64-bit version of XSI is the biggest advancement from 4.0 to 5.0. All images © Softimage unless otherwise noted.

Many of SOFTIMAGE|XSI 5.0’s enhancements and features are a direct result of user requests, such as updated rigid body dynamics, 64-bit XSI, migration tools for Maya users, ZBrush compatibility, bézier curves, laptop layouts, dopesheet improvements, weight locking, shape manager. Others were a very welcome surprise such as, rendering speed and memory improvements, Ultimapper, GATOR, tweak tool and more.

64-Bit XSI
Possibly the best thing to happen to XSI users who work on gigantic scenes is the 64-bit version. It dramatically improves performance. No more “out of memory” errors or scenes that just don’t work because you’re stuck with a 32-bit piece of software. Tests show that the 64-bit version of XSI can handle scenes with a billion polygons and thousands of textures. Not that you would want to work on a scene that large, but if you had software that could, imagine how well a typical scene would run.

Rendering Memory and Speed Improvements
Part of the new Gigapolygon core, rendering in XSI 5.0 is about scalability, and mesh splitting is just one part of a major undertaking to ensure that large scenes will render faster and more reliably. Softimage has introduced new geometry compression as well as XSI render callbacks, which will only push a minimal amount of data to mental ray, and only when a ray hits geometry will it call back XSI for the relevant data.

In addition, there should no longer be issues with requiring large lumps of contiguous memory to achieve any of this, as XSI will break large objects up into pieces can be flushed when memory usage is low and call it back later. Complementing this is a new tiling option (again mainly useful for very dense meshes), which will work for the render region as well as when rendering to disk. Ultimately, this now means that when you can load it into XSI, it renders. All these new features are present in both the 32-bit and 64-bit versions. Plus XSI 5.0 has integrated mental ray 3.4.6 just about as closely as is possible. This is one of the reasons it’s able to render so many polygons at once.

As for .mi files, they are generally smaller (around 10 times) and now use subdivision geometry shading as well as the callbacks. This development work is almost all under the hood and does not require the user to fine tune, tweak and jump through hoops to simply render large scenes. Again, all these new features are present in both the 32-bit and 64-bit versions.

Rendering & Texturing
There is a new material port node for fast subsurface scattering and skin shader. It seems to render much faster than previous versions.

Pass Channels
You can automatically generate standard or custom rendered channels such as specular or reflections.

Render Options
Some of the new render options include: Multiple final gathering bounces, final gathering density control and the ability to filter color for global illumination, caustics and final gathering.

There are also loads of new animatable functions in the Render PPG. Jitter Sampling now has a slider, not just on/off. Photon GI and Caustics now have an RGB filter. FG Bounces now has a slider and so does Pre-Sampling Density. Automatic Compute now has a filter. Logged messages have check boxes rather than levels, allowing you to show only errors or only progress etc… Export MI2 now has subdivision geometry shaders export options.

Companies at IBC that announced new products or upgrades to products in the DI and data-centric space included Quantel, Autodesk, Digital Vision (which acquired Nucoda earlier this year), Assimilate, MTI, Imagica, Celco, Bluefish444, Zaxel and FilmlLight. Actual product announcements detailing the new software, hardware and systems can be found at each company’s website.