SEARCH
diglloydTools™
Turbo Boost and the Intel Xeon W CPUs in the Apple iMac Pro
Need help deciding? Lloyd offers one-on-one consulting on choosing the best Mac and its best configuration, backup protocol, etc.
Is the iMac Pro faster or slower than the 2017 iMac 5K? Yes.
MPG will be testing the iMac Pro in very early January. OWC has offered its lab facilities and the iMac Pro to test with, and there might be some bonus tests with faster CPUs, which could prove interesting. As well as a slew of Thunderbolt 3 tests.
It remains a mystery as to why Apple is using downclocked Intel Xeon W CPUs in the 2017 iMac Pro. It doesn’t look to be TPD (total power dissipation) since all the CPUs have the same TDP. See Intel® Xeon® Processor W Family. Possibly it is a pricing thing—one quality bin down with disabled/dead CPU cores and/or CPUs that did not make the full clock speed cut.
One other test of serious practical interest to Lloyd is iMac Pro power consumption with screen lit up but the machine more or less idle. For the 2017 iMac 5K, this is 48 watts, which is wonderful when running on battery power for 8 or 10 hours a day.
SOFTWARE VENDORS: if you have multi-core software you want tested, contact Lloyd and provide a 4-CPU license (no demos, must allow 4 machines for simulataneous testing).
If we assume that all cores are used, no memory or cache contention (impossible in any real-world task), then one computation goes as follows as to maximum computing power, which is almost never realistic for desktop computers:
8 cores @ 3.7 GHz = 29.6 GHz ~= 9.9 cores @ 3.0 GHz
8 cores @ 3.2 GHz = 25.6 GHz ~= 8.5 cores @ 3.0 GHz (hobbled Apple offering)
10 cores @ 3.3 GHz = 33 GHz ~= 11.0 cores @ 3.0 GHz <== sweet spot
10 cores @ 3.0 GHz = 30 GHz ~= 10.0 cores @ 3.0 GHz (hobbled Apple offering)
14 cores @ 2.5 GHz = 35 GHz ~= 11.7 cores @ 3.0 GHz
18 cores @ 2.3 GHz = 41.4 GHz ~= 13.8 cores @ 3.0 GHz
The best balance is 8 cores @ 3.7 GHz or 10 cores @ 3.3 GHz because contention and overhead come to bear. My guess is that 10 cores @ 3.3 GHz is the best all around CPU, but Apple won’t sell you that CPU—Apple hobbles both the 8 core and 10 core CPUs by downclocking them.
The compute command of diglloydTools MemoryTester provides graphable data which shows how well the CPU cores of the machine scale to a multithreaded job.
Very few tasks can scale linearly* to use 14 or 18 cores, therefore these are pointless for most all users.
* To scale linearly (impossible even if only because of OS scheduling overhead), means that twice the cores would get the job done in half the time. Since clock speed drops quickly away from Turbo Boost speeds, scalability across core count can never occur with the Xeon W processors. Worse, contention for cache memory and main memory slows all the cores down (think “tollbooth”), hobbling real-world performance significantly. Any disk I/O at all kills scalability by placing a hard upper limit on throughput that might limit, say, at most 5 or 8 or 12 of 18 cores being able to run without waiting for data.
Intel Xeon W Turbo boost data?
Choosing the right CPU for the 2013 Mac Pro was made easier by knowing my own workflow together with having the data for the turbo boost speeds.
So far, turbo boost information seems to unavailable. The Xeon chips in the 2013 Mac Pro had this information as shown in the table and graph below. It would be a lot easier to predict some of the performance behaviors with such information, and it is why Lloyd spent the money for the 3.3 GHz 8-core CPU upgrade—see graph below—no other CPU beats it, regardless of the number of cores (up through 8 CPU cores‚. So it is always a win on speed, or at least never a loss.
Below, turbo boost speeds for the 2013 Mac Pro with Intel Xeon E series CPUs. Data not available as this was written for the Intel Xeon W series as found in the iMac Pro.
Grayed-out rows are CPUs not offered by Apple
Below, turbo boost speeds for the 2013 Mac Pro with Intel Xeon E series CPUs. Data not available as this was written for the Intel Xeon W series as found in the iMac Pro.
