Cory's Blog

As an extension to the recent realization that I have a fair number of relatively old computers, I was thinking about the physical lifespan of computers. How long is a certain system expected to run before it just falls apart. At my workplace, our policy is that we'll support the software on a system until it falls apart or can't run current software any longer. We support hardware for five years, at which point you can use it but won't receive any help with it if anything should break.

I've personally defined and re-defined my own lifecycle policy a few times. It's of course difficult to have a single lifecycle policy when your needs change over time and you get different kinds of systems. It's also of questionable value to set a lifecycle policy in an individual or home environment, as it is probably better to re-evaluate your needs on a regular basis.

As part of that re-evaluation, this post will talk about the physical problems several of my computers have and some of the methods I personally use to evaluate when it's time to replace or stop using a system.

Superslab, my ThinkPad T400 and main computer for almost six years now, is definitely in the "wearing out" phase. I've had to replace both the disks I bought when it was new, its enclosure is starting to creak and crack, and the keyboard isn't quite what it was a few years ago when the system was new. In addition to all of that, I've replaced the thermal paste at least once so far, and its batteries (the second set it has had) are completely worn out. At first, I was planning on having it three or four years for my main computer, but aside from a few things I'll mention about virtualization later on in the post it never stopped doing what I need. Conveniently, I have written about the age of the T400 a few times before, although not always directly on my blog. There's an entry on superslab's wiki page talking about replacing it with a T420 or 430 right around the time those systems were new, and about not wanting to be out of luck if the system dies unexpectedly – managing the risk of a new system by keeping the old one around as well.

Not to be down on my ThinkPad, as it's far from the only computer I have with physical problems. My ThinkPad T42p has a fan that mostly doesn't function and is using harder to find IDE disks, in addition to having an issue with its UltraBay ejection mechanism and having a keyboard that's very worn out. Despite having been a very high performance system when it was new and being nearly totally decked out, it's clear that it's too slow to use as my main computer. It is still fast enough to run Windows and Office, but it's not fast enough for development tasks or for, say, managing my photos. Of course, it is an eight year old laptop and that contributes to a system that would feel old in use, even if it was in impeccable physical condition.

Even some of my newer systems such as the almost-two-year-old Surface RT are starting to show their age (or maybe just poor component choice up front.) The RT of course became my primary portable computer almost immediately, and is showing its age. It isn't very fast to begin with, but its particular issues are definitely center around the magnetic pogo-pins power connector Microsoft chose for the device and the detachable keyboard. In addition, the paint is wearing thin or has been nicked off in many areas, and its non-replaceable storage is nearly full.

Desktops have fewer longevity issues, both because as parts wear out they can be replaced, and because physical maintenance on them is easier, but there is also the question of at what point a computer is simply over the hill.

I like to think that my computers aren't over the hill (in terms of performance), but we're at the point where the Intel i5-4300U processor in most of the Microsoft Surface Pro 2 and Pro 3 configurations is twice as fast as the Intel Core2Duo T9400 processor in my ThinkPad, and every other subsystem of the computer is twice as fast as what I've got today. This is all in a system that's about as thick as an iPhone 5.

The decision I still need to make is whether twice as fast and well under half the weight and girth and a long battery life is worthwhile, or if I have some specific need for more computing horsepower than I already have. (You knew this was the inevitable direction of this post.) Despite being old and being put into situation where physical durability are very important, the Surface RT still handily runs Outlook, Word, and Internet Explorer.

The problem with the Surface RT is, as I mentioned above, that it could use some new parts such as yet another power adapter (both "just in case" and because I believe mine are starting to wear out) and a new keyboard, and it also needs some software maintenance, and a new MicroSD storage device.

If performance isn't that important anyway, why not just add some upgrades to the T400 (such as a USB 3.0 card for its ExpressCard slot, a new wireless chip, a slightly faster processor, and more memory) and do some other maintenance to it such as getting a new keyboard, battery, and display bezel. The problem with that is that after a certain amount of time, the performance difference is so marginal it will not make a difference. One of the T400's problems isn't that the processor is slow but that the graphics chip is so old it's close to being unsupported by World of Warcraft (the game I play most often, and one of the less intensive games I play), just as an example. Even though the game still plays, it's pretty close to the lowest settings, and I'm reminded each time I use it that the drivers are old.

The other challenge in upgrading a machine so long after its prime is that not all of these upgrades are very cheap. For example, DDR3 memory still costs as much as it does for a new machine, and a new CPU won't change the fact that the soldered graphics processor doesn't get new drivers or improve the speed of disk writes.

The other question about the T400 in particular is whether or not a new battery is worthwhile. It's currently using a battery I bought for it in 2011, and I can get almost three hours of use out of it, a far cry from the fourteen hours I could get when it was brand new and running software that worked with its power management functionality. Part of the challenge on the battery front there is whether or not it's worth maintaining parts you aren't necessarily using on a system just for the sake of having them be maintained. The advent of the Surface RT means I have a highly portable computer with really good battery life that I can also use as a tablet.

But, how long should a tablet last? A year after the Surface RT was introduced, its successor, the Surface 2 was introduced, and it's a much faster device than the original RT. (Apple has done this every year with the iPad as well, but the iPad wasn't judged as unusably slow either when it was new or immediately upon the release of the iPad 2.)

The answers within the Google and Apple ecosystem for tablets aren't that much clearer. Because most of the devices are a lot more difficult to physically wear out if you're not specifically damaging them, the main answer comes from software updates. The original iPad lost new OS updates after just two years, but the iPad 2 still gets new updates after three years. There's no word yet on how long Windows RT devices will get totally free upgrades to the newest version of Windows that run on them, and whether or not Microsoft will keep adding new applications to the system (such as when Microsoft Outlook was added upon the release of Windows RT 8.1.)

Part of the problem with talking about physical longevity is that some tasks will become more demanding than average systems well before those systems are no longer physically practical. This is part of what's happening with my old Core2Duo systems such as superslab and copenhagen (a Dell XPS M1730.) Even if I didn't have "gaming" as a workload, other users of ThinkPad T400 systems have noted that they're starting to become long in the tooth for web workloads (not that I'd notice) and of course, with a maximum of 8 gigabytes of memory and with the targets of what I'd like to virtualize changing, it's not as though virtualization is practical on that system.

Virtualization is a special case because you can't just use new hardware, you also need fairly beefy hardware, ideally with multiple disks and a very large amount of memory. Virtualization and games are the main performance-sensitive tasks that make replacement of superslab make a lot of sense. (They're also the main things that put desktops on my radar.) Both AMD's and Intel's mainstream desktop platforms at the moment can handle 32 gigabytes of memory, and Intel's high end desktop platform ("enthusiast") is made primarily of systems that will accommodate 64 gigabytes of memory.

I think more than physical reliability (because let's be honest, for all its old-age quirks, my ThinkPad is perfectly usable, and I would argue my original iPad is in excellent condition) the big issues for old computing hardware are practicality and security. Virtualization is possible on the T400, but it's definitely not practical – I need a newer and bigger computer for that task. Fortunately, my T400 is running all the most recent versions of its OS (Windows 8.1) and applications (games, Office 2013, and Lightroom 5) so I'm on the up-and-up in terms of Security. My original iPad, however, is running iOS 5, and I personally don't really consider it to be "secure." In addition, even though my applications work on it, the versions of those applications needed for some functionality (such as iCloud syncing) don't work.

Even so (and ignoring budget issues for a moment) it has been challenging to determine a point at which a machine should be replaced. Even if it doesn't "strictly need" replacing. I am actually wondering if we're getting to the point where I personally can't tell when something is over the hill. I can tell that a machine such as vorlaeufer, my ThinkPad T42p, is definitely over the hill, both in terms of the physical repairs needed to make it really useful again and in terms of system performance running applications.

My biggest challenge detecting whether or not superslab is over the hill is probably related to the fact that I'm so highly personally involved with it. I bought it new in 2009 after a lot of thought about the different machines on the market at the time and research, both in terms of reading reviews and in terms of looking at them when I happened to go to CES in 2009. Over the years I've put extensive thought into the topic of whether or not this particular machine is "too old" or over the hill, and haven't come up with any conclusive results yet. The one time I thought I was actually going to replace it; I ended up hating its replacement and reinstating it after some maintenance.

It's not as difficult for me to tell when another machine, such as illiana (an XPS M1530), a friend's previous machine from about the same time-frame, is over the hill, in part because that machine had physical reliability issues. In addition, I have no problems identifying physically reliable but slow machines (such as the XPS M1730) that may have problems soon (such as with its dual GeFORCE 8800mGTX graphics chips) and existing issues that will be difficult or impractical to repair, such as loose hinges.

Issues like this are part of why I don't include used machines when attempting to prognosticate my replacement cycle, even if those used machines become important parts of my daily workflow (such as the ThinkPad X31 becoming nachibes or the XPS M1730 becoming copenhagen.) One of the things I've done to both integrate used machines and keep track of what I'm actually using is start an inventory of my computers' roles. I update it every now and again to reflect what I'm doing at the time. For example, you'll note that eisbrecher is idle even though it's one of the most powerful endpoint computers I own. I don't know if this list will always stay in this particular form factor, but it has been helpful for me to keep an inventory of my machines and store information about what is happening to them or what they're being used for.

When buying a new machine, there's a certain amount of thought that has to go into predicting what a system will be used for over the course of its life and how that task will change. Part of why I have yet to buy a new desktop, even though I've been thinking about it near constantly for years, is that I have been having a difficult time determining a priority for the new system. (Let alone my personal priorities for whether a new portable computer or a new desktop is more important.)

The primary example is that I haven't decided if the priority is to create a space or to overbuy resources and keep the same machine running for a very long time. With each of those two things as potential priorities, it's challenging to make the determination of what the machine should have. Some of it's based on budget, but because I lack the ability to set any deadline other than "I'd like to order it tomorrow" and "whenever I can afford it," I often don't have a solid idea of what the budget for a machine purchase should be.

The purchase of superslab is likely my best executed computer purchase to date. I had a particular need in mind, a specific budget, particular form factor goals I wanted to meet (weight, screen size, etc) and some particular goals in mind in terms of lifespan and the performance of some applications. I used those goals to pick out a few different systems (of course, with laptops you get the advantage that you're mainly looking at pre-built OEM systems) and then pick from those systems based on more information I'd generated early in the process. After that, superslab became my most beloved computer ever.

It'll be interesting to look back on the decision process for "computer upgrade 4" as you might call it, because I have a very challenging time determining what the priorities are.

I suppose it's an enviable position. Despite showing signs of wear and tear, everything I'm using works and does so acceptably quick. There are just a few more tasks I'd like to do, although I've done without them for so long that justifying them to myself is a challenge. Even if computing hardware is so efficient today that a significant powerhouse of a machine with a lot of hardware installed can sit idle and use very little energy, for example, I worry (without need) about the power consumption of something that even has that hardware.