NETWORK COMPUTERS AND INTERNET APPLIANCES.
The advent of commercial uses of the Internet and related technologies in the mid-1990s triggered a movement by corporate buyers away from conventional PCs and toward so-called network computers (NCs) and other specialized network appliances. In essence, these devices were cheaper, simpler computers that were optimized for network-based tasks such as Web browsing, e-mail, and so forth. The theory was that corporations were wasting a great deal of money and time by buying and maintaining full-featured PCs when some employees may only needed the limited services of an Internet browser and a word processor. Also known as thin clients, these scaled back computers relied on a network to supply them with most of their computing power, centralizing maintenance tasks such as software upgrades and reducing users' abilities to inadvertently corrupt software on their computers. Although estimates varied widely, some experts believed that companies could save up to 80 percent of the cost of buying and maintaining a new computer by installing an NC instead. Many corporate buyers have viewed NCs not as a replacement for their PCs, but as a supplement to them. In some cases, NCs have been bought to supplant older, terminal-based technology rather than PCs.
While some of the initial product offerings in this category received only a tepid welcome from business customers, by the late 1990s NCs appeared to be taking firm root in some markets. Although the overall NC market was estimated at only a few hundred thousand units as of 1998, some forecasters called for volume to rise to more than 2 million units by 2002. This was still only a small fraction of the projected shipments of PCs by that time, which were expected to be well over 100 million units.
SPEED AND PERFORMANCE
Computers are traditionally categorized by size and power; however, rapid advancements in speed and processing capabilities at relatively low costs have blurred many of these categories. The major determinants of computing power include (1) clock speed, (2) the amount of main memory (random access memory or RAM), (3) the size and architecture of the processor's memory caches, (4) the number and capacity of pipelines to feed data, (5) the number of processors the system has (for higher power systems), and (6) other features of the processor's design and its connection to the motherboard. In fact, the chip's connection has been one of the biggest impediments to speed, as the materials used often don't transmit data nearly as fast as the chip itself. No one of these factors guarantees a faster or more powerful computer, though. For instance, computer marketing has commonly focused on clock speed, measured in megahertz, as an indicator of overall speed relative to other computers. While clock speed is important, if other aspects of the processor's architecture are less efficient, there may be no advantage to having a computer with a higher clock speed. Instead, what matters is how all the processor components work in tandem—even modifying how instructions are sent to the processor from software applications can influence speed (e.g., RISC versus CISC instruction architectures).
Computer performance was once regularly gauged by how many instructions the processor could handle per second, measured in million instructions per second (MIPS), and how many floating-point operations it could perform in a second (FLOPS), measured in megaflops or gigaflops. By the late 1980s, these were increasingly viewed as poor predictors of a computer's actual performance. The response has been to develop new measurements that better summarize a processor's abilities in the context of supporting components. One of the most widely known is a battery of performance tests developed by the Standard Performance Evaluation Corporation (SPEC), a consortium of most of the world's top computer manufacturers and related companies. The SPEC benchmarks are updated periodically to address current trends, and the results of SPEC tests on many computer models are made public.
No comments:
Post a Comment