eWEEK Labs: 802.11n Is a Gamble
eWEEK Labs: 802.11n Is a Gamble
The newest wireless LAN specification802.11nisnt ratified yet, but thats not stopping vendors from releasing products based on a draft version of it. Is this gear worth the gamble? Not if you expect your wireless experience to be easy, reliable or consistent, eWEEK Labs tests reveal.
The 802.11n spec promises throughput well north of 100M bpsa tempting proposition for users whose performance thirst cannot be slaked. But products based on the draft version of the spec may not be compatible with the shipping spec, which means that users could be looking at significant upgrades or even total overhauls of the draft-based gear once "real" 802.11n products are released (sometime next year).
After last months joint announcement by Atheros and Broadcom, trumpeting the interoperability of products using their respective wireless chip sets based on Draft 1.0 of the 802.11n standard, eWeek Labs decided it was time for a reality check.
We found that interoperability among products based on different manufacturers chip sets is possible out of the box, but it wont come easy or reliably. For corporations and consumers alike, our purchase recommendation for these products remains the samedontbut we do recognize the quick maturation of these wares in the brief time since they were introduced this spring.
We brought in four sets of products representing the three chip-set makers that are currently delivering gear based on the draft 802.11n standard. Netgears RangeMax Next Wireless RouterGigabit Edition (WNR854T) and RangeMax Next Wireless Notebook AdapterGigabit Edition (WN511T) are both based on Marvells TopDog technology, while Belkins N1 Wireless Router (F5D8231-4) and N1 Notebook Card (F5D8011) represent Atheros Communications XSpan chip set. Representing Broadcoms Intensi-fi chip set are Buffalo Technologys Nfiniti Wireless-N Router and Access Point (WZR-G300N) and Notebook Adapter (WLI-CB-G300N).
We also brought back for another look the Linksys Wireless-N Broadband Router (WRT300N) and Notebook Adapter (WPC300N), both of which are based on the Broadcom chip set. (See "First 802.11n Products Show Standards Promise" at eweek.com.) Since we last tested these products, Linksys updated the router firmware to Version 93.3, which improved performance slightly at longer distances in our tests.
Netgears products proved to be the unrivaled speed champs at close distancestopping 130M bps of real throughputbut performance started to lag considerably as the client moved away from the router. (See "Unencrypted throughput" chart, Page 33.) The Belkin pairing also fared poorly at longer distances, and we could not crack the 100M-bps barrier at close range.
All the Linksys and Buffalo gear based on the Broadcom chip set worked together easily at high speeds. Power users looking to mix and match draft 802.11n products at this time are advised to stick with Broadcom-based products. But users also should take care to ensure that driver and firmware revisions are of similar generation on the client and router.
This can be a herculean task because Linksys and Buffalo use different versioning schemes for their respective routers firmware. Of the products we tested here, we determined that Buffalo is using a newer client driver than Linksys is, while the Linksys router firmware is more up-to-date than that of the Buffalo Nfiniti Router.
As noted in a joint release, Broadcoms and Atheros products worked together at high speeds out of the box. (See "Interoperability matrix" chart, above.) However, we saw this compatibility only in certain pairings. Connecting the Atheros-based Belkin adapter to either of the Broadcom-based routers we tested produced solid resultsin the 70M- to 90M-bps range. While this performance is less than Broadcom and Atheros have claimed in their joint testing, we found it acceptable at this time.
Using the Broadcom-based client adapters with the Belkin router was another story, unfortunately. We saw poor perform-anceless than 20M bps at short distancesdespite the fact that the connections link rate hovered at about 300M bps.
The Marvell-based products did not interoperate at high speeds with any other products in our tests. The link rate topped out at the 802.11g-maximum 54M bps, which produced the expected 802.11g performance (in the neighborhood of 20M to 22M bps).
This leaves Netgear in the unenviable position of having two draft 802.11n product lines in its stablefrom Marvell and Broadcomthat will not yet work together at high speeds.
Customers taking the plunge with Netgears draft 802.11n gear need to ensure they purchase an adapter and router that will work well together at high speeds. Netgear draft 802.11n products with a model name ending in "T" are Marvell-based products, while a "B" in the model name stands for Broadcom.
According to David Henry, Netgears marketing director for consumer products, Netgear engineers have Marvells interoperable code in their labs and are fine-tuning it to improve throughput performance. While we could not get our hands on the beta interoperability code for these tests, we expect to see a software upgrade available on Netgears Web site within the next month.
Next Page: Testing out backward compatibility, security.
Because products based on the 802.11n draftor 802.11n itself, for that matterwill likely be used in conjunction with 802.11g devices, we tested the backward compatibility of the draft 802.11n routers and access points with WiFi-certified 802.11g-compliant products.
We tested each draft 802.11n router in conjunction with the Intel Pro/Wireless 3945ABG adapter included in our test Lenovo Group ThinkPad T60s and tested each draft 802.11n client adapter with a Cisco Aironet 1200 access point. In each case, we tested for maximum throughput, distance performance and compatibility with encryption.
When used together with 802.11g access points, draft 802.11n client adapters provide undeniably improved performance at longer distances to the legacy devices. In our tests at 95 feet, the Cisco access point with the Intel adapter could muster only about 3.3M-bps throughput. Using the various draft 802.11n adapters with the Cisco access point at the same distance, we saw three to five times better performance.
The Linksys Wireless-N Adapter tallied the lowest result at this range, at 9.5M bps, while the Belkin N1 Notebook Card pulled in more than 15M bps (with the Buffalo and Netgear cards in between).
We also saw glimpses of range improvements for 802.11g clients when using a draft 802.11n router, but we also encountered more interoperability problems: When we used the Intel adapter and the routers from Linksys, Buffalo and Netgear, we experienced unexpectedly limited bandwidth under certain circumstancesabout 6M bps when we expected about 22M bps.
We determined that this problem is due to Centrinos incompatibilities with the way Broadcom implements the standard narrow 20MHz channel that 802.11b/g products use within the wide 802.11n 40MHz channel. With both Broadcom routers, we could choose whether to locate the narrow channel high or low within the wide channel. (If we set the wide channel to 7, the narrow channel could be either 5 or 9.)
We determined that backward compatibility with the Centrino client suffered when we set the narrow channel to the high position, so our published interoperability numbers are with the narrow channel set low.
Both the Belkin and Netgear products advertise the narrow channel on the same channel number as the wide channel, so we cant explain Netgears poor performance in this scenario. Were still looking into it.
Representatives of D-Link, whose draft 802.11n products we expect to start reviewing in a few weeks, told eWeek Labs they encountered a similar problem in their own lab when using an Intel adapter with their Atheros-based router. They said they resolved the problem by upgrading to the latest Centrino drivers, but a check revealed that we were already using the latest drivers (Version 10.1.1.3, according to the Intel and Lenovo Web sites).
This channel problem may be specific to the Centrino adapter, as we could not duplicate the issue when using a D-Link DWL-650 802.11g adapter. For the time being, well say that mileage will varya lot.
More Data, More Security
Now armed with at least the potential to wirelessly transmit large chunks of data in a short time, we wanted to verify the ability of the devices under test to encrypt this data.
We secured each wireless session using WPA2 (Wi-Fi Protected Access 2) with AES (Advanced Encryption Standard) encryption. We tested each product pair at the same distances as we did in our throughput tests, but, because we were concerned primarily with high-speed results, weve published only the scores at 40 feet (see "AES-encrypted throughput" chart, Page 33).
The pair of Netgear products suffered most from encryption, falling a whopping 23 percent from their lofty unencrypted performance (but achieving a still-impressive 97.8M bps). The Belkin pair dropped about 14 percent with encryption, to 79M bps, while the Linksys scores remained about the same.
The Buffalo pairs performance actually improved slightly. However, the lack of full support for WPA2 marred Buffalos scores, so we tested Buffalos products with WPA using AES. (AES was an optional component of the original WPA spec, although Temporal Key Integrity Protocol, or TKIP, is the primary encryption algorithm.)
WPA2 is not a selectable option on the Buffalo Nfiniti Router, nor could we successfully complete a connection to a WPA2-protected network from the Nfiniti Adapter when using Buffalos Client Manager 3 wireless configuration tool. We could sometimes connect to a WPA2-protected network when we used the Windows XP supplicant (with the WPA2 patch installed), but we found this capability inconsistent from network to network, and it should not be trusted to work in the field.
At first, we could not get encryption to work at all with the Belkin N1 Wireless Router. In conversations with Belkin representatives, we learned that the N1 implementation does have a known bug: When the router is set to support both AES and TKIP, the client does not recognize that the network requires encryption and instead unsuccessfully attempts to associate unencrypted. However, this was not the problem in our tests. In the end, we needed to reinstall the router firmware to get encryption working properly.
Next Page: Frequent updates and revisions from vendors are expected during the next few months.
A Look Ahead
With the 802.11n standard in flux and lots of work still to be done to improve interoperability within the current draft standard, owners of draft 802.11n products can expect to find frequent updates and revisions of both client drivers and router firmware on their vendors Web sites during the next few months. As such, users should pay close attention to code availability, as interoperability, throughput and distance performance can vary greatly depending on the code revision on both ends of the connection.
Near the end of our test cycle, Belkin provided us with beta versions of both its client driver (Version 220.127.116.11) and router firmware (1.01.17). With the beta code, we noted slight throughput performance gains at shorter distancesa 6 percent boost, to about 105M bps at 40 feetbut significant improvement at longer distancesalmost a 175 percent performance gain, up to 31M bps. (See "Belkin shipping code vs. beta" chart, Page 36).
Interoperability results changed dramatically with the beta code as wellsome better, others worse (much worse).
With the beta firmware, the Belkin N1 Wireless Routers wireless throughput spiked to almost 120M bps when using either of the Broadcom-based adapters, but the performance was highly irregular and unreliable.
As we sat watching either adapters client monitoring software, we noted that the link-rate cycle went from 54M to 300M bps and back again, even with the laptop stationary at a short distance and with no environmental interference sources detected by our spectrum analyzer. Depending on the channel we allocated the link on, we were just as likely to get 7M bps or not be able to pass traffic at all.
With the upgraded Belkin client connected to our Linksys Wireless-N Router, we found that performance worsened, from 87.6M bps down to about 67M bpsnot good, but not nearly as bad as when we tried to connect to the Buffalo Nfiniti Router. In that case, our client machine consistently coughed up a Blue Screen of Death rife with NDIS (Network Driver Interface Specification) error warnings.
Just after we concluded testing, Linksys released an updated driver, Version 18.104.22.168. This latest version promises improved performance and upgrades the client to a WiFi-certified stateallowing Linksys to start using the WPA/WPA2 naming conventions for its security settings instead of the baffling nomenclature we complained of in our initial look at the product.
Buffalo also briefly released a beta firmware revision for its Nfiniti Router in early July, but the code was removed from the Web site before we could test it.
Technical Analyst Andrew Garcia can be reached at email@example.com.
Check out eWEEK.coms for the latest news, reviews and analysis on mobile and wireless computing.