D-Link High Speed Home Network

Making Networking Experts of Us All

by Josh Walrath

Testing

            I took a very long time with this review, and was very thorough in my testing and procedures.  I took a good look at the Gig-E performance that these products bring to the table, as well as the wireless performance with 100% D-Link products, as well as 3^rd party wireless products.

            Gig-E testing was done with a computer using the MSI K8N Neo Platinum, which uses the integrated NVIDIA Gig-E controller, on one end.  On the other end I used the Gigabyte GA-8KNXP which features the Intel based Gig-E and CSA (Communication Streaming Architecture).  The strong point of both of these implementations is that neither utilizes the PCI bus for Gig-E to chipset communications.  This should increase the real world bandwidth and efficiency of the Gig-E connection.  Both computers use the Seagate 120 GB 7200 RPM SATA hard drive, which gives around 50 MB/sec average throughput.  Windows XP Professional SP 2 was used in each system. 

            We can see by this quick test that around ½ of the theoretical bandwidth can be achieved.  The file being sent is not large, and fits into main memory, so no hard drive activity would hinder overall performance.

            The second test used was taking a 728 MB file and transferring it between the two computers.  This of course will have the effect of diminishing overall performance because the hard drives are not nearly as fast as the Gig-E connection.  Three trials were timed, with an average calculated between the three. 

            The hard drive performance is definitely holding the Gig-E back, and most likely a striped RAID array would have more adequately pushed this particular setup.  The 57 MB/sec throughput that Sandra reported is probably close to the theoretical maximum of this setup.  While this is a far cry away from 125 MB/sec, it is significantly better than the theoretical 12.5 MB/sec that 100 Mb Ethernet provides. 

Wireless Testing

            This was a bit more complex to do, but the results are very interesting.  Anyone seriously considering buying a wireless setup should pay close attention here.  To get a good idea how this wireless setup works I split this into two separate tests, along with four distinct wireless settings.

            Because security is now a primary concern to any user, I utilized the 64 bit WEP (Wired Equivalent Privacy) functionality that the wireless gives.  D-Link supports up to 128 bit WEP, but I figured that would be overkill for the average home user.  If wireless were to be used in a corporate setting, then 128 bit is recommended.  I did not enable WPA or WPA-PSK authentication, but I did keep it as Shared (and not Open).

            I set the wireless to 802.11G only, which means that the access point will not support or respond to 802.11B devices.  This should theoretically give 802.11G devices a performance boost.  I used the DWL-G650 and the integrated Broadcom 802.11G unit in the eMachines M6807 laptop.

            The wireless network was then established and a 496 MB file was transferred from a computer on the wired network to the laptop on the wireless.  This process was timed, and both seconds and effective megabit per second ratings were recorded.

            The laptop was set approximately 30 feet away from the access point with two intervening walls between them.  The Broadcom reported an “Excellent” signal from this spot, while the DWL-G650 reported “Very Good”. 

            Except for the Static Turbo setting, the D-Link product is performing quite a bit slower than the Broadcom.  I think a large portion of this is due to the strength of the wireless signal.  Only when Static Turbo is enabled does the D-Link distance itself from the Broadcom.  No matter where the Broadcom was in the house, it achieved similar scores.  The Broadcom solution is not able to communicate with the DI-624 when Static Turbo is enabled, so no results were given for this setting.

Next: More Testing and Conclusion

Copyright 1999-2004 PenStar Systems, LLC.