Maya Render Shootout

Originally posted March 14th, 2005, by rob-ART morgan, mad scientist. Updated March 28th, 2005 with more render tests. We hope this page is the start of an interesting Maya performance comparison page. We prefer to test ourselves so we can verify the preferences and settings, but will consider accepting test results from remote mad scientists. To "prime the pump," here's what we observed when we tested a Dual Xeon and Dual G5 side by side: SAMPLE MAYA PROJECT FILE We discovered the Maya Test Center while "googling" around to see what others use to benchmark Maya. We downloaded the test file and got busy. SOFTWARE RENDER For the Software Render, we opened the test file in Maya (all preferences and settings at factory defaults). We set Render Globals to "Maya Software." We clicked the Maya Software tab and chose Production Qualty. Then, after selecting the upper right window, we clicked the Render icon to start the rendering process. On completion, we opened the Console Window and noted the "Time For Frame Render" (second line from the bottom). HARDWARE RENDER We then selected Maya Hardware in the Global Render Settings window. We clicked the Maya Hardware tab and made these mods: Presets = Production Quality with Transparency Number of Samples = 36 Color Resolution = 1024 (Bump Resolution = 1024, too) Texture Compression = Enabled Motion Blur = Yes (checked the box) (WE DIDN'T CHANGE ANY OTHER SETTINGS.) We selected the upper right view and clicked RENDER icon to render. When it was done, we opened the Console Window and noted the total elapsed time next to "hardware rendering frames" (top line). We noticed if we clicked RENDER again, the second run was much faster, but we used the results from the first run in the graph above. ANOTHER MAYA RENDER TEST WE TRIED A reader suggested we try the ZooRender render test sample. It's pre set to run a "mental ray" type rendering if you just click the render icon after opening it. We didn't see any render times in the console (or any other) window, so we used our stopwatch. We got a consistent 57 seconds on our G5/2.5GHz Power Mac. TEST FOR GRAPHICS CARD ADVANTAGE After more "googling," we found a test that measures the effect of different graphics cards on Hardware Rendering in Maya. It's a Particle test you can download. Following the directions, you can get playback speed in frames per second and hardware render speed. As you can see from the graph below, the graphics card you have in your G5/2.5GHz Power Mac does affect how fast you do hardware rendering, not just how fast you can playback an animation. INSIGHTS and COMMENTS Right out of the chute, the G5 Power Mac was beaten by the Dual Xeon. If you look at the database of results on the Maya Test Center, there are faster times posted for both the Xeon and G5, though the Xeon still wins. However, there is no description of what settings were used by those submitting entries, so that database tends not be a useful reference. That's why we're starting our own Maya results "database." We did a fresh install of Maya on our test units and left the settings at factory defaults except where noted above. The only reason we tweaked the settings for the Hardware Render test was to make it harder on both machines. (Both machines complete the Hardware Render in under 2 seconds when the default settings are used.) Although the sample test file incorporates the use of shadow maps, ray tracing, glows, 3D textures, and displacement mapping, we were not convinced it was "be all end all" sample possible. That's why we solicited advice from our readers who use Maya and did more searches on the net for Maya test files. Now that we found the ZooRender mental ray test and Particle graphics card stress test, we'll be going back to test the Dual Xeon and hope to add the Dual Opteron results soon, too. DOES MEMORY CAPACITY AFFECT MAYA's RENDER SPEED? We tried both render tests with 2GB, 4GB, and 8GB of memory in the G5. The results were the same. When we checked memory usage with Activity Monitor, Maya never used more than 175MB of real memory while loading and running render tests. By contrast, Motion used 883MB of real memory when we did a RAM Preview on one 300 frame project. When opened a second 660 frame project and previewed it, Motion grabbed a total of 2.2GB. So, unless you have multiple RAM hungry processes active, you won't gain speed by throwing gobs of memory at Maya. CALLING FOR MAYA TEST RESULTS We ran our tests with both version 6.0.1 and 6.5 with virtually the same results. So if you have either version, we would love to see your render results. So whether you have results to share or a better sample to try, please me. RELATED LINKS Maya Test Center database of results (Macs and PCs) or download their test file. ZooRender's Maya render test is worth a try if you have Maya 5 or 6 or scroll down that page to see their list of results. PainfulUrination has four Maya tests including one that stresses the graphics cards. Renderman benchmark results for Macs and PCs. WHERE TO ORDER YOUR APPLE PRODUCTS When ordering products from Apple Store USA, please click THIS TEXT LINK or any Apple display ad as your "portal" to the online store. In so doing, you help to support Bare Feats. For factory refurbished Macs, check with Small Dog and Power Max. WHERE TO BUY WINDOWS PCs There are many places to buy PCs. We want to plug WhisperPC since they provided the Intel Dual Xeon 3.4GHz. Has Bare Feats helped you? How about helping Bare Feats?