Summary
This is the fourth version of this brief report, addressing the differences in browser file download speeds between WindowsXP browsers and MacOS browsers. This report includes the recently updated Apple Safari 1.2. Additional tests using a GbE (Gigabit Ethernet) network Xserve server and workstations connected via GbE is being worked on in the Netherlands. Results are not yet available.
Background
These tests were originally conceived to answer questions about performance differences and potential productivity losses associated with various WindowsXP and MacOS browser configurations, operating in Intranet systems on 100MBps Switched Ethernet (typically 10,750KBps) networks and accessing the Internet via high-speed connections ranging from 3Mbps (384KBps) to 54Mbps (6912KBps) in speed. Planned network upgrades to GbE to improve Intranet productivity also raise the question of browser performance in a GbE Intranet environment. Are these browser already maxed-out running on 100Mbps networks?
Browser Tests
Observed browser download speeds using 64MB binary and text format test files, downloaded from my local AppleShare IP Server v6.3.3 http and ftp services via direct connection 100Mbps CAT5 Switched Ethernet network from 500MHz PowerMac G3. Server supports minimum 84Mbps (10.752KB) sustained read/write transfer data rate. Highest speed reported by each browser connected to the the server is noted below.
Windows PC configurations:
WindowsXP SP1, Compaq Presario 6000 1.6GHz Athlon, 768MB RAM
WindowsXP SP1, Compaq Presario 5400NX, 3GHz Pentium 4, 1GB RAM
PC Browser reported KBps
and version http ftp
IE 6.0.2 - 10854 366
Firebird 0.7 - 6390 3077
NetScape 7.1 - 5780 1808
PowerMac configuration:
B&W G3 Rev.2, PowerLogix 800MHz CPU, 1GB RAM, builtin Ethernet and ZYNX 100Mbps
PCI Ethernet used.
Mac Browser reported KBps
and version http ftp
Safari 1.2 - 8806 0 (didn't work, Error -50)
Safari 1.1 - 7782 0 (didn't work)
IE 5.1.7 - 1454 542
iCab 2.9.7 - 4128 578 (OS 9.2.2)
iCab 2.9.7 - 2201 601 (Classic)
iCab 2.9.7x - 1619 522 (OS X)
NetScape 4.8 - 2549 614
NetScape 7.0.2 - 3333 541
MSIE 5.2.3 - 0 0 (see attachment)
Comments
- Opera 7.2.1 browser for Windows reported large variations in download speeds, as multiple download tests were performed. Data is suspect and not included here.
- NetScape 4.8 (Communicator) was used since it is the only Mac OS9 browser that can still support many Java based web pages.
- Browser reported KBps numbers are rounded up.
- Windows TCP/IP was optimized with DrTCP, to enlarge RWIN and turn on Selective Ack features, both resulting in large speed increase. Mac TCP/IP was not tweaked prior to these tests. Have to find OS X tool to do this.
- iCab (very popular with my clients) was run in multiple Mac OS configurations to see what effects they had on performance. Effects were significant.
- Safari 1.1 did not work with my ftp server, reporting that it couldn't find the URL. Safari 1.2 reports Error -50 and offered Retry does not work, just repeats Error -50.
- IE 5.2.3 Download Manager file download progress info chokes after reaching 6.0 MB/sec and starts reporting a negative number string. The download process slows down and eventually completes, but total time is excessive, resulting in a long period to download the test files.
- Tests using the 3GHz Pentium did not show file transfer speed improvements, compared to the 1.6GHz Athlon unit, except for many online speed test web pages involving Java Applets, which the 3GHz Pentium 4 ran significantly faster! The resultant numbers are considered bogus and grossly inaccurate, suggesting errors in Java applet designs... way too many of them!
About The Screen Captures (not included in this version of report)
Screen capture were executed before the browser completes the download process (with the sole exception of MS IE for Windows OS) because they do not show the total KBps figure for the download. There is no way to see this important data, other than to use a calculator on the total elapsed time, versus file size. Huh? Not reporting final KBps is a serious omission, in my opinion and experience.
Matisse Enzer's bit-byte converter at
http://www.matisse.net/mcgi-bin/bits.cgi? was quite helpful.
Regards,
-Rod
__________________________________________________________
Rod Paine: Retired desktop computer systems consultant
and systems management guru, who keeps his eye on the
technology and stays in touch with key industry friends,
associates and a few past clients of the ASTEC Company,
where he was President from 1988 to 2003.
|
