steve@edm.UUCP (Stephen Samuel) (06/16/88)
> Was there any mention of the contents of the test file? That can make a > lot of difference. I would like to assume they used several different > files with different contents and averaged the results, but it doesn't sound > like it. ARE YOU KIDDING ME??? They used RANDOM DATA - not exactly what I'd like to call reasonably representative data. Once again, a case of BYTE doing benchmarks under seriously artificial conditions to the point of being almost meaningless for large groups of users. In this case, I think that there is a consensus among trailblazer users that the numbers they gave are quite a bit off from normal usage. Similarly: I have yet to see data for the '386 running in '286 or '086 mode from them even though this is the state that most MS-DOS / OS-2 users will are stuck with (easily half of the 386 market). The least they could have done on the modem test is to also try transmitting some 'real' data and give some compression (speed) ratios so people could see where that would make a difference. Similarly: I don't agree with their decision to run all the modems at 9600 baud. Since most users are going to try to obtain the best speed possible from their modems, it is unreasonable to artificially limit the faster machines to the slower baud rates. There isn't even anything to indicate that they used the same data for each modem (a random data generator was used, so there may have been different data for each modem). When you consider that an old BYTE benchmark article pointed out that benchmarks should try to provide as good an emulation of 'normal' usage as possible, their blunders are almost inexcusable. BYTE really has to take a look at their benchmark philosophy. I would strongly recommend that they read some of their own (~1980) articles on precisely that subject. -- ------------- Stephen Samuel Disclaimer: You betcha! {ihnp4,ubc-vision,seismo!mnetor,vax135}!alberta!edm!steve BITNET: USERZXCV@UOFAMTS
Paul_L_Schauble@cup.portal.com (06/18/88)
Steve Samuel, in a recent posting, took Byte Magazine to task for using "unrealistic" random data in their recent modem review. Granted, this is primarily a Unix newsgroup, so perhaps I have the wrong perspective. In my usual universe, most files are compressed by arc, pkarc, or compress before being transmitted. I suspect that random data is a very good match for a compressed file. Without returning any of Steve's snide comments, this seems like a good test to me. I am a bit disappointed that they didn't also test text. Paul