With the help of "FENN" on the MacRumors Forum, we have some early GPU benchmark results for the 27-inch 'late 2012' iMac with the GeForce GTX 680MX GPU. For now, we compare it to the 2011 iMac with Radeon HD 6970M GPU and a 'rogue' Mac Pro with the GeForce GTX 680 Classified GPU -- for perspective.
MP 680 C = 'mid 2010' Mac Pro 3.33GHz Hex-Core Westmere with unflashed GeForce GTX 680 Classified GPU (4G VRAM) connected to 27" LED Backlit Cinema display (2560x1440)
iMac 680MX = 'late 2012' iMac 3.4GHz Core i7 with GeForce GTX 680MX (2G VRAM) and 27" LED backlit built-in display.
iMac 6970M = 'mid 2011' iMac 3.4GHz Core i7 with Radeon HD 6970M (2G VRAM) and 27" LED backlit built-in display.
All tests were run under Windows 7 since that's what FENN was using. We will, of course, post the results under OS X in a few days.
WHAT DID WE LEARN?
1. The 'late 2012' iMac 27-inch with the optional GeForce GTX 680MX GPU is signficantly faster than the best GPU in the 2011 iMac -- with the exception of the LuxMark OpenCL benchmark.
2. We included the Mac Pro with the "unflashed" EVGA GeForce GTX 680 Classified GPU to make two points:
- The mobile GPUs in the iMacs and MacBook Pros are not equivalent to similarly numbered desktop GPUs that can be installed in the Mac Pro.
- Many serious, hard-core gamers and pro 3D creators would rather have powerful desktop GPUs if Apple or third parties will make them available for the past, present, and future Mac Pros.
MORE TO COME
We're just getting started. Many real world tests will follow and we will add other models and configurations of Mac.
BE A REMOTE MAD SCIENTIST
If you want to contribute to the effort, you can run the following free benchmarks on your Mac and send us the results. Just be sure to indicate whether they were run under OS X or Windows. And please use screen capture (Command+Shift+4, drag, select, click on OS X) to save the results. That preserves the integrity of your results.
Unigine's Heaven Benchmark runs as OpenGL under OS X and DirectX under Windows. If your display is set to 2560x1440, it will run at that resolution (called 'System' in the status window). We prefer that rez for this comparison. Don't change the other default settings. Click "Run" and then "Benchmark". It will "fly" through 26 scenes. The average FPS will be displayed in a pop-up window after all scenes are processed. Send us a screen shot of the pop-up window or save the results in a file and send us that.
On the second run, use 1280x720.
Luxrender's LuxMark is a good OpenCL GPU render test. The default test is Sala scene with GPU only -- which starts running at launch. That's the one we want your results for. When complete, a Pop-up window will display average Ksamples/second. Send us a screen shot of that window. (You can render the scenes with GPU only, CPU only, or Both but we want GPU only.)
Geeks3D's GPUtest has 3 OpenGL tests. Run FurMark and GiMark. Set at 2560x1440 and default settings. We get slightly faster results with "fullscreen" unchecked. Click "Run benchmark" button and wait for the average to be posted in a pop-up window. Send us a screen shot of that window.
Make a second run of FurMark and GiMark at 1280x720.
FutureMark's 3DMark 11 is a Windows only DirectX 11 benchmark. Run the default "Performance" preset (1280x720). Do NOT use custom settings. You can upload your results to their website, but we want a screen dump of your results.
Thoughts? Questions? Contact
Also, you can follow him on Twitter @barefeats