SAN FRANCISCO (10/27/2003) - Space: the final frontier of networking. For two decades, users have linked PCs with ethernet cables, but sending data through the open spaces of offices and homes without wires is still relatively new. Last year's most widely used networking standard--802.11b--allows transfers at a maximum speed of 11 megabits per second under ideal conditions. That pales next to wired speeds, which now reach 1 gigabit per second.
However, the new wireless standard, 802.11g, promises 54-mbps speed while running in the same 2.4-GHz range as 802.11b and remaining compatible with the older hardware (unlike 802.11a, which offers 54 mbps but has a shorter range and no compatibility with 802.11b). The anticipation for 11g products was so great that vendors began releasing them months before the Institute of Electrical and Electronics Engineers ratified the 802.11g specification in June. This is our first review since IEEE certification, and it kicks off the PC World Test Center's regular evaluation of wireless kits for reviews that will appear several times per year.
The 802.11g products we saw were certainly faster than the 802.11b models in this review. And the products using proprietary technologies were even swifter--though you'll get those speeds only if all your hardware is from the same vendor.
The new speeds allow 11g networks to boldly go where no wireless has gone before. While even 11b is faster than your broadband connection to the Internet, 11g's extra throughput lets you more easily transfer large files across office and home networks and even stream video from computer to computer (though not yet flawlessly, as we explain in "Still Waiting for 11g-Rated Movies").
The new products also provide better shields against intruders, by supplementing the weak Wired Equivalent Privacy security system with Wi-Fi Protected Access--the first chunk of a developing wireless security standard called 802.11i. In September, WPA capability became a requirement for any new product seeking certification from the Wi-Fi Alliance industry group. (Already-certified 802.11b products don't have to support WPA.) Wireless products that carry the Wi-Fi logo--even ones manufactured by different companies, such as a router from one vendor and a PC Card from another--are supposed to work together.
The Wi-Fi Alliance's efforts to ensure compatibility seem to be succeeding. We mixed and matched several routers and cards, and connected reliably with all the products we tried, though not always at top speeds. For our performance tests, however, we put together wireless "kits" all of whose components--a combination router/access point, a PC Card for notebooks, and a PCI card for desktops--came from the same company. We also tested USB adapters in the two cases in which they were available (both for 11b kits); some 11g USB adapters will be out on the market by the time you read this.
One note arising from our compatibility findings: A product first and foremost had to work with its own kit. In our tests, that didn't always happen.
We also encountered some chronic problems with Wi-Fi networking--from buggy software to finicky Internet connections. For those reasons, we put a high value on kits that were easy to install, had helpful documentation, and came from companies with friendly, easy-to-reach tech support. Performance and price were still important, of course. On the basis of overall scores, we chose Netgear's 802.11g kit--consisting of the WGR614 router, the WG511 PC Card notebook card, and the WG311 PCI card--as our Best Buy.
After weeding out glitchy products, we ended up with six 802.11g kits, from Belkin Corp., Buffalo Technology (USA) Inc., D-Link Systems Inc., Linksys, Netgear Inc., and US Robotics Inc.; a combo 11g/11a kit from Linksys; a standard 11b kit from Netgear; and an 11b kit with speed enhancements from D-Link.
Our tests confirmed that 802.11g products are dramatically faster than 11b ones. As our test report indicates, the 11g kits from Belkin, Buffalo, D-Link, and Linksys posted similar throughput scores that were roughly four and a half times higher than those for Netgear's 11b kit--close to the 11g spec's theoretical five-fold advantage over 11b. Also in accordance with the specs, in our tests most of the 11g kits performed at a level similar to that of the 11a kit from Linksys.
We measured speed not only in terms of the wireless connection but in terms of moving data over a network. We could not, however, measure the effect of distance and interference--from devices such as 2.4-GHz cordless phones, microwave ovens, and other wireless networks--because interference conditions change so often. To eliminate the background radiation for the tests we did, we sealed the kits in an enclosure that insulated them from all interference.Because you will likely encounter the environmental variables we eliminated, your real-world performance with any of these kits will probably be lower. But the difference in throughput we measured among the kits still indicates how they may perform relative to each other.
In our Internet tests, we measured throughput by downloading and uploading a 50MB file. (We eliminated the variability of Internet performance by creating a faux Internet in the form of a local Linux server.) For each product, the upload and download performance was roughly the same, because there was no cap on speed. In the real world, however, DSL and cable broadband services usually offer slower maximum rates for uploads than for downloads. Home broadband connections top out around 1500 kilobits per second for downloads, less than half of what even the slowest kit in this review, the Netgear MR814 router and accompanying cards, delivered.
In addition, we measured throughput across a local area network, to simulate file transfer performance in an office. We moved the same 50MB file between two PCs, one connected to the router via wireless and the other attached via an ethernet cable. In most cases, the local-network throughput exceeded the Internet throughput--hardly surprising, since LAN traffic is not slowed down by having to go through the router and its firewall. The U.S. Robotics 11g kit posted nearly identical performance for Internet and local-network traffic, as did the D-Link 11b kit; the vendors say that the routers' use of powerful ARM9 processors is responsible for the results we obtained.
Those two kits were also the top performers in their respective wireless technology classes. For this success, the vendors credit the use of Texas Instruments Inc. wireless chips that contain some performance-enhancement technologies not incorporated in the 802.11g and 802.11b specs. The U.S. Robotics products, for example, employ a TI technology called packet aggregation to combine small data packets into one large packet, thereby reducing transmission overhead. This performance boost happens only when all products in the kit use the same TI chip set.
Netgear's 802.11g kit also performed above the average for the 11g kits. Company reps attributed this to close cooperation between Netgear and the wireless chip maker Intersil.
In our informal tests, we could easily get routers and cards from different vendors to connect, but not all pairings hit full 11g speed and some managed only 11b rates. Most of the 802.11g kits were not yet certified for 11g interoperability and may require firmware upgrades; vendors are pursuing certification for all kits.
To make clear in the future what functions of a specific wireless product are certified, the Wi-Fi Alliance has introduced a new label that presents separate check boxes for 802.11a, 802.11b, 802.11g, and WPA capabilities. If a product doesn't have the label, you can look up the product's certifications on the Wi-Fi Alliance's Web site.
Decrypting WPA security
Like 802.11g certification, WPA capability is currently at varying stages of implementation. Only four of the routers we tested had firmware support for it: Belkin's F5D7230-4, Buffalo's WBR-G54, D-Link's Xtreme G DI-624, and Linksys's Wireless-G WRT54G. Only the Buffalo and Linksys routers were Wi-Fi-certified for WPA, and only Buffalo posted directions on its Web site for enabling WPA. All the vendors are planning WPA firmware upgrades for their entire lines of 11g routers, and for some 11b products.
Especially confusing was figuring out how to enable WPA on the PCs in which we installed the vendors' wireless cards. Windows XP is the only version of Windows that has built-in wireless support. Installing a wireless product on other versions requires a driver from the manufacturer, which will also have to provide WPA upgrades. For XP, which we ran on our test systems, you must install Microsoft's Service Pack 1 and patch Q815485. All of the wireless vendors say that they are developing WPA software for Windows 98 SE, Me, and 2000 on their 11g products. WPA works only if every component in the network supports it; for backward compatibility with older devices, WPA-enabled products will retain WEP as an encryption option for some time.
With the advent of WPA, we gladly bid adieu to the 10- or 26-character (depending on encryption strength) keys we had to enter on the router and the client cards to identify the user and to encrypt all data using WEP. Some kits that we looked at--those using Belkin's F5D7230-4, Linksys's WRT54G and WRT55AG, and Netgear's MR814 and WGR614 routers--made WEP setup easier by generating keys for us based on a short pass phrase.
Though WPA looks to be merely WEP with a pass phrase, it's much more secure. You enter the pass phrase to get on the network, but once the connection is made, the router continually generates and distributes new encryption keys via TKIP (Temporal Key Integrity Protocol). Thus, WPA is far harder to crack than WEP, which uses the same encryption key over and over again. With WEP encryption, a hacker needs only a wireless card and software (freely available on the Internet) that analyzes your wireless traffic and can quickly extract the static key.
Besides measuring the performance of the wireless products, we learned a lot about how often they fail. Of the 13 kits we intended to review, only 9 made it through our testing, and even some of those passed only because we ordered replacements from the vendors (after first trying to install driver or firmware upgrades on the originals). Between the original submissions and the replacements we were able to get during our testing period, eight kits failed because at least one of their components couldn't reliably connect or consistently transfer data.
When we mentioned our difficulties to Frank Hanzlik, the Wi-Fi Alliance's managing director, he agreed that some faulty product is in circulation. "We certainly do see problems, especially with the (inexpensive) products," he said. Most of the defective kits we encountered were not Wi-Fi-certified models, but one was.
The failures we experienced don't seem unique. In our most recent Reliability and Service survey of PC World readers, many participants reported mysterious problems with their wireless kits.
One Linksys owner's story was typical of the complaints we received regarding many vendors: "The wireless functionality stopped working for some reason or another.... All of a sudden it came back again, without (my) doing anything." And an exasperated D-Link customer said, "It's difficult to avoid feeling like an unpaid beta tester. Frequent firmware/driver updates make the products seem rushed to market." In our survey, readers gave Belkin poor ratings for the reliability of its routers; other vendors scored better.
The most dramatic bug, however, afflicted several models manufactured by Netgear, including the MR814. Netgear says that its customers should upgrade the MR814's firmware to at least version 4.13; otherwise the router may send a crippling flood of requests to Internet time servers while updating its internal clock. According to Netgear, this problem doesn't affect its 11g router, the WGR614.
Setup Highs and Lows
Other readers reported problems with the setup process. For example, one U.S. Robotics customer said: "I had trouble configuring the router. I had to return it to the store because tech support...said it was a defective router. The store put it on their system, and it worked OK. I will try to configure it again."
In fact, though the U.S. Robotics Wireless Turbo kit took top honors for performance in this review, we rated it near the bottom for usability. Each component came with a thin installation pamphlet that didn't cover all of the setup possibilities. For example, the DSL account we used for testing utilizes Point-to-Point Protocol over Ethernet technology that requires a log-in to establish a connection to the Internet service provider. The U.S. Robotics pamphlet mentioned PPPoE as a possible connection type and referred the user to an appendix on the bundled documentation CD-ROM. That appendix referred to yet another part of the CD manual, which finally recommended contacting the ISP for assistance.
Still, this setup approach was better than the one for the Buffalo Technology router, which we could not set to automatically dial the PPPoE connection; we had to log in to the router and manually enable it every time. The Buffalo equipment also suffered from confusing instructions, often with awkward translations. The thin setup booklet first advised users not to perform the initial configuration of the router by logging in from a wireless connection. It then dedicated the next few pages to explaining how to log in wirelessly, before discussing how to make a wired connection and how to configure the router. On the plus side, both the U.S. Robotics and the Buffalo routers pack many advanced features, such as sophisticated firewalls and event-logging capabilities, that should appeal to power users.
In contrast, the Belkin, D-Link, Linksys, and Netgear kits provided a lot more help for non-geeks. Three routers--the Belkin F5D7230-4, the Linksys WRT54G, and the Netgear WGR614--for instance, came with setup wizards that could detect a PPPoE account and step the user through setting the router to negotiate the connection automatically. The Netgear WGR614 was the most reliable model for establishing a PPPoE connection. Other units, such as the Linksys WRT54G, required some prodding.
We appreciated the CD-based wizards from Belkin and Linksys that walked us through the setup operation. (Belkin was also the only vendor that provided a full printed manual for easy reference; the other companies supplied them on CD only.) Each of the remaining kits required us to log in to the router by typing its IP address in a Web browser (though like the others, the Belkin and Linksys routers insisted on a log-in for further adjustments after initial setup). The kits from D-Link, Netgear, and U.S. Robotics made operations easier by providing context-sensitive help menus to explain topics such as DHCP (which assigns local IP addresses to computers on the network) and filters (to restrict access to the network).
Even the best kits might leave users confused, so reliable tech support is key. We evaluated vendor support by calling and asking for help with setting up wireless encryption.
Just getting through to most of the companies was a challenge. We called in the early evening on weeknights and often spent a long time on hold. Only Buffalo Technologies answered our call immediately. A few vendors said that they had been inundated with calls during the first week of September as a result of higher-than-expected sales. However, all of the companies in our review offer toll-free support lines, and all of them except U.S. Robotics provide 24-hour, daily access. (U.S. Robotics offers support for 12 hours per day, Monday through Friday.)
Once we got through, we were pleased to find that each company's support staffers were generally polite, though their knowledge varied. The Belkin representative, for example, did a good job explaining WPA, but he also told us that it was impossible to use the router with a PPPoE DSL account. The very friendly rep from Buffalo failed to mention the need to install Windows XP Service Pack 1 and the additional patch in order to enable WPA. The Linksys representative told us that the Microsoft WPA patch was faulty. (Spokespeople for Microsoft, Linksys, and other wireless companies subsequently told us that they were not aware of any problems with the patch.)
Many of our tech support requests were routed to overseas call centers. While the reps all seemed to have a good command of English, some were difficult to understand. Many respondents to our Reliability and Service survey also reported problems understanding support personnel.
Wireless--Well Worth It
While 802.11b is good for Web surfing from the couch, 11g provides extra muscle to local networks for power home users and small offices. And WPA not only will make home networks safer but will also appeal to businesses that have shied away from wireless because of security concerns. Going wireless is still far from painless, so we advise you to stick with vendors that provide the easiest setup routines and the best tech support.
None of the kits we reviewed was perfect, but we recommend Netgear's 11g ensemble as our Best Buy. It had the second-best performance of all the kits, a relatively simple setup routine, and one of the most intuitive user interfaces. Our biggest concerns were about its lack of WPA security and full Wi-Fi certification. But only a few products had these items at the time of our review, and Netgear was working on both as we went to press.
We will keep an eye on these and other developments with a new Top 5 chart of wireless kits that will appear several times a year in the Top 100. Stay tuned.
Router Interfaces: The Slick and the Sloppy
Both Linksys kits include an intuitive disc-based setup wizard to walk you through tricky steps like setting up a PPPoE log-in. Buffalo's Web-based router interface can be difficult to navigate and sometimes literally impossible to decipher.
Lab Notes: Wireless, Wireless Everywhere
One measure of Wi-Fi's overwhelming success is that you run into it almost everywhere. Its ubiquity is good news for people who are seeking an available network, but it posed a serious challenge for our performance testing. Using a Pocket PC equipped with a wireless card and Wi-Fi-sniffing software called AirMagnet, we searched high and low (including in the basement of our building) for a suitable testing location that was free of interference from other Wi-Fi networks, as well as from the noise of common appliances such as cordless phones and microwave ovens. But in San Francisco, at least, that mission proved impossible.
Our next step was to see if we could get relatively consistent performance in spite of external interference; but we saw performance variation as high as 40 percent from the same wireless kit in the same location, depending on the day and time.
In the end, we decided that the best solution was to purchase a Ramsey Electronics STE5000 Shielded Test Enclosure to eliminate electromagnetic interference. This equipment prevented us from testing wireless performance over a distance of more than about a foot, but it did guarantee reliable results for every kit, every time we conducted the test. And our results, though obtained under ideal conditions, still indicate how well one kit performs in comparison with another.
--Elliott Kirschling and Sean Captain
Still Waiting for 11g-Rated Movies
About a year ago, PC makers began selling systems that offer TV tuning and recording functions, in addition to the PC's standard ability to play music and display photos. But in many households, the PC lives in a different room from the TV, so getting the content stored on the PC over to the TV is difficult. In response, companies such as HP and Linksys have introduced wireless devices that pull content from a PC in another room and deliver it to a television. The first generation of such products has been limited mostly to retrieving music or photos, as 802.11b networking is too slow for delivering full-motion video and 802.11a is too limited in range for a large house.
Promising the speed of 802.11a and the range of 802.11b, 802.11g networking brings users much closer to wireless video streaming, but it still can't rival the quality of a wired connection. We tested our Best Buy 802.11g kit from Netgear with the $250 Prismiq MediaPlayer--a set-top receiver with an ethernet port for wired connections and a PC Card slot for wireless adapters. (The Netgear WG511 is one of the wireless cards that Prismiq recommends for the MediaPlayer.) We connected the device to a television via an S-Video port and RCA stereo jacks.
We first attached both the MediaPlayer and a PC to the Netgear router via ethernet cables; in the video window of the Prismiq interface, we saw a perfect reproduction of our video clip, a medium-quality MPEG-4-encoded trailer from The Lord of the Rings: The Two Towers. (We would have used the actual DVD of the movie, but the Prismiq doesn't support DVD streaming; doing so over ethernet is illegal.) The quality was almost as good when we switched the Prismiq to a wireless connection (with the PC still connected by wire), though we noticed a few jumpy video frames and an occasional gurgling sound in the audio. When we also switched the PC to a wireless connection, the audio glitches increased and the video occasionally broke down into a jumble of colored blocks. The video was barely viewable when we dropped to 11b speeds by switching to the D-Link DI-714P+ router.
We certainly saw an improvement with 802.11g as opposed to 802.11b, but not enough to make us believers in wireless video streaming (even for video of less-than-DVD quality). However, video streaming may improve with the introduction in mid-2004 of 802.11e, a so-called quality-of-service wireless standard intended especially to ensure a smoother flow of data for streaming audio and video.
The extra capacity of today's 802.11g did completely eliminate the glitches we occasionally heard when streaming music to the Prismiq over 802.11b. So if you're happy with the video offerings in your living room but you'd like to pipe in music from PCs farther afield, 11g hits the right note.
Five Reasons Not to Go Wireless
Wireless makes networking simpler and less cumbersome for many people, but it's not the right choice for every PC user. If you don't mind stringing some cable and you don't care about Internet surfing from the couch, hooking up your computers in the traditional style may make more sense.
For example, accountant Chuck Gerughty, of Pacifica, California, learned the pitfalls of wireless the hard way. For years he had relied on carrying disks by hand to exchange files with the employees in his ten-person office. But when he learned that a wireless network would allow all seven PCs in his office to access files on his computer--and that it would avoid the hassle and expense of drilling holes and running cables in his leased office space--he decided to install 802.11b Wi-Fi technology.
But Gerughty soured on Wi-Fi when he noticed that opening a large, complicated tax file over the network took a good 30 seconds--an eternity during the busy tax season. "It was a nightmare," Gerughty recalls. He wound up dismantling his wireless network and paying a contractor $1250 to install a much faster wired network.
Gerughty's experience is just one example of why wireless isn't always the best choice for networking a small office or home. Here are five reasons to stay cabled.
Usability: As reader feedback and our own testing confirms, setting up Wi-Fi can be tricky and annoying.
Price: Most PCs these days come with built-in ethernet ports, and if you don't need to rip out walls, running cable is cheaper.
Security: The new Wi-Fi Protected Access security is better than the original Wired Equivalent Privacy, but WPA availability remains spotty, especially for PCs not running Windows XP. Even an unencrypted wired connection has a measure of security in that a would-be hacker has to gain physical access to the network.
Consistency: Interference can slow down or even kill a Wi-Fi signal, especially in areas where several Wi-Fi networks overlap. This is less of a problem for 802.11a products than for 802.11b and 802.11g networks, which share the 2.4-GHz portion of the spectrum with other devices such as Bluetooth products, cordless phones, and microwave ovens.
Bandwidth: Though more than adequate for browsing the Internet on a shared DSL connection or for occasional file transfers, wireless throughput may not suffice for people who, like Gerughty, must frequently move large files between PCs. There's a reason that gigabit ethernet is gaining popularity--it's up to 90 times faster than an 802.11g connection.