As much as I love new technology, for me, it’s all about performance. I’ve tested a number of 802.11 draft "n" products, and while they are significantly faster than 802.11g products, all of my computers in my office/lab are connected to my network via wired Ethernet either at 100Mbps or 1000Mbps. It’s just faster, and, frankly more reliable.
Last winter, I tested one of the first "Cable Free" USB products for another publication, and frankly, was disappointed in the throughput. However, the new IOGEAR product is based on a completely different chipset and UWB implementation. So, I was hoping that its performance would be dramatically better. I found that CWUSB is better, but not by a lot.
It really should come as no surprise that CWUSB products are going to be significantly slower than their wired counterparts. In his review of the Belkin Cable Free USB Hub, Tim Higgins uncovered a big reason why "hub and dongle" devices are so slow. In fact, the Belkin reviewer’s guide included a "Pyramid of Wireless USB throughput" diagram that showed that HWAs (host wired adapter – (hub) and DWAs (device wired adapter – (dongle)) are at the bottom of the performance pyramid. Belkin’s notes said that hub and dongle type products use additional communication protocol layers in order to deliver wireless capabilities to the wired USB host and device.
Indeed, Tim’s summary of the performance pyramid (below) was borne out by my test results on the IOGEAR Wireless USB product.
"The Belkin product falls into the bottom "HWA > DWA" area of the pyramid that represents a connection that has the UWB radio connecting to the CPU via a USB 2.0 connection on each end of the connection. You can see that the additional interface slows things down considerably to a 30 - 45 Mbps range—over a factor of 10 reduction from the advertised number. Note that this will be the performance range of any of the UWB/USB "hub+dongle" products that are expected out over the next year—Wireless USB flavor included."
One of my most frequent uses of USB is to copy files from my camera’s memory card to my computer. On a typical day of shooting, I generate about 1GB of data. For this test, I connected a SanDisk Image Mate 12-in-1 card reader to the hub. I then copied 921 MB of data both to and from a Kingston Ultimate 2GB 120X SD card. I ran each test three times and recorded the results.
I then connected the SanDisk Image Mate directly to a USB 2.0 port on the same notebook and re-ran the tests. Since I copied the same files to/from the same card using the same notebook, the only difference in the test was the wireless link.
I realize that SD memory isn’t the fastest memory, so I repeated the same tests described above, but I replaced the card reader/SD card with a USB 2.0 external drive. I used the same set of files for the file copies. Summarized below are the results.
|USB Hard drive||Kingston Ultimate SD|
|Speeds in MB/s||Read||Write||Read||Write|
|% of Wired||9.3%||10.1%||23.5%||23.8%|
Table 1: Read / Write performance: CWUSB vs. USB 2.0
From these results, you can see that the throughput on the IOGEAR wireless link was consistently about 2.25 Mbytes/sec. Whether the test was reading/writing to a USB hard drive or the SD card, the throughput was approximately the same, indicating that the link speed, not the memory device, was the bottleneck.
Looking at the "wired" speed tests, you can also see that the USB hard drive was significantly faster—more than twice as fast—as the SD memory. What’s important to note, however, is the percentage of wired speed. For the USB hard drive, the effective wireless throughput was only approximately 10% of what was achieved on the wired connection.
To put this in context, on average, it took 38 seconds to copy the 921 MB of data on the wired connection, and on average, 6 minutes and 46 seconds to copy using the wireless USB hub. Similarly, even when comparing to the slower SD memory, the percentage of wireless speed was just a little under 25% of the wired speed.
Perhaps some of the poor performance was, in part, attributable to the "very low" signal strength. But with the test hub and host adapter only separated by seven feet, I feel it’s a fair test. However, I retested read and write speeds with the hub and dongle separated by only three feet, and the write performance was virtually identical. The read speed tests were surprisingly, slightly slower at three feet than at seven feet, averaging 1.97 Mbytes/sec.
To test the 10 M range claim, I also moved the product out of my office into the hallway about 13 feet from the dongle, where the hub promptly disconnected. Thus, it would appear that, at least in my environment, the working distance is about 10 feet.
Finally, to see how WUSB and "Cable Free" USB compare, I pulled test results for the Kingston SD card for both products into Table 2. (The Belkin data comes from my previous testing. Note that the Belkin product has had driver updates since this data was taken.)
|IOGEAR Wireless USB||Belkin Cable Free|
|Speeds in MB/s||Read||Write||Read||Write|
|% of Wired||23.5%||23.8%||16.6%||10.8%|
Table 2: Read / Write performance: IOGEAR CWUSB vs. Belkin Cable Free USB
You can see that the IOGEAR WUSB performed better than the Belkin "Cable Free," particularly for writing. Even at its best performance, however, WUSB throughput is only slightly better than 12.5 Mbps (1.6 MB/s) USB 1.1 products.
Don’t get me wrong. I love new technology, and if it outperforms what I currently use, I’ll probably adopt it. One of the stated benefits of CWUSB is that it eliminates cable clutter from your desktop. However, in its first-generation form, you’re only eliminating the single cable that connects the hub to your computer. All of the other USB peripherals still plug into the wireless hub. If you saw my office, you’d realize that eliminating one cable is like removing a drop from a bucket of water—it wouldn’t make any difference.
Certified Wireless USB holds an exciting promise—high-speed wireless data transfer without the hassle imposed by setting up a Wi-Fi connection. However, the underlying additional communications protocols associated with CWUSB means that the technology will never achieve "wired" speeds—even with additional driver tweaking (see the "Pyramid of wireless USB Truth"). In addition, we’re still two steps away from the "native" host and device implementations that will provide the highest speeds.
So, at least for now, my USB connections are going to remain like my network connections—wired...and cheap!