Continued from page 2
Curated Test Results and Closing Comments
What we know about Apple’s new Mac Pro is that it has two GPUs standard and that all of them are blazing fast in OpenCL performance. The D500 and D700 utilize the Tahiti Architecture while the D300 is Pitcairn-based. We covered a lot of details on these CPUs in our popular article, “The Mac Pro, So what’s a D300, D500 and D700 anyway? We have answers,” 24 Oct 2013)
Somewhere on the Internet I read a comment where a user assumed Adobe would never update its applications to truly take advantage of the new Mac Pro’s innovative dual built-in GPUs as standard. However that will likely not be the case. On Jeff Tranberry’s official Adobe blog he writes of the new Photoshop CC 14.2 update that Adobe has updated the latest version of Photoshop CC for the new Mac Pro.
Specifically, Adobe will modify Photoshop so it takes full advantage of one of the two available GPUs. He also says that Adobe expects to “add support for both of the new Apple Mac Pro GPUs, and to continue to optimize” for their customers.
Final Cut Pro X Testing
Here are references to Final Cut Pro testing and benchmarking:
- FCP.com magazine generated the test results we first spoke about.
- Macworld Mac Pro benchmarking series Part 1 (Final Cut Pro X benchmarks)
Adobe Photoshop CC Testing
- Macworld Mac Pro Photoshop CC testing (focused on OpenCL)
- MacPerformance Guide Photoshop benchmarking, great set of tests
General Graphics sand Gaming Testing
We want to point out some key findings in these various tests. Firstly, for 3D and CAD users the Anandtech Cinebench R15 tests truly showcase what Steve Bell of Archiform 3D talks about in his criticism of the new Mac Pro–that is that you want lots of cores AND high frequency as well. Clearly, the multi-threaded test shows that the 12-core Mac Pro is a monster and you can also see it in the Anandtech Blender 3D tests as well.
For Maya and modo users digging into David Girard’s tests over at ArsTechnica is a worthwhile trip.
- Macworld Mac Pro (Unigine Heaven and Valley testing)
- Macworld’s Cinebench R15 OpenGL Testing
- Electronista Mac Pro Cinebench R15 testing
- Ars review by David Girard which covers Maya, modo, etc…does a correction from a previous review that discovered a memory issue in Maya that limited performance. Tests the D700 GPUs
- Ars original Mac Pro review. This review does a superb job of digging into the details that Architosh broke with about what the D300, D500 and D700 were actually. We had all the essential correct and the clock decreases were due to TDP (thermal design requirements) for Mac Pro
- Motion and After Effects Time Lapsed testing. This is a very detailed review.
- DaVinci benchmarking of the new Mac Pro
- BareFeats DaVinci benchmarking on the new Mac Pro
- Anandtech’s Blender tests on the new Mac Pro. This test really shows the advantages of more cores in rendering for a well-architected, highly threaded application. It also points out the disadvantage to Intel’s i5 series chips which don’t support support Intel HyperThreading (virtual threads)
- Anandtech Cinebench R15 Single Thread Test – Mac Pro vs Late 2013 iMacs — these results are very useful to see just how impactful low frequency is for 3D apps when in single-thread mode (editing, scene creation, general use, most modeling functions, etc)
- Anandtech Cinebench R15 Multi Thread Test – nothing can touch a 12 core Mac Pro
General Apps and Benchmarks Testing
- Macworld Mac Pro Speedmark 9 Tests
- Geekbench by Cult of Mac tests (download Geekbench here to test your own Mac)
Boot Camp Testing
Some people may want to buy a Mac Pro and run Windows apps in Boot Camp, making their Darth Mac their favorite Windows box. People have been looking at things like SolidWorks, CATIA, and of course Windows games to run on Macs for years. Anandtech did test the Mac Pro under Boot Camp but with respect to gaming.
- Anandtech Boot Camp Mac Pro Testing — the only point of this was to show CrossFire in Windows mode, but measured against games on actual Windows machines also in CrossFire mode the new Mac Pro doesn’t do as well because it’s a workstation dual GPU setup against high end gaming cards. Importantly however, the BioShock tests, also in native OS X without CrossFire, of course shows that if CrossFire did come to OS X or something like it, the results are nearly 2x as good.
Closing Comments and Recommendations
This compendium of notes and discussion, along with our two companion features, highlight a very healthy dose of key issues involved in the new Mac Pro purchase decision process. The key take-away points should include:
- The new Mac Pro was nearly purposely made for Final Cut Pro X and is the ultimate box for that tool
- OpenCL 1.1 or 1.2 accelerated apps will shine best with the Mac Pro
- Apple’s apps have been specifically rewritten to leverage the two GPUs in tandem with one GPU being dedicated entirely to computational work via OpenCL
- MARI is a model 3D app written to leverage dual GPUs with one GPU entirely dedicated to accelerated OpenCL work while the other GPU powers the requirements of drawing to the display(s)
- CPU bound applications benefit if they are highly multi-threaded, more so for the 8 and 12 core models
- The quad core i7 iMac with Turbo Boost to 3.9 GHz may beat the quad-core Xeon E5 with Turbo Boost up to 3.9 GHz on many single-threaded operations that are not long-duration or sustaining operations. We qualify this against “sustaining” single-core features because Turbo Boost is responsive to internal CPU temperatures and the i7 in the Mac Books and iMac have a much lower thermal threshold point where Turbo Boost shuts down than on the Mac Pro–which has superior cooling with its Unified Thermal core.
The last point above deserves closer attention. What single-threaded functions will push a CPU? What multi-threaded functions will push a CPU? The latter is easier to answer for most of us. Rendering animations is a very clear example for the latter.
In helping to decide which computer to get–between the fastest iMac and the lower entry Mac Pro–for a single-threaded dominated workflow, the reader is advised to consult their software application developer directly for a detailed explanation. Especially if it is not accounted for in any of the testing results referenced above. —- ANTHONY FRAUSTO-ROBLEDO