r_harveySelecting | Legacy | Installation | Sealed boxes | About | Home
r_harvey
Every few months, computers reach a new level of performance, and a few months later, there's a new release of our favorite software. We could keep on top, or slip behind, or stay somewhere inbetween. I'm usually a generation or two behind, because that's what I expect others to use, and because I feel that being a hardware junkie is waste of resources (time, money, landfill).
When somebody asks me what kind of computers I use, I'm hard pressed to answer. Some parts are new; a few are eight years old.
Computers are not a fad; we don't have to buy a new one whenever advertising convinces us that it's time. They're more like cars -- we service the engine and replace worn tires, we don't buy a new car. Think in terms of installing new components when they break or become outdated.
I use desktop computers, not portables. This is mainly because of cost and expandibility. While I would prefer to use portables, they are fragile, sealed boxes, that are difficult and expensive to upgrade.
I recommend standard computer cases, whether desktop or tower, as long as they support standard (ATX) motherboards with some expansion slots. This is little different than the form factor we had in the '80s, and it still works today. Before buying a computer, turn it around and look at the back of the case, to make sure it has 4-8 slots at a right-angle to the motherboard. Avoid new AT or Baby AT motherboards, except to upgrade existing cases.
If you don't mind opening the box every now and then, I would not recommend buying a new computer every time something new is released. Often just one major component has changed enough to matter, and you'll be able to incrementally upgrade. For example, I'm using a current motherboard, but until recently I used a sound card that's nearly seven years old, because I never play games. Now that PCI-based sound cards are available, I've upgraded. The old sound card has been demoted to a computer that's used just for testing.
Those big, faceless, generic-sounding mail order companies sound like terrible places. They aren't. Their goal is to make a low-hassle sale, and turn you into a customer who will buy from them again. You and they share one goal: to build you a computer that suits your needs.
Your sales representative has heard your problems a hundred times, and knows solutions to a thousand other problems you haven't even thought about -- after all, that's what they do all day. You may spend a few dollars more than at the local discount house; probably not, though. You will buy into a family that acts like they want your business.
Don't worry about having the very latest processor. Every new generation since the introduction of the 133MHz Pentium has been an incremental improvement; much less than you would expect. For most applications, it makes little sense to spend hundreds of dollars extra to gain 10-percent greater performance -- especially when that 10-percent turns out to be 4-percent.
Pay attention to bus speed and its relationship to the core CPU speed; try to get processors that are multiples of the bus speed. For instance, on a 100MHz bus, 700, 800 and 900 are multiples, but 750 and 850 are not. The greater the disparity between bus speed and core processor speed, the lower the effective performance increase (the performance difference between 500 and 550MHz, is at best 10-percent, but 4-percent in the real-world).
When you get a new processor, remember that a 10- 20-percent increase won't make any difference; get twice the performance, or don't bother with it. Installing a new motherboard is not tough, but you will likely have to by new RAM, and reinstall the operating system -- and therefore, all of your software -- because some drivers work with just one chipset or one processor.
If you want the best performance for the price, look for a 100MHz or faster memory bus. Fast SDRAM or DDRAM memory is a good investment; RAMBus (RDRAM) is not proving to be an econimical option, but it is effective as processor speeds exceed 2GHz. If you don't like opening the case, start with more RAM than you think you need; 128MB is a fair minimum today -- even for a computer built with innexpensive components. If you use expensive graphics applications or desktop publishing programs, begin at 192MB.
Now that memory prices are falling, it's tempting to get lots. Keep in mind that Windows 95, 98 and Me, may not recognize more than 512MB, while Windows NT and 2000, as well as Linux can handle a few gigabytes. Bugs in operating systems and motherboard BIOS' result in a ceiling of 1.5GB for most computers.
Don't look for value-for-dollar. You probably don't need that much storage anyway, so spend the money on the best brand and model, not the biggest drive. If you buy much more capacity than you need now, you will likely replace the hard drive before you ever need that extra capacity.
Here's a quick way to tell which brand of hard drive to buy: go to the manufacturer's Web site and look for data about reliability and things like drive fitness tests; if you don't see it, but they do talk about how big or fast or cheap their drives are, try another company. If they say their drives will run for 500,000 hours, and leave it at that, they're lying -- nobody's hard drive will run for 57 years (and nobody will want a paltry few gigabytes in the year 2059). See Hard Drives.
If you are using a standard Pentium-compatible computer made before 1999, the motherboard may support 256MB or more of memory, but it cannot cache memory beyond 64MB, meaning that you will see little performance gain, and possibly reduced performance, by adding more than 64MB of memory. Imagine what would happen if an application or the operating system were loaded into non-cached memory.
These older computers generally cannot recognize hard drives larger than 8.4GB (the older the PC, the lower the hard drive size limit... as low as 540MB in the mid-'90s). Most computers from the late-'90s have problems with hard drives larger than 35GB.
If you are happy with your Pentium-class computer, it's economical to replace innexpensive components, such as CD-ROM drives. If your Pentium or K6-2 needs major repairs, it's probably best to retire it, and move to a new mid-range computer.
Installing most new peripherals takes several hours, regardless of what it says on the box. You'll have to uninstall the driver for the old device, turn off the computer, install the new device, then start-up again and plug & pray. Drive letters or resources will change, and some software will become confused. And, of course, you'll want to play with it for a few hours.
You can let clerks at the local appliance store install peripherals for you, but they surely have less computer experience than you, and they clearly have no interest in your needs. Do you want to leave your computer, with all of your business data, in a store for two days while they install a CD-ROM drive? You're effectively out of business while the computer is in the shop, so why not spend that time effectively -- ensuring that the repair is done your way.
You are in charge of computer configuration and reliability. Don't install unnecessary software or drivers. Do add extra internal fans. Do make sure there is plenty of air movement inside the case. Don't move the computer while it is running; this includes banging on the desk and slamming desk drawers. Don't keep a hard drive beyond the warranty period.
Over the long haul, unless you're a gadget freak, you'll find that you open the computer box every six to ten months, and each time will cost a day or two of productivity. The other option is to hotfoot it down to mega- hyper- biz- compu- city- mart every two years for a whole new computer. And when you get that new computer, your work schedule will be disrupted for two days moving data, and getting software installed and configured.
We invest 2-3 days a year in computer hardware maintenance. That's 1-percent of the work year. The computer makes us more productive the other 99-percent.
It's time to contradict everything in the previous section.
Wouldn't it be great if we could keep the computer for years, and never have to open it -- and even more importantly, never have it get out of date?
That's the goal of technologies like Plug-and-Play, USB (Universal Serial Bus), and Fire-Wire (IEEE-1394). We will be able to add devices like scanners, printers, keyboards and video cameras, without opening the box. And the box itself will be smaller, lighter, and easier to connect.
In full-swing by the turn of the century, these technologies may eliminate the need for grounding straps (or guilt about not using them). Still to be resolved is the problem of what to do when the box becomes outdated; once that's sorted-out, we'll be ready -- make that happy -- to discard the standard AT/ATX motherboard and case.
USB is heading from version 1.1 to 2.0; the process is confusing everybody. Incompatibility may be a problem -- or not. What will likely happen (it seems from this early stage...) is that USB will be accepted for low-speed devices (slower than 1MB per second), and Fire-Wire will be used for high-speed devices like cameras and data storage. It's not all about technology: historically, the main issue with Fire-Wire has been a one-dollar per unit licensing fee, which convinced keyboard- and mouse-makers to choose USB. It's likely that Microsoft will not allow USB 2.0 drivers to work in Windows 95/98/Me, to encourage users to upgrade to later operating systems; these operating systems will be designed to work best/only with new computers.
Here are some suggestions for getting ready for legacy-free computing:
Top | Notice | Home | | © Copyright 2007 R. E. Harvey, All rights reserved.