Trace: » survey1_results
Table of Contents
Survey Results
Freeform Comments
| Please provide comments about your comparative experience, and explain how Aqsis could improve when compared with alternative solutions. |
|---|
| Lack of raytracing is of course a slight detriment. The eyesplits problem with shadowmapping is a little strange to deal with too. |
| Aqsis was definitely slower than prman. Also it doesn't support the current prman feature set. Aqsis also lacks the toolset that a modern prman environment comes with (alfred, etc). |
| I like aqsis it seems to provide more support for renderman specs. It has the best importers,exporters on par with 3delight. aqsis more stable than pixie and only recently 3delight has as good support for CSG than aqsis. However aqsis is slower when use with DOF, Motion blur. and take sometimes too much memory in particular with DelayArchive(). |
| The main issue with Aqsis is the connection to my main tool Maya. There is Liquid, but getting it to run is a pain. And it is a pain with every version change, that is why I gave up for now. I can only endure so much pain in a row. For Aqsis to make a difference to me, it has to integrate painless. It is ok if it has some things missing, but if I am not even able to press the render button and out comes a render, then we are at a stage of pain again. |
| Aqsis should have native ray-tracing support, and maybe support for photon maps rendering/shaders. |
| A Graphical User Interface for creating and administrating shaders with autogenerated preview samples would be nice. Something like the old Pixar RenderApp and Glimpse applications. |
| no raytracing no global illumination |
| It depends on what area you look at to decide whether Aqsis performs better/comparable/worse than other renderers. Every renderer I've used so far has had its weaknesses. In general, Aqsis has been slower (sometimes even dramatically) than other renderers, so with this respect it usually performs worse. |
| Renderering speed needs improvement. Equal Macintosh support needed. |
| Could be faster, should be multithreaded. |
other than the lack of raytracing, and somewhat inferior performance/speed, it's comparable, and quite good in fact. There were some problems with extreme displacements, revealing grid cracks as well, but i'm sure once these problems are fixed, and performance gets improved things will be much better, but they're quite good already. One possible direction would be the creation of a aqsis DSO to interface with a streaming language for GPUs, like 'Brook for GPU', to offload highly parallel/processor intensive tasks to the GPU (such as matrix/vector operations, etc…). Use of this in the RSL sources would either transparent, with no user interaction, or via a subset to specify the streaming language use on GPU. Nvidia cards are very well supported on UNIXes as well, and since a NV45 GPU has around 54GFLOPS peak performance, and a P4@3GHz peaks at 6GFLOPS, this would mean a tremendous performance boost, for some very specific situations, and aqsis could benefit from this a lot, since rendering is one of the tasks that can be easily parallelizable, as well as the already mentioned matrix|vector operations. This seems the trend in the industry atm, to offload computation towards the GPU, using it as a dedicated stream/vector coprocessor. Development of parallelism on the CPU could also be interesting, either classic SMP like, in which parallel buckets are ran simultaneously, or SMT like, in which multiple threads contribute to a single bucket. As for raytracing, www.gpgpu.org has quite a extensive series of articles about raytracing/radiosity on the GPU, this might be another interesting area to follow, and overall, all the “hacked algorithms” available for realtime use (ex: games) work quite well, and produce a huge boost in performance (look at prman 13's ambient occlusion and new 'radiosity' when compared with earlier versions, as an example). That's pretty much it, i would however urge the aqsis team not to follow the quick raytracing way that pixie followed, pixie is notoriously problematic in this area. As for the mentioned streaming language, Brook for GPU: http://graphics.stanford.edu/projects/brookgpu/ it's a C extension basically, with stream/kernel data types, etc… (you can read more info on the url), and i'm sure there would be a huge boost in performance in having a DSO with Brook versions of the ri specs. The ri specs would be unchanged, the classical non-Brook/GPU functions, and as an alternative to nvidia users, the option to use Brook/GPU DSO versions of the ri specs. Note that i refered nvidia more than once, only because it's well supported across several platforms, while ati unfornately still has a long way to go in regard to driver stability, and it's not as well covered in GPGPU projects atm. That's it, keep up the good work. |
| I like BMRT because its lite, its easy, and the RM compliance is there, its fast however it is becoming old . Povray is good , i am too lazy to learn it. Not sure how Aqsis can improve as i am only just looking. |
| I like how Aqsis handle any RIB files. I found this program to be slower than many of the rest. |
| Launch a campaign to get TDs working on actual productions to throw RIBs at Aqsis in their spare time. Motivate them to report the results back, even if they can't give out the RIBs. Nojthing fast forwards the stability and production-readyness of a renderer like testing with real-world RIBs |
| (perhaps I shouldn't speak here since I'm not a really regular user, but…) a good gui for import of .obj/.3ds files, with ability to change parameters of materials, camera angles,lights, gi alternates etc (along lines of Kerkythea) |
| It needs to be faster |
| Slow. No raytracing. |
| More documentation, Raytracing option. |
| too slow for on some task |
| You are doing an awesome job with Aqsis. |
| Aqsis is slow. But Pixie doesn't have the CSG feature, and this is critical. |
| Rendering speed is the major bottleneck. |
| Performance could be better (speed). Some minor bugs here and there. |
| the real big problem is speed… |
| Negative: -no Raytracing -ridiculous Ambient Occlusion -(hence) no GI Positive: -stable -good color quality -trim Curves -relatively fast |
| It is slower than 3Delight. It lacks certain features such as raytracing, GI. |
| By providing ambient occlusion or/and global illumination using raytracing. |
| I tried to use Aqsis with cutter in a Linux environment. I hope Aqsis can provide more packages for different Linux distributions. |
| Aqsis was better than pixie (both play the same game), but lately Pixie has improved a lot, so that it has more modern features and a bigger community. |
| Please use the below textbox to share any further comments/toughts about the Aqsis project itself (256 characters max.) |
|---|
| Keep up the good work. What you guys have done so far is remarkable. |
| A great project with *real* potential… back-pats all-round !!! |
| The project itself is great. The people are nice and helpfull, which make a huge part of the experience. Unfortunately, the product integration is completely lacking. There exist some tools, but none of them work without pain. |
| No clear focus - where does Aqsis want to stand in 1 year? Not clear to me … Maybe it'll help to promote Aqsis for example as the baking tool of choice for game developers (just an example). |
| Aqsis is great. I learned a lot on this project and I love the community around Aqsis. |
| I'd like (if i had time) to make the shading engine more vectorized (SIMD). 1st step would be to modify the way grid parameters are stored (to make them homogenous vector). The long run would be to put aqsis on a cell processor, with PPE doing split/dice and SPEs doing shading. |
| Aqsis is an excellent renderer, and I'm very glad to see that it is an open-source project. I look forward to what the Aqsis team produces next! |
| I love it! |
| please work towards total integration in Blender. Blender now has Node-Shaders which makes it fully possible to integrate any Aqsis feature. |
| Aqsis has a relatively low awareness amongst graphics community but is well known by rendering enthusiasts. for it too become succesful it needs tighter integration with main authoring environments if to be as succesful as technology deserves to be. |
| More optimzation is required to match 3Delight and Pixie. |
| it needs to be faster. |
| Please don't give up. |
| Keept it real, kewlers! Great work so far. :) |
| I am POV-Ray user, but I would like to use it through rib style interface. |
| doing a quite wonderful job (and, more importantly, a constant improvement). |
| I've followed Aqsis (and used it) since it first went open source and its certainly come a long way. I think it's a hard sell though because in the end, it can only do what any other RM-compatible renderer can do. Improving speed and stability are the most important things that you can do to make it stand out. |
| I love renderman. And I love open source. I think that you have started a wonderfull project, better than other valid projects like pixie. Good work :) |
| The source isn't reachable with ease from the main site |
| Awesome work! Keep up the great job! |
| - The rendering looks good, better i mean as e.g. BMRT on my Solaris station. - Shadowing : not perfect at this time for me, but probably at this day i'm missing something in several parameters. - The speed is a real problem by the usage. - The homepage and the community features are greatly installed. |
| Hold on, even if development slows down sometimes. Wish I had time to give more contribution to the project. |
| It would be good to have even rough placeholders for all the essential features needed in an animation production. More specifically, please add at least a crappy motion blur feature to Aqsis. Get in bed with Blender (just because it is your best bet currently), and be loud about it. |
| I like Aqsis, its open source filosophy, its support organisation and it slow but continuous evolution |
| You have a great render but some of the worst sample imagery I have seen online. Fisrt impressions…. Two words for this project “Team-up”. K3D is a good software to look at becoming a “partner” with and there is of course the now dead Shaderman, project witch I believe would be a glorious tool to bundle with your render, ( You seriously need a node based preview tool that alone will cause an interest that is more than passing ) but if you had a modeling enviorement ( like K3D ) wow. BTW it also python scriptable. |
| Blender support is essential. |
| The one thing that I would like to see is an increased community, increased community participation and more documentation on how to get into the code. |
| The documentation isnt all that great, but over all its a nice package, I plan to so some serious production work with this in the next 6 monts |
| Very nice tool! Very helpfull! Is there away to cut down the rendering time for a single image with a cluster? |
| Please use the below textbox to share any further comments/thoughts about other ideas and requests (256 characters max.) |
|---|
| I'd love some profiling tools for shaders. |
| The documentation is in need of more loving. |
| Speed improvements when aqsis is used with DOF, MB is critical IMHO. Along with memory, resources, memoryleaks cleanup, … For the previous items well we need stats, with real examples and real dev tools to provides such stats on weekly basis. DSM is also overdue. |
| Get in contact with the Liquid Team and do a little bonding. When you have them as your friends - not just as neutral guys related to you - they might help out with getting Liquid to work on Maya. Which is my main concern. A renderer is no good, if I cannot use it for my main package. And Maya is a good target besides Blender, because it shares a large audience. When people see it working there, some developers might even join. |
| Excellent support, great community, no or small user base keeps new users away (devious circle …). |
| I'd love to see more integration onto other tools as well as a improved user friendlieness (which isn't bad at all but could be improved compared to non renderman renderers) |
| Better Integration with blender is also a strong need. But it seems to be some blender work. |
| I would like to see some speed improvements in the renderer. Support for parallel or multi-core processors would be appreciated. |
| You really really need to hire some artist, or at least you could try to get the permission to convert some hi-quality scene. |
| GPU accelerated versions of (some) RIspecs, particularly useful in raytracing, or (big amount) of shadow maps generation (for multiple shadow map techniques such as SSS/ambient occlusion). A GPU accelerated relighting tool. |
| Ease of use with Blender is the upmost of importance, otherwise this probably nice renderer will remain unknown,unused and unseen by most people. |
| Create plugins for better integration to convert files from Blender and Maya. |
| a shader gui. links with a modler Ayam or blender |
| POV-Ray engine is becoming under GPL I guess. Please somebody hack the RenderMan interface for it. |
| would love to see better integration with blender (as in yafray or blend2pov or povanim) |
| Only one: a more frequently updated roadmap and more info on the development :) |
| A couple questions ago, it said to rank how well Aqsis was doing in the following areas… I think Aqsis is doing awesome in all those areas… so 'lower' doesn't meen that it needs attention as far as I am concerned. |
| - Blender integration : but this is in progress, as everyone know (and normally this would be a Blender related question) ; - Eventually, a subforum for Ayam3D, as existing for K-3D and Blender ; - Scons : it should be useful to find in the aqsis package : . any configuration file or howto for each OS ; . otherwise, for one or two years, the “old” Makefile way, i mean not only Scons but the two ways (e.g. Blender has actually the two ways). |
| Approach integration with Blender with real-world workflow in mind. Give rough support for the whole of the process. Here, I'd recommend supporting the motion blur option in Blender. Try to get users to easily produce results with Aqsis. This would be good to the gallery. |
| My intrest in Aqsis will raise tenfold if and when Blender will have 'acceptable' integration with Aqsis (too bad the fork with aqsis integration died, although I of course understand the reasons….) |
| You need a shadetree/slim/darktree tool guys. I cant stress that enough. Again even using an Open source model ( teapot or Buddah or rabbit ) they are all great example with smooth edges. The gallery should reflect the features of the renderer. Does it do motion blur, HDRI, , raytracing, CsG, Toon, linestyles. Come on I Know it more than what I saw. I wouldnt even bother if it dint. You guy are right about the generic rib file issues but I think teaming up with K3d could really be a great thing. The k3d project is ruff around the edges also badly documented but combined with your tracer extremely powerful. Give it some thought. It never hurt Blender to collab with yafray. |
| It would be nice to see Aqsis releases a bit more frequently, but perhaps developer time constrains it. Nice to hear about the progress with automated regression testing and stuff, looking good. Keep it up! |
| ray tracing would be fun but keep up with the advanced things renderman is doing. It would be worth it to offer support at a fee to help fund development |
| I need very fast rendering of single images on a large machine: some 10 cpus. Is there a way to do this in a simple way with aqsis? |
Voting Results
The remaining questions fromt he survey are of the form of voting and ranking, the results of these have been gathered and are presented in the form of a PDF of graphs below. I can provide the Open Office spreadsheet used to produce these graphs if anyone is interested.