[ Previous Article | Next Article | Index of Articles]
This article first appeared in the June 1999 issue of the Louisville Computer News. It was written by Lee Larson.
Former British prime minister Benjamin Disraeli once said "There are three types of lies: lies, damed lies, and statistics." I was reminded of this quotation many times last month as I reread Stephen Jay Gould's excellent book The Mismeasure of Man, in which he shows the impossibility of using a single number such as I.Q. to measure intelligence. Intelligence has many facets, and thinking we can measure it with one number is just as silly as thinking we can tell everything about Louisville's climate by knowing the average annual rainfall is 44 inches.
We have the same problem with computers. The convenient number used by computer sellers is CPU speed. The implicit message is that a 450 MHz box is always better than a 350 MHz box. This one number simplifies advertising, but it's just as ridiculous as if car dealers only used engine R.P.M. in their advertisements and ignored other features such as price and gas mileage.
As do cars, a computer has many subsystems contributing to the whole, and the proper mix determines which machine is best for a particular job. A kick-butt game machine will have fast graphics and memory, while a file server needs no fancy graphics, but must have huge storage and powerful networking. The game machine is like a sports car and the file server is like a truck; neither can do the other's job very well.
Apple has long been saying their PowerPC machines blow the doors off comparable Pentium class machines, and they cite benchmarks to prove it. The most recent ones they've latched onto are the BYTEmark benchmarking numbers, derived from a series of speed tests written for BYTE magazine.
In an effort to be somewhat more accurate, the BYTEmark tests are a series of programs designed to test different aspects of a computer's speed. The results are a set of several numbers, instead of just one. Running the BYTEmark programs, a 433 MHz G3 Macintosh is about twice as fast as a 500 MHz Pentium II at floating point math and about 12% faster at integer math.
The problem with these BYTEmark numbers is that they tend to check processors in very particular ways, and if the test is changed, the same processors might look very different. For example, running the RC5-64 code cracking software (www.distributed.net) on a 500 MHz Pentium II yields about 1.6 million keys checked every second and on the 433 MHz G3 about 1.3 million. The program is highly optimised for both processors, but some obscure details about how the chips handle this particular type of internal data make the Pentium II slightly superior, even after adjusting for clock speed. (If you really want to know, the Intel chips have more versatile hard-coded variable bitwise rotate instructions.)
Many people say the best way to benchmark a computer is to see how well it runs the software you actually use. After all, do you really care how well your machine can run a game like Angry Nuns with Guns, if you spend all your time in Photoshop?
Since there are many sites on the Internet giving shootouts between different computers running Adobe Photoshop, it would appear to be a simple job to pick the best computer. But, a reason to mistrust such sites is that the benchmarker might have an agenda. Just as with political polls, asking questions in just the right way can shift the results in a direction the questioner wants.
Photoshop is a very complicated program, and there are hundreds of possible speed tests. A proper mix of tests can probably favor any high end machine. Different testers have found the DIGITAL Alpha processor, Intel Pentium II, Silicon Graphics and the Apple G3 to be fastest.
Another example of the issue of who is doing the benchmarking has recently been bringing the Linux community to a boil. On April 13, with great fanfare, Mindcraft released a major benchmarking study comparing Linux to Microsoft Windows NT as a server operating system. According to Mindcraft "Windows NT Server provides over three and a half times the performance of Linux as a Web server...and as a file server, Windows NT Server provides over two and a half times the performance of Linux." This shocked people who were experienced with both operating systems.
After the study hit the 'Net, Mindcraft revealed that the testing was paid for by Microsoft. Not only that, but Mindcraft made extensive use of Microsoft expertise to fine tune the NT server, while doing little to optimise the Linux box. Linux advocates have cried "foul," and after a major firestorm on the 'Net, it appears other, more independent tests will be forthcoming.
So, knowing all this, what conclusion can be drawn? My conclusion is that for almost everybody it really doesn't matter. If any modern computer had thumbs, they'd be twiddling 99.9% of the time. The wait between keypresses is an eternity for a 300 MHz processor, and speedy Web browsing is helped more by a speedier connection than a faster computer. For e-mail, browsing and typical business use, most benchmarks are meaningless.
Of course, just as with cars, there is always a person whose sole aim is to be first with the biggest, fastest, and most expensive toy.
Miscellaneous
For the last several months, Apple has been placing beta test versions of QuickTime 4.0 on the Internet. Now, the final version is available for free. QuickTime is all-encompassing cross-platform media playback and authoring software, supporting an enormous variety of data formats including MPEG, Windows AVI, Photoshop images, PNG, and FlashPix, along with a slew of compression and transport technologies.
QuickTime 4.0's most prominent new feature is the ability to handle streaming media such as live broadcasts or pre-recorded audio and video. Apple is seeking to extend QuickTime's dominance in digital media production to real-time Internet-based content, competing against RealNetworks' well established RealPlayer and Microsoft's Media Player. QuickTime 4.0's most visible new features, however, revolve around radically redesigned playback and application interfaces designed to look like consumer electronic devices. Apple's QuickTime 4.0 announcement claims the new controls are "stunning" and "intuitive." However, the revised look and feel has drawn consistent criticism from some who feel the new look has more style than substance.
QuickTime 4.0 is available for free via the Internet for both Mac OS and Windows. On the Mac OS, QuickTime 4.0 requires a 68020 or better processor running System 7.1 or higher, although some features require a PowerPC processor. The installer itself is a scant 380K, but then you must choose which QuickTime components you wish to install, with typical selections ranging from 6 to 15 MB of downloaded data.
QuickTime Pro 4.0, which adds authoring, editing, export, and enhanced playback capabilities, is an additional $30. Owners of QuickTime Pro 3, can upgrade to QuickTime Pro 4.0 for free.
Here it is another month and there's yet another upgrade to Netscape Communicator. Version 4.6.1 is available as a hefty download. This release is a minor bug fix to version 4.6 which offers an updated AOL Instant Messenger 2.0, fixes to improve security, and has new 56-bit DES ciphers added to both export and US versions. This version also does not include G2 player, although the release notes say otherwise, and does not include the QuickTime plug-in. A 128-bit secure version is also now available for eligible US users
For the last year or so, some of us have been waiting impatiently for the SETI@home project to finally get under way (setiathome.ssl.berkeley.edu). SETI@home is an Internet-based distributed computing project to analyze signals from radio telescopes for signs of intelligent extraterrestrial life. In late May it started off with client software for Macintosh, Windows and several different flavors of Unix.
The client software contacts the SETI@home servers and is given a block of data to analyze. It performs complicated signal analysis on the data, sends its results back to the server and gets another block of data to chew on. The analysis of a single block takes from 12 to 40 hours, depending on the speed of your machine and how much other work it's doing.
The first Macintosh client is not without some problems. Although it's designed to work as either a screen saver or a regular application, most people find it's a lot more efficient to use it as a regular application, even though it requires more memory in that mode. It runs much slower when the Mac is set to millions of colors rather than thousands of colors and it doesn't know about multiple monitors. There are some limitations when it is used behind a firewall. All these problems will be fixed in the next version.
Nobody anticipated the success of the project. In less than a month, more than 600,000 people downloaded the client. This phonomenal interest has caused quite a few problems. The SETI@home servers have been overwhelmed and are sometimes nearly unreachable. Most annoyingly, the servers ran out of new data early on and began serving the same blocks out over and over again.
All the server problems will soon be over, thanks to a donation of equipment from Sun Microsystems. The astronomers assure everybody there are more than enough data to go around.
Louisville Computer Society
The topic for the July 27 meeting of the Louisville Computer Society had not been announced as this was being written.
The Louisville Computer Society meets from 7:00-9:00 P.M. at Pitt Academy, 4605 Poplar Level Road, at the intersection of Poplar Level Road and Gilmore Lane. Everyone is welcome to attend. For more information, on the web go to www.aye.net/~lcs, or e-mail lcs@aye.net.
[ Previous Article | Next Article | Index of Articles]
/home2/lee/www/cgi-bin/textcounterdata/ [TextCounter Fatal Error: Could Not Write to File __lee_macwritings_LCN9907_shtml]