With this version of Windows, Microsoft is promoting a choice of operating system that hasn’t been an issue in more than a decade. You now have the choice of 32-bit Windows or 64-bit Windows, and it’s not a decision to be made lightly.
For most users, the move to Windows 7 comes when they purchase a new computer. Most older computers are not going to be upgraded, although you may have that option. If your computer can run Windows Vista tolerably, it will certainly run Windows 7 and run it better than Vista. That’s because Microsoft did a superb job of reducing the overhead of the OS.
As a rule, you probably should not consider upgrading a computer that predates Vista. If the computer was manufactured after Windows Vista was rolled out in January 2007, then the hardware likely is new enough to handle Windows 7. I migrated my personal home-built computer system to 64-bit because it had the power. It’s got a quad-core CPU, 8GB of memory, a fast 1TB drive, and an ATI 4650 graphics card. Windows 7 64-bit runs flawlessly. But I wouldn’t even think of putting it on my father’s old Pentium PC with 2GB of RAM and a slow 160GB drive and integrated graphics.
So let’s assume you are getting Windows 7 on a new computer (or deploying them to your users by providing them with a brand-spanking-new system). Right away, many of your problems are solved because the new computer will come factory installed with all of the latest device drivers. Because Microsoft retained the Windows Vista driver model, the library of device drivers that grew over the three years since Vista’s release ensures quite a bit of compatibility for older devices.
During the Windows 7 development cycle, Microsoft made very sure to keep in touch with device makers and keep a two-way line of communication open so they didn’t get caught flat-footed as they did with Vista. That OS shipped and oops, a lot of people didn’t have new drivers ready to go. Microsoft made was motivated not to repeat that mistake.
At the same time, Microsoft pushed hard to get as many 64-bit drivers made as possible. While Windows Vista had a 64-bit version, it wasn’t that heavily promoted. With Windows 7, Microsoft is pushing 64-bits hard, and it needed driver support to do that.
Here’s what you have to remember: While you can run a 32-bit application on a 64-bit OS, you cannot use a 32-bit device driver. It must be 64-bit. So before you make the move, check that every device you use is supported. You can do that at Microsoft’s Windows 7 Compatibility Center.
Drivers are your most objective measure for whether to make the move to 64-bits. If the drivers for a device you need aren’t there, then the decision is made for you. But before we get too deep into the issues of 32- and 64-bits, let’s go over the differences and issues surrounding the bits.
The last time we went through this migration in bits was in the mid-90s, when Microsoft came out with the hybrid 16/32-bit Windows 95. It wasn’t a pure 32-bit OS, however. That’s because Windows 95 worked by booting 16-bit MS-DOS and then loading the 32-bit GUI and application layer over it.
For the 16-bit and 32-bit layers to communicate, they used a not-so-elegant trick called “thunking,” which mapped 32-bit code to 16-bit pointers, memory models, and address spaces and vice versa, and allowed 16-bit apps to run on Windows 95. It also allowed Windows 95 to have long filenames and get away from the 8.3 character file name limitations of 16-bit MS-DOS.
Starting with Windows 2000, Microsoft added a layer called Windows on Windows (WoW, not to be confused with World of WarCraft) that added some extra 32-bit support, like true long file names. WoW has been integral to Windows XP and Vista for running 16-bit applications, but over the years, virtually all 16-bit apps have disappeared.
Starting with Windows Vista and now in Windows 7, as well as in Windows Server 2008, Microsoft added WoW64, which does the exact same thing as the earlier version: It maps a lower bit application to a higher bit OS. In this case, it allows 32-bit applications to run on 64-bit Windows.
Natively, these applications won’t work without the compatibility layer, and because it requires a little extra memory to run the software, running a lot of 32-bit applications on a computer running 64-bit Windows might consume more memory than you expect.
Of course, there are two qualifiers to that issue. First, it’s not a lot of memory being used, just a few kilobytes per apps; and more importantly, the big benefit of 64-bit Windows is a lot more memory.
The great limiter of 32-bit computing is how much memory it can address, which is 4GB. When Intel introduced 32-bit processors in the late 1980s, 4GB of memory was almost incomprehensible. Today, you can purchase 4GB of RAM at your local Best Buy for $200. Progress is both truly amazing and scary at the same time.
Here’s a geeky experiment. In Microsoft Excel, type this in one cell:
=2^32. In the cell next to it, type
=2^64. Convert them to comma style formatting, reduce both cells to 0 decimal places, and widen the cells. You’ll see how many bits of memory each type of processor can access. So going from 32-bit to 64-bit isn’t a doubling of addressable memory; it’s several orders of magnitude larger.
This limit on 32-bit computers lead to a great computer mess that we still live with to this day. x86-based servers grew like mushrooms, with the buildout of servers to accommodate the Internet era a decade ago, and the ascendancy of both Windows Server and Linux as viable alternatives to Unix servers for hosting Internet Web servers .
Once you boot Windows on a 4GB system, by the time the OS loads, the apps load, protocols load, and everything else, you’ve lost half of the system’s memory. This meant servers could only handle a small number of concurrent connections. Companies had to deploy banks and banks of 32-bit servers, each handling a tiny amount of traffic.
At the same time, Intel was coming out with some decent server processors that ran fast but had nothing to do, since they could only handle a small amount of traffic. The result was row upon row of servers at 1-2% utilization, because Web serving requires very little CPU power.
This is why servers moved to 64-bits a lot faster than desktops. Unless you are a power user, 4GB is more than enough for most tasks. But on a server, memory is performance. A friend told me this 20 years ago and it still holds true. Put 16GB of memory in a server and it will be faster than when it had 8GB. Put 32GB in, and it still speeds up because it has to do less and less disk swapping.
Microsoft was one of the first to do this. When the company finally had a production-ready version of Windows Server, it rolling out servers using AMD’s Opteron processor. AMD was the first to embrace 64-bit computing in the early part of the decade.
Unlike an OS, no compatibility layer is needed for a lower bit OS on a processor. A 64-bit processor can run a 32-bit OS with no problems. If your computer is newer than seven years old, it’s likely doing just that. AMD was first with 64-bit processors in 2003. Intel got its act together with 64-bit Xeon server chips in 2004, desktops in 2005, and laptops in 2006.
In 2006, Microsoft replaced 250 32-bit servers running the MSN network with 25 64-bit servers, with 8GB of memory each. The performance increased dramatically because the servers had a great deal more headroom and they did significantly less paging memory to disk. Also, each server could handle more concurrent connections.
Making the 64-Bit Decision
That’s the true advantage of 64-bit computing: addressable memory. Your programs won’t run any faster. Any speed increase is likely to be more due to advances in processor performance courtesy of Intel and AMD.
So, unless you are doing CAD/CAM, video/audio editing, programming, desktop publishing, desktop database work, or working with massive spreadsheets, moving to a 64-bit system won’t do you much good. For a user running Word, a browser, Outlook, and a few background apps, there’s little to be gained.
Mind you, it can add up. Even basic business apps can consume memory fast. If you have a 4GB system and load Outlook, Word, Firefox, Photoshop, Premier, and a few more applications, yes, things will start to slow down. But for basic everyday work, no, 64-bits won’t get you much and you won’t need more than 4GB of memory.
So your first rule for making the decision to move is need. If you don’t need it, don’t rush. The world will eventually migrate to 64-bits but it will preserve 32-bit compatibility for a long time.
If by some miracle you are still running an old 16-bit application, it won’t run on a 64-bit OS. Microsoft looked at the work that would be required to make 16-bit apps work on WoW64 and decided, rightfully I think, that it wasn’t worth the effort for the amount of 16-bit apps left in the world.
The second rule of measure for whether to go 64-bit is driver availability. Microsoft has done a commendable job of getting current hardware up to 64-bit parity but there are no guarantees for older hardware. Are all of the drivers you need to run all of your hardware available? You can always run a 32-bit application on 64-bit Windows, but you cannot use a 32-bit device driver on a 64-bit system.
Also, because Windows XP is so old, there is a very rich library of device drivers and CODECs for audio and video, and not all 32-bit CODECs have been ported. So that’s something you have to check along with your device drivers. If you are doing audio or video work, this is an imperative.
The best argument for making the purchase is future-proofing your computer. You want it to last three to five years, right? Okay, suppose you find yourself running low on memory regularly. If it’s a 32-bit system and you are already at 4GB, well, you have a problem. But if it’s a 64-bit computer, you can drop another 4GB in it and you’re off to the races.
One last tidbit: if you are going to splurge on a new computer, and go 64-bit with lots of memory, do it now. Every market researcher is predicting the price of memory has been increasing for the past year and will continue to do so this year for a complex series of reasons. Memory has been obscenely cheap in recent years. It won’t stay that way.
Want more like this? Sign up for the weekly IT Expert Voice newsletter so you don’t miss a thing!