doncl

So, this entire article presupposes that the 'heavy workload' == 'graphics intensive'.   I'm a member of a neglected (in these kinds of articles) Mac-user demographic, i.e. a software engineer doing iOS.

And really, this demographic is not all that rarefied and unimportant.   We're the ones that actually make all the stuff that you guys run on your Macs and iPhones and iPads.

In any case, whether you think our viewpoint is one to be considered, or is an exotic edge-case, here it is:

0) Graphics performance matters almost not at all.   As long as it can run 3 4K monitors and see lotsa code, I'm good.   The integrated graphics on my MBP 13 are fully adequate.   Any sort of discrete graphics card is a waste of money, heat, and battery life.   I don't care about games, I don't care about Final Cut Pro performance, I don't care about Adobe Photoshop.   All that matters, to me, is how fast Xcode can compile Swift and Objective-C code, link it, and deploy it to a simulator or device.

1). My company bought me a 2018 i7 six core, and I personally have a 2018 i9 six core.   I can tell you the i9 finishes compiles incrementally faster than the i7 does.   It's not dramatic, but a 2 minute 58 second build comes down to 2 minutes, 20 seconds.   This is real-world performance that, magnified over the course of a workday, translates to substantially more code getting written.

2) My old despised 2013 Mac Pro 12-core 64 G 1 TB beast compiles almost as fast as the 2018 i7 six core.   And it's a lot quieter.    And it doesn't degrade compile speed no matter how hard I push it.

3) With both MacBook Pros, it seems to help the compile speed a little bit to run an app (TG Pro in my case) that takes control of the internal fans and runs them full-out.  I also put them on a cooling laptop pad, but it's not clear to me that this has any effect, except to raise the ambient white noise level in my workspace.

4) Configuring Xcode to put the DerivedData products on a RAMDisk helps quite a bit.    I suppose that's irrelevant to this discussion, though.

All I'm asking is that, when AppleInsider does analyses like this, you include some Xcode runs - get some large Swift codebases from GitHub (they abound there), and pick one as your testbed baseline.  Xcode is free from the Mac App Store, and for my money, is the single-most important source of compute workloads you should be evaluating. Run clean builds and incremental builds and time them.   Include that along with CineBench and Geekbench and Handbrake video file transcodes and Final Cut Pro renders, and whatever else you're doing.    I promise you, you will find that your video card matters less than nothing, when you view performance from this viewpoint.    And this is kind of an important viewpoint - software developers represent a significant subset of 'pro' users; Apple claims over 50% of GitHub open source submissions are done from Macs.

my .02
Don