots, anyone feel free to mess with it.
http://speedy.sh/xbN9p/fails-message.png
http://speedy.sh/zGkAZ/malpka-a-odbicie.blend
Oh, alb3530, thank you for all your builds. :)
http://www.youtube.com/watch?v=Ma78frwQtYE
all of my CPU works together computing the subdivision surface of my cubes (not only 1 cpu 100%)
I ll try to explain with my bad english:
- add a subsurf modifier on the default cube
- Open your CPU inspector
- Press Alt + A in blender to launch animation
- duplicate your cube
- Your fps will go down faster (depend of your CPU), you can set the fps to 60 to see the affect faster
- when your fps begin to go down, watch your CPU inspector
- You are "out" of cpu, but only 1 cpu is working at 100%, others are about 5/10%
- I tried with other build (OpenMP SSE2 build for windows 32b, here on graphicall), and my fps is really better with the same scene
- all of my CPU works together computing the subdivision surface of my cubes (1 cpu 100%).
- Gain is something like 300% faster in viewport for me
Thanks again
Doesn't work here. Not SSE2 optimized.
I tried on a simple scene with subdivides cube(subsurf), and only one core works (13% of my CPU)
Optimized SSE2 build use about 60% on the same scene.
Cpu : Sandy bridge 2600K
Could you do that?
Thanks
Thanx for the compilation. How you succedeed ? i tried MSVC 2010 (Express&Ultimatum), 2008 and still get boots library linking error-libthread etc.
so compilation fails...
http://www.mediafire.com/file/27phrg0mrd3aaud/tracking_bad_packed.blend
do you want the video footage i am tracking packed as well.??
Here is what is happening in detail.
Load up this build,
open video file.
leave it at 250 frames for the scene, just for testing purposes, do a "detect features' and then track all markers for 250 frames. got around 20-50 tracked markers. Set camera data (nikon d7000, 24mm, set for tracking as well as scene camera) . add camera solver function to scene camera. select A and B frames. solve (some times crashes, not new) --- result is massive error with scene camera barely moving.
** just tried with default camera settings, this time the camera actually moved, although still had severe problems (jumping).
^ tracked on KLT.
Tracked this same footage in previous Tomato branch, no problem.
- tek
yes actually, I tried this twice when I saw the comments bellow mentionning that issue... I did exactly what you say C:/BLENDER/blender.exe
still have an error... but not the same one !! lol
here is the console error :
openCL build failed : calcCompile failedError : Creating kernel kernel_ocl_path_trace failed !
tek5828: can you please share the .blend file you're having problem with?If not asking too much, can you explain your problem with m5ore details?
cobalt31: make sure that full path to blender.exe doesn't contain any spaces on it.It should be something like C:\BLENDER\blender.exe and not like C:\BLENDER 2.60\blender.exe
thanks for the suggestion though... I'll just wait and see
To all who helped - Thank you very much for all the responses, I'm assuming this error thing is about Cycles being a work-in-progress feature so I'll just watch out for any future fixes for it (or better yet, switch to an nVidia card soon...need to save up still though XD).
Thanks for your awesome efforts,
rhui-
10.5 (5/26/2010)
11.1 (1/26/2011)
11.5 (5/9/2011)
11.7 (7/27/2011)
11.9 (9/28/2011)
11.10 (Latest - 10/31/2011)
all of them gave me the same error and console message..
*I used Phyxion's driver cleaner to clean the drivers before installing another driver.
CPU render works..I'll just play with that for the meantime..
More power,
rhui-
rhui- have you tried updating or rolling back your drivers?
If any one knows with that is ! please let us know ;O)
"Compiling OpenCL kernel ...
OpenCL Build failed:
Invalid Option: Study/Blender/cycles_x64.Win7-41652/cycles_x64.Win7-41652/2.60/scripts/addons/cycles/kernel (expected - at the beginning)"
btw I'm using a Radeon HD 5770 crossfire x... Looking forward to a solution from you guys.
Keep it awesome,
rhui-
Best regards
Is there anyone here who have the same problem? Or better yet is there anyone here who solved the problem already?
Thx in advance! :)
You might want to go to http://www.geforce.com/Drivers and update the drivers....
Also, what application are you using to check CUDA Cores?Is it GPU-Z?
brick: Try to download now and let me know...If even so you can't, I strongly sugget you to contact your ISP, as you've stated you can't download the file even using a proxy server..
Best regards
Please, can someone help me?
regards
Thx in advance...
Best regards
Lets wait it's fixed soon, cause it can be of no use if you're going to work with nodes in your current project....
Best regards
i am not sure what this is, if i mean this is a build problem or something.
thanks again
Best regards
Best regards
cobalt31: Unfortunately, I don't have an ATI card to test the build on it.But I'll see what I can get anyway....
Speroangelus: What software are you using to check CUDA cores?Is it GPU-Z?
And what is the card and driver you are using?
Best regards
I have notebook with a Cuda graphicscard, that's supposed to have 48 GPU-cores and an i7-CPU quadcore (emulated 8 cores).
When I use autodetection for Cuda I still only get 8 cores, could it be that only my CPU's cores are detected instead of the GPU's.
I've had this with every build of Cycles I've tried so far.
Has anyone an idea that might help me?
Thanks alot for these great builds though love it.
Thx in advance.
Invalid Option : Files/Blender (expected - at the beginning )
Thank you, Alb.
And it's got OCEAN SIM!!! :D
MartianStories: I'm adding added experimental CUDA support to this build right now...please test and let me know..I've tested with a simple render and it seems like it's working.... ;)
Best regards
However, your Cycles Fastest is NOT THE Fastest. That's because there is no GPU CUDA rendering here.
You make a Fastest, and make it so that NVidia Cuda cards can run on GPU, you'll have a smash hit on your hands. And if you apply all the little tricks you've done to BI to get your Fastests, you'll have an army of Cycles addicts like me sending you love poetry.
:)
but openCL doesn't work for me,
error like this in the console :........ http://www.kaliposer.de/sonstiges/error.jpg..............
have you any idea ?
thank you.....
Best regards
Best regards
It seems like some component is missing....Have you installed Microsoft Visual C++ 2008 Redistributable Package (x64)?
I always perform some basic tests before uploading here.More precisely, what I do is:
Open the application;
Check contrib scripts, to see if they are there;
Perform a test render;
Best regards
Also how come no CUDA?
I am not very good using Cycles yet and I just made little mistake.
I didn´t know that you can´t create or upload new image in Edit Mode if you are not added Texture Coordinate node to your object.
I can create new image in Object Mode, but I can´t do it in Edit Mode to specific object without adding Texture Coordinate node in Node Editor.
In blender Internal Render this was posibble.
Hopefully you get my idea =D
DanielCookie: I'll try to reproduce this problem here
Thanks for help!
The error is:
OpenCl error (-1): Device not found.
OpenCl error(-34): Invalid context
OpenCl error(-34): Invalid context
OpenCl error(-36): Invalid command queue
OpenCl error(-34): Invalid context
OpenCl: failed to initialize device.
I searched for nvidia gt 540M and there was OpenCl mentioned.
My System : Win 7 64, intel core i5-2430M, nvidia GeForce GT540M, 8GB DDR3.
Any ideas?
From commits:
Revision 41689 by blendix
...
* Windows: CUDA kernel compile at runtime is failing, probably will have to do precompiled kernel again.
...
But you can be sure it will be here as soon as it's implemented...
Best regards
There're some visible features though, such as node frames.
It should work on Windows 7 x64, as this is my very machine and I use Blender on it.....
Intel Core2Duo E8400
Win 7 Pro 64bit
Can't launch this app because of wrong parallel architecture
Thank you for the feedback
Still defaulting to exr on save image...
make sure you have latest update by (campbellbarton)
http://lists.blender.org/pipermail/bf-blender-cvs/2011-August/039016.html
lux render, yafaray and mitsuba exporter.
along with all the contrib and external addons.
this one..
http://graphicall.org/76
It includes the External scripts & some more stuff.
Basically it's your build with the extra scripts & file structure added.
If you can contact me on irc freenode #blenderpython
I can help with issues that sometimes arise with the scripts/modules or any other info you may require. :)
But as far I can remember, that one was some 2.58 build
Thanks for the suggestions
I'm thinking about including external addons soon as well...
this is much nicer than all the scripts in the one folder.
and adding lux render, yafaray and mitsuba exporter.
Maybe have 2 builds "fastest lite" without all the addons.. and
maybe something like "fastest complete with external addons/lux, yafaray, mitsuba exporter" or something :)
(it would basically be like the nice bat3a's builds, just with a faster blender internal render,
would the build optimizations make external renders faster as well? hmmm dont know)
if you help you'll sve me the ordeal of converting to ubuntu
Blender has to be OpenCL entirely - composer, rendering (Cycles), animation playback and so on. Graphics card MUST accelerate calculations.
Log in to leave a comment.