Jump to content

Fabric Engine Splice Performance


abvfx

Recommended Posts

Tonight this interesting video popped up in my inbox. I m not quite sure what to make of it. I am not a software developer or even a HDK coder, optimization technically is something that i know very little about.

http://vimeo.com/70423041

From what i know, black-box style specific plugins tend to perform well compared to features developed over time. But Splice looks a little different from that. Enabling TD's access to a language to work in.

I would like to hear the thoughts of those more knowledgeable on the subject, software/plugin developers or just someone who knows why things appear to be much slower in houdini.

Obviously it goes without saying let us not take this into the Splice vs Houdini territory but this is od|force so i doubt this thread would be hijacked by such petty arguments.

Thoughts?

Edited by phrenzy84
Link to comment
Share on other sites

what? this is the most shortsighted comparison I've seen in a long time

how can they just take any asset from orbolt and assume it is optimized for speed as their code when it's clearly not, if they'd understand how houdini works and how to get the most of it I bet the numbers would be at least comparable

even though I believe that Fabric Engine may still be a bit faster as it is clearly being built for the best performance possible (considering that their team has a few guys from former Softimage team and ICE core developer so plenty of experience on multithreaded procedural stuff)

  • Like 1
Link to comment
Share on other sites

He also complained how Houdini took very long to open the file and maya opened it instantaneously. Houdini doesn't cache its result, and I think there might be materials in the Houdini scene.

A cheap shot from fabric engine I think. They are mad because they are not in the same league as SESI :)

We need to create the same fence asset for performance and call them out.

EDIT: I made this in 15 minutes, it's not great but it does the same thing, and it's over 120 fps with the screen recorder running. I recorded high def but the website compressed it, that's why it looks blurry:

http://videobam.com/ZOzrA

Edited by magneto
  • Like 1
Link to comment
Share on other sites

Houdini simply looks slower as they lack the skill and tact to present a fair comparison. It's poor marketing. Why would you then trust the other services offered on their site.

Link to comment
Share on other sites

EDIT: I made this in 15 minutes, it's not great but it does the same thing, and it's over 120 fps with the screen recorder running. I recorded high def but the website compressed it, that's why it looks blurry:

http://videobam.com/ZOzrA

This is why i felt the need to post this thread. Watching the video and the performance i didnt feel it was a balanced comparison in speed. And now you can be sure i am going to download that fence asset from orbolt to see what exactly are its features.

Personally i still feel houdini's speed can hold it back, but is something to be expected in favor of openness? Maybe. More importantly can i live with it, yes it isnt awful, but who wouldnt like to see performance do better.

Either way i do think it is healthy to have such a discussion, if for the very least to educate people like myself.

  • Like 1
Link to comment
Share on other sites

I looked at the fence asset for 2 seconds and I knew what was up. It's not copying geometry like he shows in his demo, it meshes a geometry that's far more complicated. If you actually look inside, you can see the complex boolean operations, smooth operations, etc. Then of course it's gonna be slow :)

I bet they knew this but never expected to get called out. But now with Symek at the helm of the Engine Wars, they will be very sorry indeed :)

  • Like 1
Link to comment
Share on other sites

I wouldn't worry too much about it, SESI is continually improving Houdini's performance. While there are still aspects that are underachieving, they are being systematically addressed.

Perhaps the video was removed from the public domain as a retraction :)

  • Like 2
Link to comment
Share on other sites

Dragon, Thanks for raising concerns - I will get the assets reviewed and detail what's happening in each scene. We are travelling until later today, so I will lock the video until we can verify the data.

The video is still available for those who want to watch it and review the comments.

https://vimeo.com/groups/fabric/videos/70423041

Link to comment
Share on other sites

I've prepared simple fence scene if anybody want's to test since I was curious how would Houdini really perform, not that it matters that much, but it was really low blow from FabricEngine

https://www.dropbox.com/s/7xgn6z1o09skqej/fence_test.hip

64 poles

2000 side bars

53412 polys

changing ground geometry (which recomputes everything) ~ 60fps

the same with changing parameters on the asset

i7 2720QM 2.2Ghz, FirePro M8900

  • Like 1
Link to comment
Share on other sites

They removed the video because they got called out and their marketing effort just got holed below the waterline. They see the Engine as a threat to their nascent business model, which as far as I can tell, is another layer and yet another (and proprietary) programming language on top of what you have to do anyway to get Maya beyond the M-O-O-N-that-spells-production stage of usefulness.

The Houdini Engine is a force multiplier. The next couple of years are going to be very interesting.

Link to comment
Share on other sites

I partially see it as an ignorance from their part and not the intention, they were just so thrilled to see Houdini perform so badly on the asset they downloaded that they thought to be a good idea to compare

They have a top team who knows the stuff so they don't have to feel threatened by anybody and the competition to Houdini Engine is just healthy for both products

also blocking the video for now shows they may be considering they made premature decision about houdini

  • Like 2
Link to comment
Share on other sites

I'm sure their dev team is awesome.

But the two products aren't even in the same category and any comparison at this point would not be apples-to-apples, especially since the asset they downloaded is one of Jeff Wagner's really early ones, that does an insane amount of stuff under the hood.

  • Like 1
Link to comment
Share on other sites

on the other note, Fabric Engine seems really impressive in all of it's areas, I especially like the focus on speed an flexibility, I think they have good thinking of providing high performance CPU/GPU language to all areas of the production (like VEX, but seems more powerful and in H I really miss VEX for direct manipulating parms and obj transforms, even in simulation fashion) (also their python implementation seems to be pretty fast)

and even though I like nodes the ability of having code based nodes in all contexts is great for TDs, Houdini is close to that, but not quite there, the rigging context is really missing (CHOPs don't count for that as they are not flexible enough)

I think SESI can only benefit from keeping a close eye on FabricEngine

  • Like 1
Link to comment
Share on other sites

Hi guys - just got pinged about this thread. We took the video down as soon as we realized a mistake had been made - there was no mendacious behaviour on our part, just a screw up. I've contacted (through Vimeo) each person that commented on the video (I see a few of them in this thread) to explain what happened, and to explain that we'll redo the comparison after Siggraph (with a Houdini TD). I'm not sure it's that constructive anyway - generally we do performance comparisons to show Application vs Application + FE, as the goal is to show how we can help rather than just a performance delta. I may wait until we implement Splice for Houdini (dependent on demand) as I think that would be a better demo that showed something useful. Interested to hear thoughts on that, I know you guys have VEX already.

Anyway - my apologies for the video going out without proper research done. We were/are flat out on Siggraph preparation and cut a corner by not checking things with a Houdini specialist. I'm glad it was called out so soon as it could have turned into a really bad situation for us. It still wasn't great (I was boarding a flight to LA when I first saw the comments, and damage control on an iPhone with fat fingers isn't easy!), but it could have been a lot worse.

If you want to talk to me directly about it, you can email me at paul.doyle at fabricengine dot com (or PM me)

Thanks,

Paul

Link to comment
Share on other sites

I may wait until we implement Splice for Houdini (dependent on demand)

It's interesting what you can do for rigging, so I can see demand there (in Houdini).

I could also be interesting to have it in Houdini for a TD to be able to use tools already developed for Splice (in a mixed pipeline that would be useful).

That would put you in direct competition with Houdini Engine (and that's a good thing).

Dragos

Link to comment
Share on other sites

Definitely. Riggin is an area where Houdini is needed of much development and F.E might find a lot of interest there from studios with Houdini in their pipelines.

You screwed, but the correction was as elegant as the screw up. :)

Edited by Netvudu
Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

×
×
  • Create New...