http://www.pugetsystems.com/labs/articles/AutoDesk-Maya-2014-Professional-GPU-Acceleration-509/
I guess that pretty much answers all performance questions. At several times the performance level of a Titan, how much more future proof performance does one need right?
The W5000 is a pretty solid Maya-card, and you can always add a second one in the future in XFire or GenLock, but AMD has already implemented the connectorless GenLock/XFire system that just works over the PCIe-bus, that feature has even already made it to the consumer cards, i.e. the R9 290x. The W5000 is GCN architecture, which is the current AMD architecture that is entirely being embraced, and has the "AMD mystery connector" for future functions that AMD has not revealed yet, but will probably have to do with some kind of deeper system integration for enhanced performance.
It's not a gaming card, it's a professional level card with supreme performance, but at a very good price/performance level, and it's a card that you could use in your next system 3 years from now for instance. AMD always goes for bandwidth and streaming processors over number of transistors, they focus on overall performance and compute performance of their cards, not on software acceleration technologies, but on raw power and sheer data throughput speed. That's what you need for professional 3D graphics, because it means a huge poly count per second, and that's what it's all about. Probably in the next generation of gaming, the poly count will also become more important than things like anti-aliasing and other post-processing technologies, because the source textures are likely to become much denser because of more available VRAM and higher monitor pixel densities. So the W5000 is a wise choice in my opinion in more than one way, again, especially for the price. The W7000 has more bandwidth, but not that much more performance, and it does cost 50 % more, and it doesn't come is such a convenient form factor as the W5000, which is even compatible with an ITx build, it's really small, discrete and doesn't require that much power, when you see the card it's hard to believe that it spits out so many triangles per second, great stuff.
I use SuperMicro EEB boards, which might not fit your case, but if you don't mind EEB, I would definitely recommend the ASRock EP2C board with the Intel C602 chipset, which costs less than 300 dollars, and has some serious features and seriously good build quality. I like it better than SuperMicro boards and than the Gigabyte EEB board I bought recently. The ASRock gives the impression of a better overall finish, ASRock has made the BIOS with obviously more care than others, the implementation of the PCIe ports is great, with 4 full x16 PCIe 3.0 slots, and the parts used are of good quality. Yet it's not that expensive.
The ASRock board also runs on a single CPU if needed, so you can buy one now and add a second one later. I would go with an E5-1620, which costs less than 400 dollars, and is therefore cheaper than a consumer grade and feature-wise handicapped i7-4770k. That is a quad core with hyperthreading, so 8 logical cores. If you get a second one, you have a total of 8 physical/16 logical cores on your system, all running at 3.6 GHz. I wouldn't worry about longevity of that configuration, it's plenty powerful, especially on a board with 4 full blown PCIe 3.0 16x lanes, which can be filled up with up to 4 AMD firepro cards if needed for HPC grade 3D performance. And those E5 cores don't even require that much power, you don't even need a crazy PSU for them.
If you already have an SSD, I wouldn't invest in a further SSD solution quite yet, prices of PCIe SSD solutions will drop to consumer level in the future.