Thanks to the Other World Computing's test lab, we have some preliminary benchmarks on the entry level 'late 2013' Mac Pro 3.5GHz Hex-Core. We decided to see how it compared to the 'mid 2010' Mac Pro 3.33GHz Hex-Core.
CPU GRAPH LEGEND
2013 MP 6c 3.5GHz = 'late 2013' Mac Pro Xeon E5-1650v2 3.5GHz 6-Core with 16G of 1866MHz DDR3, Dual FirePro D500s, and 256G PCIe flash storage
2010 MP 6c 3.33GHz = 'mid 2010' Mac Pro Xeon W3680 3.33GHz 6-core with 24G of 1333MHz DDR3, Radeon HD 7950, and Sonnet Tempo Pro SSD
Both were running OS X Mavericks 10.9.1.
CPU CRUNCH TESTS
Cinebench R15 -- Render 3D Model using all cores. (LARGER number means FASTER CB Rating)
Geekbench 3 -- Here is the overall score for 64-bit performance of Integer, Floating Point, and Memory (HIGHER score means FASTER.)
PCIE FLASH STORAGE TESTS
STORAGE GRAPH LEGEND
2013 MP 6c Factory = 'late 2013' Mac Pro Xeon 6-Core with Factory 256G PCIe flash storage
2010 MP 6c Accel*2 = 'mid 2010' Mac Pro Xeon 6-core with two Mercury Accelsior_E2 PCIe SSD cards striped (RAID 0) using Disk Utility
2010 MP 6c Tempo = 'mid 2010' Mac Pro Xeon 6-core with Sonnet Tempo SSD Pro PCIe card.
LARGE SEQUENTIAL TRANSFERS
We used AJA System Test to test file level, large sequential transfer speed using a 16GB test document that simulates a video whose frame size is 4096x2160. (HIGHEST number means FASTEST.)
SMALL RANDOM TRANSFERS
We use the Standard Random test in QuickBench as a predictor boot volume performance and a simulation of an application that does multiple small random transfers. (HIGHEST number means FASTEST.)
OPENCL GRAPHICS TEST
LuxMark 2.1 is an OpenCL benchmark that renders scenes of various complexity. For this round we used the default Sala scene (488K Triangles) and rendered using GPUs only. (LARGER number means FASTER in Thousands of Samples per Second.)
LUXMARK OPENCL GRAPH LEGEND
2010 MP 6c 7970*2 = 'mid 2010' Mac Pro 6-Core with dual Radeon HD 7970 GPUs
2013 MP 6c D500*2 = 'late 2013' Mac Pro 6-core with dual FirePro D500 GPUs
2013 MP 4c D300*2 = 'late 2013' Mac Pro 4-core with dual FirePro D300 GPUs
2010 MP 6c 7970 = 'mid 2010' Mac Pro 6-core with single Radeon HD 7970 GPU
2010 MP 6c 7950 = 'mid 2010' Mac Pro 6-core with single Radeon HD 7950 GPU
OPENGL GRAPHICS TESTS
Valley Benchmark "flies" through forest-covered valley surrounded by vast mountains, presenting a birdŐs eye view of 64 million meters of extremely detailed terrain down to every leaf and flower petal. It features advanced visual technologies: dynamic sky, volumetric clouds, sun shafts, DOF, ambient occlusion.
We used Extreme preset: 1600x900 windowed resolution with 8x Anti-aliasing, Ultra Quality for Shaders and Textures, and with Occlusion, Refraction, and Volumetric Shadows enabled. We posted AVERAGE frames per second. (LONGER bar means FASTER.)
OPENGL GRAPHS LEGEND
2013 MP 6c D500*2 = 'late 2013' Mac Pro 6-core with dual FirePro D500 GPUs
2013 MP 4c D300*2 = 'late 2013' Mac Pro 4-core with dual FirePro D300 GPUs
2010 MP 6c 7970 = 'mid 2010' Mac Pro 6-core with single Radeon HD 7970 GPU
2010 MP 6c 7950 = 'mid 2010' Mac Pro 6-core with single Radeon HD 7950 GPU
2010 MP 6c G780 = 'mid 2010' Mac Pro 6-core with single GeForce GTX 780 GPU
2010 MP 6c G680 = 'mid 2010' Mac Pro 6-core with single GeForce GTX 680 GPU
GPUtest - Volplosion - OpenGL (2.1 or 3.2) pixel shader test run at 640x480 Windowed. (LONGER bar means FASTER.)
GPUtest - TessMark x64/x64, OpenGL 4.0 tessellation test run at 640x480 Windowed. (LONGER bar means FASTER.)
INSIGHTS
When it comes to CPU crunching, the 'late 2013' Mac Pro 6-core is faster than the 'mid 2010' Mac Pro 6-core in both single and multi-core tests.
Since the newest Mac Pro has PCIe flash STORAGE, we installed PCIe based flash storage cards in the 2010 Mac Pro. It's encouraging to know there are equally fast flash storage options for those of you keeping your Mac Pro towers for now.
We featured LuxMark is an OpenCL benchmark since it uses multiple GPUs. We haven't tested the FirePro D700s yet, but you can see, the D300s and D500s beat the fastest solo AMD GPUs we installed in the 2010 Mac Pro. Dual FirePro D700s should match the KSamples/sec achieved by the dual Radeon HD 7970s we tested.
OpenGL Benchmarks are not multi-GPU aware -- at least not at this time. That's why the single AMD GPUs in the Mac Pro tower match or exceed the dual D300 and dual D500 framerates. When we post Final Cut Pro X 10.1 results, the FirePros in the new Mac Pro will have an opportunity to shine.
In SUMMARY, we purposedly equipped the 2010 Mac Pro tower with graphics cards and storage devices that closely mimic what the 2013 Mac Pro offers 'out of the box.' That way you can decide if you are ready to take the leap to the 2013 Mac Pro or whether to enhance your existing 2009-2012 Mac Pro tower to perform better.
MORE TO COME
This is just an appetizer. As we gain access to other models and configurations of the 'late 2013' Mac Pro, we will add more comparisons and feature real world apps.
