Review: Nvidia's Quadro K5000

Post by Matt Silverman

My company Swordfish was established in October 2012, focusing on motion design for marketing and software development. As a startup, I tried to make smart, economical decisions, and having owned a few previous companies in this industry, I’d learned a few key lessons that I was determined to take to heart. One of those lessons is that equipment quickly becomes obsolete, and I would rather invest in brains than boxes.

I found that the best bang for our buck was to hunt down a few used 8-Core Mac Pros on Craigslist to get off the ground, and continue to accumulate more of these workstations as our jobs required more artists. These Macs averaged $1,200 each, with some requiring a few hundred dollars extra for additional RAM to bump them all over 32GBs. Over the past 18 months, we accumulated 12 of these workstations. Uberware Smedge software allows us to render across these workstations in the background while artists continue to work, and turns the studio into a full renderfarm at night. We wrote custom submit scripts for After Effects, Maya, Cinema4D, Houdini, Nuke, and Handbrake to help automate queueing, allowing freelancers to quickly get up to speed. If renders are too intensive for our internal farm, we send the renders to the Amazon cloud using Zync Render (Maya/Nuke), or Houdini’s native support. We also use the cloud for nightly backups to Amazon S3 and project archiving to Amazon Glacier.

While the cost effective CPU solution we found helps crunch through the heavy lifting, it doesn’t help improve the user experience while setting up our projects. That’s where a GPU comes into play. Traditionally associated with 3D acceleration, in recent years many 2D applications started to take advantage of the GPU, accelerating everything from color correction to motion tracking. The GPUs contained in most of our used Macs were the stock cards, adequate for most of our design tasks. For heavier 3D tasks, we added Nvidia Quadro K5000s shortly after they became available. read more...