The 64bit Question

Okay, so you’re pricing out a server or workstation and run-across an AMD64-or Intel 64 -branded processor. But what’s the difference between the 32bit technology and 64bit technology?

32bits refers to the processor’s bus size: in one data stream, how many bits can we shoot through the processor per cycle at one time? In parallel, we can bring 32-bits to the processor for every cycle and, if your processor is operating at 3.2 ghz (3.2 billion cycles/second), you can see where 64bit technology may have a significantly higher performance benefit. A 3.2 ghz 64-bit architecture could literally double processing throughput.

There’s also some additional benefits to 64bit microcomputer architecture. We’re not limited to 4gb of RAM. 32bit systems can only logically address (“see”) 4gb of RAM. 64bit technology doesn’t have this limitation. As a 64bit platform, it can actually support up 16 exbibytes (17,179,869,184 gigs) of RAM. WOH! 16 exibity-whatas?! Dude, imagine Doom 3 with 32 gigs of RAM! Yeah, a machine that can address that much RAM on my desktop computer? Sign me up!

Well, okay, hold on, ace, because there’s a dirty little secret or two you need to know first.

One is the 16 exbibyte myth. Today’s processors can only address up to 40 bits of _physical_ memory, so there’s a _physical_ limitation on RAM here of up to 1 terrabyte, and a _virtual_ limitation of 48-bits, or around 256-terrabytes. Operating systems like the 64bit version of Windows can only address up to 128 gigs of RAM but virtually address (through the swapfile on the hard drive) the 16 exibytes of RAM. So there are some practical limitations here.

Also, 64-bit platforms require 64-bit operating systems. The kernel for the o/s must be compiled as a 64bit version to take advantage of the platform. The same goes for applications. This means, for example, you need a 64bit version of Windows, a 64bit version of Explorer, 64bit drivers, a 64bit version of Office, a 64bit version of SQL Server, etc.

Managing this is a chore. A lot of people don’t realize that most, if not all, of their applications are compiled as 32bit applications and – although they’ll run on the 64bit platform – they won’t be able to address all that memory or take advantage of the speed benefits of a 64bit platform. 64bit drivers come out late on the market. Most apps aren’t compiled as 64bit versions. So although you’ve paid premium, you’re still going to get mediocre results.

Once you start pricing the 64bit version of desktop and server applications, you’ll start thinking: wait a minute. I have a 64bit platform and I’ll end up paying a premium for it, but I have 32bit operating systems and applications. I’m paying for something that I won’t use.

Where 64bits really matter, in todays terms, are in two spaces. The professional workstation space where you’re dealing with a souped-up microcomputer to do some amazing things with CAD or graphics (say, you want to build a movie company to rival Pixar). And the server space where you need this kind of footprint (and expense) to run your company and manage its growth over time.

If you aren’t planning to completely replace your o/s and applications, or need exceptional speed and memory beyond traditional commercial capabilities (to run Office, for example), think twice about the 64bit investment.

If you’re looking to scale your computing experience and literally make your box sing – because your needs truly justify the expense, and your mission-critical application set comes in 64bit versions – think about the 64bit technology.

R