Hardware Reviews - Ethernet

How we test

Updated 11/02

For our Ethernet reviews, we currently use three main tests. 

For evaluations and reviews of Ethernet equipment, we use our two test machines as detailed below. We use a clean install of the Windows XP Home on our "advanced desktop machine" and Windows 98 Second Edition on our "basic desktop machine" and laptop. We install the operating system clean for every new test and have no other software or peripherals installed other than what you see listed. We follow the manufacturer's install instructions for network interface cards and use any drivers provided with the equipment. 

For evaluations and reviews of Ethernet network interface cards, we install the test card on the advanced machine and use a D-Link DFE-530TX+ 10/100 adapter card in our basic machine because we found it to be among the most stable cards that we've tested so far. We also use the built-in network adapter in the test laptop for some tests. All of the test machines are connected to the switch or router being tested, or in the case of testing network cards, are connected to an Allied Telesyn ATFS708, 8-port 10/100 Fast Ethernet switch.

For tests of cards, switches, routers and kits, we set up the software as follows: We install, in order, TCP/IP and then IPX. We remove the others so only one protocol is installed at a time. With all protocols, we leave the default settings as is.

For the actual tests, we currently use three main procedures.

First, we use a 60-meg file and copy it from one machine to the other. For lack of a better term we call each test a send or receive. This provides a good measure on the "real-world" speed of the network cards, hub or switch being tested. 

To more fully test the bandwidth of a hub or a switch, we use the same file along with a 30 meg file and copy them back and forth between both machines at the SAME time. We run the same test twice, average the results and report the result in seconds to complete all file copying. This test puts more demands on the network hardware and helps reveal the performance gains from higher-bandwidth equipment such as switches.

We also use two subjective tests on Ethernet equipment. We watch an AVI movie file that is transferred across the network equipment and judge how smooth the video and audio is, and report any problems. We also play a network game across the network and report on how smooth the game play is and if we've discovered any problems while running the games.

Our test machines -- after 11/02

Test Machine 1 (or Advanced Test Machine)  -- also known as T1800

Test Machine 2 (or Basic Test Machine)  -- also known as T1000

Test Laptop -- "L1000"

Our test machines -- after 11/00

We have two machines set up to help us install, set up and test hardware and software products of interest to home PC networkers. Here's a quick review:

Test Machine One - "T866"

Test Machine Two - "T400"

********

Updated 6/00 - For Reviews prior to 11/00

For our Ethernet reviews, we currently use three main tests. We will try our best to keep these as consistent and standard as possible, but expect that we'll have to make changes based on industry changes every 6 months or year.

For evaluations and reviews of Ethernet kits, we use our two test machines as detailed below. We use a clean install of the original Windows 98. We install the operating system clean for every new test and have no other software or peripherals installed other than what you see listed. We follow the manufacturer's install instructions for both network interface cards and use the drivers provided with the kit. Obviously, we connect both interface cards to the hub or switch provided with the kit.

For evaluations and reviews of Ethernet network interface cards, we install the test card on our Celeron 400 (known as T400) machine. The other machine (a Pentium II-266, or T266) has a D-Link DFE-530TX+ 10/100 Adapter card because we found it to be among the most stable that we've tested so far. This card and the card being tested on T400 are connected to a Netgear FS308 10/100 switch.

For both kits and cards, we set up the software as follows: We install, in order, TCP/IP, then IPX, then NetBEUI. We remove the others so only one protocol is installed at a time. With all protocols, we leave the default settings as is.

For the actual tests, we currently use three main procedures.

First, we use a 30-meg file and copy it from one machine to the other. For lack of a better term we call each test a read or a write. Since all tests are conducted from the Celeron 400, a read is copying the 30-meg file from the P2-266 machine to Celeron 400. The write is copying the file from the Celeron 400 to the P2 machine. This provides good measure on the "real-world" speed of the network cards and the hub or switch included with a kit.

To more fully test the bandwidth of a hub or a switch, we use the same file along with a 13 meg file and copy them back and forth between 4 machines at the same time. Specifically, we use four computers connected to the switch or hub. We'll call them Test1, or T1, T2, T3 and T4. From T1, we copy a 30 meg file to T2. From T2, we copy a 13 meg file to T1. From T3, we copy a 30 meg file to T4. And from T4, we copy a 13 meg file to T3. In short, two 30 meg files and two 13 meg files are transferred across the switch at the same time. We run the same test twice, average the results and report the result in seconds to complete all file copying. This test puts more demands on the network hardware and helps reveal the performance gains from higher-bandwidth equipment such as switches.

Finally, to ensure even more objectivity to the test, we use Qcheck, a freeware networking testing utility, that uses software "endpoints" on both test computers and a small utility to actually run the tests. For all tests, the utility sends data packets back and forth between the two computers. The utility tests true network performance and tries to avoid complications from the actual performance issues of the computers. For a response time test, Qcheck returns the minimum, maximum and average number of seconds it took to complete a transaction. For our tests, the transaction we use is sending a 32,000 bytes back and forth 10 times. We report the average response time in milliseconds reported by Qcheck. We run the test using both IPX and TCP/IP protocols.

For a throughput test, Qcheck returns the amount of data per second that was successfully sent between the two endpoints. For our test, we send 1 megabyte across the network.

For a streaming test, Qcheck returns the rate that streaming data was received by the second endpoint, the amount of packet loss that occurred and the CPU utilization for both endpoint computers. In short, the test reveals whether your computers and network connections are "capable" of streaming video performance. For our tests, we ran a 1Mbps test for 30 seconds.

Our test machines -- prior to 11/00

We have two machines set up to help us install, set up and test hardware and software products of interest to home PC networkers. Here's a quick review:

Test Machine One - "T400"

Test Machine Two - "T266"

Privacy statement

All content © 1998-2007 by homePCnetwork, inc.