Almost no one out there seems to have discussed or reviewed the new Mac Pro's performance when it comes to compiling code, other than unhelpful Geekbench reports that don't take into account real workflows where only perhaps Xcode and similar build-tools are used. All such reviews seem to focus on its GPU / VFX / Audio related capabilities. Here's my recent experience using the Mac Pro as a developer.
I just received a 16-core Mac Pro yesterday, with 32GB RAM. I'm a developer and was hoping to cut down on the time it takes to compile code that roughly takes 2 minutes on an 2019 iMac i9 3.6Ghz 8-core, 128GB RAM. I was hoping to bring this down to 1 minute at least with a 16-core, but was surprised that after numerous tests on numerous different projects, the new Mac Pro took either exactly the same amount of time to compile the same code on an iMac, or was at times only 0-5 seconds faster than the iMac in terms of compilation speed. The projects involved have a mix of C / C++ / Obj-C and Swift code. Surprisingly almost zero gains with the new 16-core.
I then compared this with a Macbook Pro 16inch (2.4Ghz 8-Core i9 with 64GB RAM) and the MacBook Pro only took 3-10 seconds longer to compile the same projects.
This was a pretty shocking discovery after having spent $9,000+ on the new Mac Pro. After a day of using the Mac Pro, I'm unimpressed with it's performance and have called Apple to return it. It seems the much cheaper iMac gives value for money and I'll wait for the 10th Generation Intel chips for the iMac instead, however long it takes for Intel & Apple to come out with these.
The Mac Pro will only be of benefit to very select workloads. An iMac Pro already offers plenty of cores, and even the regular iMacs now do with Coffee Lake Refresh, as you've discovered.
If Apple were to offer a Coffee Lake Refresh version of the Mac mini, it would probably be quite a good deal for a compiling build host. But they don't, so the iMac may be the best bang for the buck.
If Apple ships a Comet Lake-S update for the iMac ("10th generation" for this particular profile), that should bump it to up to ten cores. There may also be some slight single-threaded improvements. For example, Coffee Lake-SR only goes up to DDR4-2666, and Comet Lake-S might (might!) bump that up to 2933. So, slightly faster memory. At the same time, packing even more cores in the same process inevitably means that a 10-core version of the same CPU will actually have lower clock rates than your current chip.
We probably won't see real single-perf bumps for another two generations: not with Comet Lake-S in 2020, and not with Rocket Lake-S in 2021, which is still 14nm, but with the generation
after that. But for multicore, there's likely to be a ten-core Comet Lake-S-based iMac next summer.
This may as well reflect on how poorly Xcode utilizes multiple cores but the monitoring I performed suggested otherwise - it was using all 16-cores at peak times during compilation, yet it didn't do that well compared to the new 8-core i9s.
That's interesting, actually.
My understanding is that the Swift compiler is 1) very slow, but 2) fairly good at taking advantage of many cores. This is quite unlike, say, C#'s Roslyn compiler, which I've found to scale poorly to multiple cores (to be fair, it's a bit hard to compare JITC and AOTC compilers). However, as you've said, I've seen very little feedback on how the Mac Pro fares for developers. I have seen plenty of people praise the iMac Pro in that regard, but it's getting a little long in the tooth, and I'd be wary of choosing it over the iMac at this point. (The iMac Pro does have some non-obvious advantages, such as a better cooling system.)
I wouldn't entirely rule out that your Mac Pro is hampered by initial setup (Spotlight indexing, whathaveyou) and that your tests aren't reflective of long-term performance. Regardless, my personal recommendation, especially if you're happy to save multiple thousands of dollars, is to forego 'pro' altogether and just spec out a high-end iMac.
