SoftXSI 1.5 launch: mini report

Last week, I went to the Santa Monica, CA, USA launch of SoftXSI 1.5. Hoping to see some friends, get some free food and drink, and maybe even learn something about the software. I did bump into Jack Walsh, Evan Ricks, and Scott Kirvan (old friends). Pretty much spent the entire evening talking about rendering and production with Scott. He’s one of the partners in the “renderer formerly known as Ghost” initiative. I had a great time. The food was really disappointing but they had a free bar. Darn, I wish hadn’t had to drive.

It was definitely not full, compared to the Maya launches I’d attended at the same venue. It was perhaps 2/3 full, so I got the impression there was some disinterest in the feature film community since I didn’t see many folks from big studios. This is surprising since Santa Monica is in the heart of the new “Hollywood by the Sea” where most big and small studios have moved. I’m sure the SoftXSI/IRIX debacle has lost them significant business from the “big boys”. The people I did meet that I knew where mainly in the post-production/boutique visual effects business where NT has gained some acceptance.

So onto the demo. It was pretty good in some areas, perhaps leap-frogging competitors, and very much only “catching up” in others. For example, to compete with MAX and Maya or just implementing stuff that was in Soft3d for ages and finally moving it over to XSI. However, the same thing happened with Max and Maya, it’s just that XSI got a really late start.

I was especially impressed with the polygon modeling and sub-div rendering which are somewhat more advanced than max4. More in the realm of Mirai in terms of workflow tools, perhaps. The sub-div rendering isn’t at the PRMan level but it was superior to max4’s because it incorporates view-dependent tesselation, from what I heard. That means only the necessary tesselation occurs for a particular framing or relative screen size. max4’s subdivs use explicit tesselation regardless of whether the object is even in the camera view!!!

Even though I’m not convinced MR2 is a good general purpose renderer because of speed, the integration with XSI and their IPR (Interactive Photorealistic Renderer) seemed way beyond Maya3 or max4’s versions.

The node tree shader editor looks really powerful. The proof that it is usable in actual production seems to be evident since you can get free MR2 shaders from this website. It’s still small potatoes compared to the PRMan shaders available, but it’s significant compared to the Max/MR2 shaders. Sorry, I don’t think the XSI ones will work in max’s mental ray.

The animation mixer, termed NLA (Non-Linear Animation) was another of the strong areas. No surprise, since animation has always been Soft’s forte. There was some interesting analytical blending between animation clips. Motion mapping of mocap data between different characters looked good.

In the “catch up” category, they were very proud of having parametric primitives, which pretty much everybody has had for years starting with Prisms, then Houdini, then Max, then Maya. You could adjust the number of segments on a sphere, for example. Whoopee! (not). There were a few more controls than max4’s prims but nothing to lose bladder control over.

They also have some parametric modifiers, like “Twist”. Hmm, groundbreaking, eh?

The NURBS were also decent but unremarkable. He showed off the ability to do curves-on-surface based on two intersecting surfaces. I believe that’s in most packages and was even in Soft3d (not certain about that).

The paint tools were fairly sophisticated from the little I saw and the UV mapping tools more user-friendly than max4’s but not necessarily up with Maya, at least with the paint paradigm. A good analysis would take much more time.

The particle system had some interesting features. They called it a “full particle system”. I’m not sure what that means. Maybe Houdini would be full + 5, in that case. I really couldn’t assess the features based on the demo. Have to wait for my demo-CD.

What’s really missing is any connection to PRMan. Animal Logic support Soft3d with “SoftMan” but they haven’t seemed interested in an XSI version. In the biggest insult to my intelligence this year, the demo guy responded to my PRMan pipeline question by saying “we don’t think you need PRMan anymore, because it’s old”. And some other equally stupid remarks. Then he said “but I don’t know anything about rendering, I’m a game guy”. I could only shake my head. No wonder very few film people bothered attending.

A Linux version of XSI was also mentioned but not demoed. AMD was the sponsor but the demo workstation was a “dual 1000” (Intel), ironically.

So it was a really mixed bag, IMHO. They had some impressive technology in areas they were already strong and some other areas were merely copies of competitors. Their main problems, I believe, is they have alienated much of their core market by releasing an unusable IRIX version and losing their PRMan pipeline. It remains to be seen if they can salvage their position and reputation with a robust Linux version. I’d love to see it happen because I don’t want Maya and max to become complacent. Competition is good.

I’d love to hear others impressions of the event, either in Santa Monica or at the other locations.

Martin Foster
“Johnny on the Spot”

So you went to the demo to check out the competition to your own products Maya and Max. What is your point?

I realize you said,

Urp,

Wow! No wonder working professionals aren’t keen on participating in this forum. In the past, people have thanked me for writing these reports, even if they didn’t agree. Are people that insecure here?

Anyway, I think you mistake me for somebody who has some kind of allegiance or loyalty to a package. Those that know me will understand that it’s not true. I’m happy to compliment or criticize any package, as justified. I think my credentials and experience in the industry are adequate for this kind of analysis, since I’ve used max, Soft3d, and Maya in production on “high-end” work like commercials, feature films, and large format films. I’ve been writing reports about Siggraph and other shows for years, so it’s nothing new to share my impressions. It’s just new on this forum. (One thing that bothers me about this forum is the use of “handles”. What’s wrong with proper names?)

So the moment I stop looking at other software/technology that I might potentially use, that’s the moment I should probably retire or go into another line of business. Do you ever go to Siggraph? Do you only look at the packages that you work with? If so, I would consider that a very narrow view.

In a nutshell, I think my conclusion was SoftXSI 1.5 is maturing into a great package with the biggest problem being poor IRIX performance. This pretty much excludes it from use in the big studios who are still using pretty old IRIX workstations, like Indigo2/R4400’s or O2 R5k’s etc. So, for SoftXSI, I say “you can’t be in the game if you in the wrong stadium”.

Perhaps you can tell what Softimage is doing to improve their IRIX and/or Linux performance? NT is not going to be accepted in the big shops and the small number of prestigious feature-film seats are essential to drive sales for the rest of the world. What sells well in Hollywood, sells well in Nebraska … or Katmandu.

Other than questioning my motivation, you made some good points where I think we completely agree, but you didn’t address the IRIX issues.

Martin

Thanks for the report Martin, it is appreciated.. I didn’t get any sense that you were just giving your impressions. Always good to get feedback from all angles..

M.

Hey

Regarding the ‘we’re a tough crowd …’ you bet. We play de hardball here )

No seriously, I think Urps post was not as harsh as it might at first appear. He is, after all, agreeing with you. He merely questions the fairness of the report, in his opinion. Just because some people take everything with a healthy measure of cynicsm doesn’t make them insecure. But if you don’t intend to have people add their opinion, what is the point of forums?

Don’t mean to step on your toes here, btw. I liked reading about the event, and I am glad you took the time to write it. I guess it did come across a little biased, to me at any rate, but whether that was intended or no, only you know, if anyone.

The handles are quite common in internet forums. You are free to use your name if you want, but have the freedom to remain anonymous. What is wrong with that?

BTW, from what I hear, there will be various Irix performance optimisations in 2.0. Also Linux versions of 3D were just demoed, with XSI 1.5 to follow soon. Where are the MAX versions on IRIX or linux? Is Maya really that blisteringly fast on Irix? not from what I have seen…

Balgillow

Well balgillow, I agree that if you’re going to post an opinion or report, its fair game to be disputed. I don’t think that was the gripe with urp’s post. It was the tone of urp’s post that was in question. It did come off somewhat combatant, despite how valid his point’s may have been. While the original poster’s post may have been bias (I don’t know, up for interpretation), it didn’t at all have anything but an open and cheerful tone. At least that was my initial impression, and also that of other folks I know who read it..

BalGillow,

Well, I’ll agree that it’s an opinion piece framed with reference to my own experiences. How could it not be? I can’t evaluate it from your point of view or Urp’s. I just was rather shocked by somebody questioning my motives. It’s never happened before when I’ve written Siggraph reports and such for the 3d community.

If somebody wrote up THEIR impressions of the show, or any other show I would enjoy that. I might discuss some of the conclusions – that would be quite valid – but I can’t imagine challenging their motives for doing it. That’s why I was asking others opinions of the show.

As far as biases, I think we all have some, but I feel that I’m as open as I can be about trying new software. I’ve used most major packages over the last 10 years (3ds, Soft3d, MAX, Maya, Mental Ray, PRMan, Shake) but my new job at Sony Pictures Imageworks as a lighting/compositing TD actually involves largely in-house tools except for PRMan for rendering.

Fair enough? Perhaps we can turn this into a constructive thread yet.

Did ANYBODY else go to the event??? If so, what did they think?

Martin

Well, now that we’re all friends again… cackle

Here’s a deal, I’ll be at the German Launch Event for 1.5 and I’ll give you my impressions of that one when its over. Launch event is on the 25th, in the CeBit. Anyone else going to be there?

I’m also curious to see if the audience they target will be different. Bit of a different crowd in Germany, less games, more postpros. Sounds like your launch event was tailored to suit the games market. Also the German launch will be a smaller affair, might make a difference

Balgillow

Hi all - just wanting to forse feed my two cents.

Thanks for the launch info -

A couple of pointers - Not everyone likes prman, - not many people can afford prman, if all the companies running maya consisted of all the people who ran it with prman - AW would be horribly broke… Mental Ray - people keep talking about how slow it is? - if you want to do a fair comparison - run Mental ray in scanline from a command prompt, then run prman from a command prompt - I wont open up a can of worms by saying wich is faster - turn on shadow maps, and open it up for me. Mental ray is easier to use. Mental ray can raytrace - Pixar subds are a pain in the ass - if anyone would like me to elaborate - give me a sign… Prman is pixel based -thats a big plus - Prman is the king of motion blur.

Both renderers have a shader creation option through shadetree or wordpad if you never see the sun, - the volume of shaders created by companies around the world doesn’t benefit anyone but them. You cant actually download prman shaders from ILM in case anyone has failed to notice. Shaders for any renderer are always tricky - The scalability being spoken of in the way of creating shaders (production quality) is beyond ninety percent of the industry - In the way of default blinn and so on - well, it’s a matter of prefference.

Secondly - XSI and even good old softimage implement to date the most advanced forms of nurbs stitching available - in my opinion it is the only thing that makes nurbs production ready on complex models requiring continuity across up to hundreds of patches.

It does not have trims - they should be shot for this - and the text famine.

I must admit i love pecimism, it’s a safeguard to bullshitt, - but only if it can be justified. What i also love is how the general concensus on this forum is how subds are the way to go - and yet maya is the king of modelling? Bluntly - ninety percent of my modelling nurbs or polys consists of pulling points - which works fine in xsi or soft - proportional modelling is also a lovely thing for face shapes etc - funny also how all the maya models we work on here often end up in soft for stitching or sculpting -

thanks - and yes launches are lame, - always, we have all heard about the product, a few of us own it, and a few of us will believe anything… Interesting then to see who is there…

Adios

as long as softimage dont go under, I’ll be happy, and happier still by how difficult peoples lives are.

  • Non smokers die every day.

" Pixar subds are a pain in the ass - if anyone would like me to elaborate - give me a sign…"

Please “elaborate”

“Pixar subds are a pain in the ass”

Please elaborate - because I don’t see your argument. PRman is a renderer which uses the Renderman Interface Bytestream (dotRIB) to store it’s primitives, animations and shader calls etc. Exactly the same as dotMI for Mental Ray. Ease of use is dependant on the designer of the interface that exports to the RIB file that is then rendered in PRman. If you saw an application that handled Sub-Ds poorly, then that isn’t PRmans fault.

"Both renderers have a shader creation option through shadetree or wordpad if you never see the sun, - the volume of shaders created by companies around the world doesn’t benefit anyone but them. …- The scalability being spoken of in the way of creating shaders (production quality) is beyond ninety percent of the industry - In the way of default blinn and so on - well, it’s a matter of prefference. "

PRman and Mental Ray are for different purposes. There’s no competition between shading control - PRman wins hands down. Softimage’s Render Tree connectivity with Mental Ray makes it second best.

“…You cant actually download prman shaders from ILM in case anyone has failed to notice. Shaders for any renderer are always tricky…”

D’uh. You can’t download their Maya scenes either. Or you can’t download their textures they used. Surprise, Surprise - a corporation who makes millions of dollars not giving away their prize posessions - what were they thinking?

Just remember, PRman and Mental Ray are aimed at different markets. An animation team composing of 3 animators and no TDs has no need for PRman simply because they don’t need that high level of control. Mainly because that high level of control comes at a price (of PRman licenses & some kind of Technical Director for shader design - among other expenses).

-Vijay

elaboration 1

  • I and most other people here dont work at pixar, therfore we do not have acess to a reasonable texturing option (in my experience) - UV’s can work on them but do show splits and or holes when a group of uvs meets another group of uvs on a continual surface = pain in the ass - in my experience again mtor does not even give you the ption of using uvs at all making it an even bigger pain in the ass. - The application in question is prman - it and it alone subdivides the polys - making your question a little unfair. To compare how a file format stores geometry data is in no way reflective of how it actually renders - especially in prmans case as it renders shall we say subds based on pixel based resolution rather than polygon tasselation - which i do believe could be the root of a lot of problems, but lovely in other respects.

Prman and mentalray are for different purposes - exactly why i said not everyone likes prman - I for one use it often, but not for everything. Tree conectivity in shade tree is easier and faster than any method i have seen used for prman - what on earth thi has to do with shader flexibility is beyond me - MR is also a ray tracer, which gives a few more options in terms of flashy effects (opinion)

No shitt about not being able to download shaders buddy - thus my sarcasm, - you didn’t seem to notice. The point of this and a lot of my rambling is the justification of peoples reasoning revolving around what people like ILM are doing with prman - My opinion is to worry what you and i can do with prman - as their isn’t a conveniently located individual nearby who helped create the program / programmer who cares about 3d in my case etc .

High level of control? TDs? - buf = cell + fight club … believe me prman’s prefference has less to do with high level of controll and more to do with supposed speed and definite speed in terms of motion blur and so on.

Thank you four your patronizing ILM comment - hope you realise it was ultimately a reflection of your intelligence.

Happy rendering

  • Non smokers die every day.

Martin,
Sorry for the late response. I

"To compare how a file format stores geometry data is in no way reflective of how it actually renders.. "

True, but you said Sub-Ds were a “pain in the arse” which I took to mean difficult to work with.

“especially in prmans case as it renders shall we say subds based on pixel based resolution rather than polygon tasselation..”

I believe the word you’re looking for here is micro-polygon which is equivalent of tessellating down towards what Sub-D people refer to as the “limit surface”. MR tessellates everything and everyone knows PRman handles Sub-Ds effectively and rather powerfully over other renderers. Smoothness is implicit for PRman whereas in MR you have to adjust the tessellation so it appears smooth.

“MR is also a ray tracer, which gives a few more options in terms of flashy effects (opinion)”

True, all raytracers have more in terms of effects. But why use a raytracer when PRman can do the same thing with some ingenius shader trickery in half the time of MR. If worse comes to worse and you need a Raytracer you can connect to BMRT - not a big drama.

"believe me prman’s prefference has less to do with high level of controll and more to do with supposed speed and definite speed in terms of motion blur and so on. "

Speed differences are phenominal, even without severe optimisation. I heard one claim to achieve 10 times the speed increase over MR. Also, PRman handles millions millions of polygons effortlessly. You should read some articles about the movie Dinosaur. In one story a TD said after they finished modelling & texture mapping one dinosaur (including displacement mapping) he attempted to tessellate the dinosaur model (including the displacement maps) down to polygons. The workstation with 2 gigs of ram choked within 1 hour and it was no where near complete. Sure there are other renderers with this power but PRman is quite mature and dependable which is why it’ll be used for big time projects all the time.

-Vijay

I went to the Sydney XSI 1.5 launch -
the demo guy is a very funny (actor turned animator/demo artist) and knowledgable guy called Mark Chinny.(UK)

He gave a very thorough and entertaining demo of 1.5- one
of the best demo’s ive seen. If anybody gets a chance to
see this guy demoing XSI dont miss out. In particular, he gave a thorough
and impressive demo of the animation mixer.

Talking to Mark afterwards he admitted there are still some
things missing from XSI but he had nothing but admiration
for the product - you knew he really believes in it.

Rotifer

I signed up for that launch however I didn’t get emailed about the details

Where was it?

-Vijay

thanks, Urp. I appreciate the clarification.

Martin

Hi County,

it was at the australian animation and effects festival, Darling Harbour, Sydney-
I think it was heading out to Melbourne later on.

I think you need to be registered with HM3d to get the invite.
Incidentally, the XSI 1.0 launch was awesome- nite club,
go-go dancers, free beer.

Rotifer