Multicore Strategy: Microsoft vs. Apple

I thought the headline Apple Steals Microsoft's Multicore Thunder with Grand Central (by Randall C. Kennedy) sounded like an interesting read. Quite honestly, I didn't know that Microsoft had any thunder to steal.

I haven't seen many articles talking about how Windows 7 was going to be a multicore wonder OS. Last year, it didn't show up as a big highlight on Paul Thurrott's website or even in the list from Microsoft developer blog. I know, because I wrote about this last year.

So where is this thunder hiding that the headline is referring to?

After reading the article, all I can say is, "Randall, did you just wake up after being drunk for several years, because you aren't talking any sense." This has to be one of the most out-of-touch stories I've read lately.

"As it stands, you'll need to reach into the 20- to 30-core range before the multicore efficiencies of the Vista/Windows 7 kernel finally allow the OS to overtake Windows XP in terms of raw performance."
  -Randall C. Kennedy, June 2009

I'm not of fan of Windows XP or Vista, but I can't believe that you would need a 20-30 core processor for Vista to equal XP. I've used both, and there just isn't that much difference. In fact, on a fast machine Vista is very usable. Maybe Randall hasn't tried running Vista on a quad-core Xeon computer with 6 GB of RAM (you could, if you had a Mac Pro, which also comes in 8-core versions), but it runs fine well before hitting the claimed 20 core beast Randall thinks it needs.

If I can't trust a Windows stooge to give me the straight scoop on Vista, I sure have a problem listening to him rant about how Apple does things. First off, the Mac OS and multiple processors go way back. The Mac clone builder DayStar was building dual- and quad-processor computers that ran the Mac OS way back in 1996. That predates even a dinosaur operating system like XP by five years.

I don't think Apple, which licensed the technology from DayStar, did much to fine tune the Mac OS to support multiple processors back in 1996, but they sure haven't been waiting until this year to catch up. The announcement from last year regarding Grand Central is just the progression of many years of working with multiple processor computers. (The modern era of multiprocessor Macs begins with the dual CPU "Mystic" Power Mac G4 in July 2000) Multiple processors have always given Mac OS X a boost in speed.

"Windows XP is the OS that simply will not die, and for good reason: It's mature, stable, and - most important - fast as the wind on today's hardware. In fact, Windows XP outpaced its younger siblings by a factor of two during multiprocess workload testing...."
  -Randall C. Kennedy, Jan. 2009

While Randall is right that Microsoft baked better multicore support into Vista three years ago, he is clueless on how long Apple has gradually and continually been adding multicore support to Mac OS X. Randall, instead of complaining about Apple's better skills at marketing, you should have done a better job comparing current multicore support between the two operating system families.

The issue for both Apple and Microsoft is how to deal with more than four cores. Code hasn't been optimized for that many, except for specialized hardware. Both Apple and Microsoft are in the same boat of getting developers to work with a growing number of processors crammed in to future computers.

"...these multicore-optimizing tweaks don't seem as extensive as Apple's parallel processing plans in Snow Leopard, headlined by what it calls Grand Central Dispatch.
  "What's key about GCD is that if it works like Apple says, it'll make easy for app developers to use multiple cores by handling threading for the programmers.
  -Matt Buchanan, Jan. 2009

There is no doubt that they are going to have similar sounding technologies. Are they going to "borrow" ideas from each other to further development? Probably, but common problems often lead to similar solutions.

Since I haven't heard much thunder from Microsoft about multicore support, all I can say is that Grand Central and OpenCL sound like incredible improvements. They should improve things not only for the next round of hardware, but also make major improvements for the Intel Macs I currently own. These technologies are not only about improving Apple software; Apple is giving these tools to developers to build better applications.

The bottom line is that no matter how good the headline reads, the next time a nutty yap tries telling you that Macs are only doing something to steal Microsoft's thunder, just smile and walk away as quickly as possible.