The (eventual) release of Apple's 27" cinema display led to some serious gadgetlust and a not baseless belief that my productivity would benefit from being able to keep lots more stuff on the screen at once. However, when push came to shove it became apparent that it wasn't going to be possible to hook it into my now clearly ancient MacBook Pro (which, actually, I was really very happy with - despite it being nearly four years old) and there was much bitching and whining. However, a quick look at the prices of iMacs and it becomes apparent that it's only another $1200 to have one that has a computer built in - and that computer would be heaps faster than my MBP. A little staring at benchmarks (an even more ridiculous one) was done and I found myself shelling out damn nearly four grand for a quad core i7. Good god, these guys can market.
Do I love it? Do I ever. The screen is huge, perhaps very slightly too large given that I wear glasses. It's bright and, being an IPS display, keeps it's colour at differing angles. It also does a reasonably good job of that "close to the surface of the glass" thing that the retina display does so well. Resolution at 2560x1440 is completely outrageous. BTW, I chose to not have the wireless keyboard instead choosing the "with keypad" keyboard for zero additional dollars (but none back, either).
Performance varies from pretty much what you'd expect (i.e. blocked around the disk) to just incredible under the right circumstances. I XBench'd all three machines on my desk at the moment and got the following results:
So on the surface it appears nearly twice as fast as the MBP and around 10x as fast as my circa 1998 dual G4 but this is not quite all the story. HDD read performance has gone from 43MB/sec sequential and 21MB/sec random on the G4 (a bloody great result IMHO); pretty much exactly the same on the MBP; to 124/33 on the stock drive for the iMac. Obviously an SSD would bump these numbers significantly but I don't do a lot of HDD blocked work and neither am I going to hand over that kind of money.
Quartz performance went from 6610 alpha'd rects/sec on the G4 to 55100 on the MBP (yes, nearly ten times faster) 109610 on the iMac (twice as fast again). Note that the G4 actually has a newer than stock video card in it - mainly to support quartz extreme so the actual difference in quartz performance from the older to the newer machines is even greater than it looks.
Other "down to the metal" benchmarks show the system memcpy going from 242MB/sec to 2391MB/sec and 12596MB/sec on the i7. Yeah, I had to read that again too, they are right - quite possibly XBench fits entirely in the i7's 8MB level 3 cache. It's still ludicrously fast. Streamed memory copies go from 563MB/sec using Altivec on the G4 to 2391MB/sec on the Core 2 and 8240MB/sec on the i7. So the on-die memory controller in Nehalem was worth it, then?
Best of all is the time taken to actually do some work :) I did a build of MixTape and an incomplete run (I failed to symlink some directories or something) on the G4 took six minutes. The Core 2 got this down to 46 seconds (an actually functioning build, including the installer) and 12 seconds on the i7. Twelve - for both PPC and Intel, including building three frameworks.
So I guess a tl;dr history of the last ten years of mac development would say that the biggest speed bump came with the move to Intel and, in particular, Core 2. It also helped that this coincided with the meatiest bit of development in GPU's. Since then we've pretty much "only" scored a factor of two in the last five years or so although there are rich pickings to be had for applications that can keep their working sets small and make use of all those hardware threads. None of this should be news but it is interesting to see it happening on one's desk.
Oh, and there's a new build of MixTape that uses the display link for refresh and it looks beautiful on the iMac.
Previous next big things include development of the capture and intermediate compression technology in iShowU-HD; design and implementation of a small advertising network; the refinancing, technical direction, and a lot of the donkey work for Virtual Katy; technical direction, project management and (again) donkey work for VoiceQ; creating code and intellectual property around load balancing that was acquired by Allied Telesis; and the research and an implementation of the h.264 video compression protocol.