Introduction
| At a Glance | |
|---|---|
| Product | – Buffalo Wireless-N Nfiniti Dual Band Gigabit Router & Access Point (WZR-AG300NH) – Buffalo High-Speed wireless CardBus adapter (WLI-CB-AG300N) |
| Summary | Simultaneous dual-band gigabit LAN draft 11n router and CardBus card based on Marvell silicon |
| Pros | • Two radios for simultaneous 2.4 and 5GHz band support • Automatic wireless security setup • IPv6 support • Simple priority-based QoS |
| Cons | • Not Draft 1.10/2.0 compliant • Does not automatically switch between 40 and 20MHz bandwidth modes • Reduced range in 40 MHz mode |
A lot has happened in the four months or so since I temporarily swore off reviewing draft N products. 11n draft 1.10 was approved, Apple announced a draft 11n AirPort Extreme and revealed that it had hidden draft 11n wireless inside most of its recent Macs and Intel tossed its hat into the draft 11n ring.
But the more relevant news—at least to this review—is the announcement of dual-band draft 11n gear. It seems that manufacturers are finally realizing that most 11n gear will never be able to achieve full channel-bonded speeds in the 2.4 GHz band without wiping out neighboring networks. Instead, they’ll need the relatively wide-open spaces of 5 GHz in order to achieve their much-hyped top speeds.
Apple, D-Link and Buffalo announced dual-band products at or around January’s Consumer Electronics show (and Linksys withdrew its prematurely-revealed entry). Apple was the first to ship the single-radio Airport Extreme mentioned above. But Buffalo is the first to ship a two-radio dual-band draft 11n router capable of simultaneously supporting 2.4 and 5 GHz wireless LANs. And Buffalo’s Wireless-N Nfiniti Dual Band Gigabit Router & Access Point (WZR-AG300NH) is what I’ll be looking at in this review.
Check out the slideshow for a router admin interface tour.
Product Overview
Buffalo has been doing a better job of late with the industrial design of its products, particularly in its NASes. But a different team of designers must have been assigned to the "AG". With its cabled antenna puck and homely silver/grey plastic enclosure, it’s not a particularly elegant product; and even less so when arranged in its vertical mode (Figure 1).
Figure 1: The AG standing up
The front panel has the usual assortment of status lights as you can see in Figure 2 below. Note that Buffalo has done away with the blue LED indication of a gigabit Ethernet connection, judging it to be too bright. So you’ll have to depend on your computer to tell you the negotiated link speed.
Figure 2: Front Panel
Connectors on the rear panel (Figure 3) include four 10/100/1000 LAN ports, one 10/100/1000 WAN port and power jack. All ports are auto MDI / MDI-X which means they’ll figure out how to connect to whatever you plug into them. There’s also a reset-to-factory-defaults switch and a switch to make the AG function as an Access Point. The array of three antennas is hard-wired to the back of the router, although they terminate in miniature connectors on the two radio modules inside.
Figure 3: Rear Panel
Like the D-Link DIR-655 that currently tops our router performance charts, the AG has both gigabit LAN and WAN ports. But as you’ll see later, the AG doesn’t have comparable routing performance.
Internal details
The AG’s Marvell-based board looks like it is expecting some heat to be generated. Figure 4 shows that the Marvell 88F5181 processor (center left), 88E6131 gigabit Ethernet switch (bottom left) and two 88E1112 gigabit tranceivers (bottom right) all have heatsinks.
The label on the (5 GHz band) radio module is apparently for an alternative design, since the AG has two radios, with the second (2.4 GHz band) module mounted on the bottom of the board (see Figure 5).
Figure 4: Main board top
The gigabit switch chip must run really hot, since it has a heat spreader (The black metal plate at bottom right) on the bottom of the board as well as the heatsink on the device itself.
Figure 5: Main board bottom
Figure 6 shows the single-band radio module without its shield. Both modules use the same Marvell 88W8363 MAC/Baseband chip, which I assume is an enhanced version of its TopDog 88W8360, and 88W8060 2.4 / 5 GHz RF Tranceiver.
Figure 6: Mini-PCI radio module
Figure 7 shows the board of the AG’s companion WLI-CB-AG300N Cardbus card. The MAC/Baseband chip markings are blacked out, but you can see the Marvell logo on the center chip under the shield fencing. It also uses the 88W8363 and 88W8060.
Figure 7: CardBus board
Routing Features
The AG has a few features that I haven’t seen in previous Buffalo routers. But in order to focus on the wireless aspects of the product, I’ll just summarize the key routing features here.
Check out the slideshow for a router admin interface tour.
The Setup page (Figure 8) that you see when you first log into the AG provides a peek at some of the unique capabilities.
Figure 8: Setup page
- Router and AP mode – You can switch between modes using either the hardware switch on the back of the router, or using the Change Mode button on this page. Using the AP Mode hardware switch changes the default IP address from 192.168.11.1 to 192.168.11.100 and disables the DHCP server and NAT routing. You use the WAN port, however, to connect the AG to your LAN. Using the software method does all the above, but lets you define the AG’s IP address.
- IPv6 – This protocol, which will be the way we all connect someday. But for now, these controls will be useful if the AG finds its way into U.S. military use, or by users outside the U.S. (The Department of Defense (DoD) is driving toward establishing IPv6 in all its Internet and intranet systems by fiscal year 2008.)
- NAT and routing – You can disable NAT, set static routes and control RIP1 and 2 transmit and receive
- Dynamic DNS clients – Clients for Dyndns.org and TZO.com are included
- QoS – Eight applications can be assigned high, medium or low priority for LAN to WAN packets. This will help manage the relatively paltry uplink bandwidth that most of us get from our ISP. But it won’t help keep your kids from sucking up downstream bandwidth with their insatiable appetite for huge downloads.
- Firewall filters – You get both inbound and outbound filters that are defined separately for IPv4 and IPv6.
What you don’t get:
- Some Firewall controls – Functions that would be useful, but are missing, are the ability to disable the firewall SPI functions and blocking of cookies, ActiveX and Java applets and web proxies.
- Secure remote access – Remote access is HTTP only, and you can restrict access to only a single IP and set the port.
- Admin idle timeout adjust – One of my personal annoyances. The timeout appears to be fixed at around 5 minutes.
Routing Performance
I put the AG through our suite of router tests, and the results are shown in Table 1.
| Test Description | Throughput – (Mbps) |
|---|---|
| WAN – LAN | 201 |
| LAN – WAN | 187 |
| Total Simultaneous | 192 |
| Firmware Version | 1.46 |
Table 1: Routing throughput
This puts it in second place in our Router Performance Charts, right behind the D-Link DIR-655. Figure 9 shows what the throughput variation looks like on the simultaneous routing test.
Figure 9: Simultaneous throughput test results
But those of you hoping to put the AG into service for BitTorrent or other P2P filesharing will be a bit disappointed with its ability to support simultaneous connections. I was able to reliably achieve only 64 simultaneous (32 up, 32 down) connections. This is much better than the Belkin N1’s unusually low 10, but less than the D-Link DIR-655’s 120.
Wireless Features
Most notable among the AG’s wireless features is its support for both Buffalo’s proprietary AOSS and the took-way-to-long-to-appear Wi-Fi Protected Setup (WPS). WPS (Figure 10) is a compromise specification brokered by the Wi-Fi Alliance in order to finally make secure wireless LAN setup as easy as pushing a button, even if the connecting products come from (gasp!) different vendors!
Figure 10: WPS setup
Unfortunately, the early version User Guide that came with the AG had no information whatsoever on WPS and the Buffalo Client Manager 3 application sported only an AOSS button.
So I tried using AOSS, but initially couldn’t get it to work. After some futzing around, I got it working by disabling WPS in the router (both WPS and AOSS are enabled by default). I also had to initiate the AOSS session by using the soft button in the admin interface (Figure 11), since disabling WPS seemed to disable the hardware button on the router, too.
Figure 11 shows the results of a successful AOSS session, with both radios secured by WPA-PSK (AES). If you have to set up security manually, you have the options of WEP 64/128, WPA-PSK (TKIP and AES), WPA2-PSK and a mixed mode that supports both WPA TKIP and AES clients.
Figure 11: Successful AOSS setup
Wireless Features – more
The other wireless controls are essentially the same for each radio, so I’ll show just the 11a / 5 GHz radio’s (Figure 12). The Basic controls are pretty straightforward and require only a little explanation. The 11a radio supports only four of the eight channels allowed in the U.S. (36, 40, 44 and 48) and defaults to 40 MHz bandwidth mode (channel bonded). You can manually select the Wireless (primary) channel in 40 MHz mode but the "Extension Channel" is automatically set for you.
Figure 12: 11a basic radio settings
In a nod to being a better wireless neighbor, the 11g radio defaults to 20 MHz mode (not channel bonded). But you can switch it to 40 MHz mode, which, like the 11a radio, doesn’t let you set the Extension channel if you choose to set the primary channel.
The Advanced settings (Figure 13) are a bit more obscure. Unfortunately, the User Guide is no help and the online help isn’t much better. The AG isn’t compliant with the 11n 1.10 draft and doesn’t support the three mechanisms added in that draft to solve 11n’s 2.4 GHz "bad neighbor" problem. To the contrary, if the 802.11n Protection control really "gives priority to 802.11n devices in mixed mode" as the online help explains, then 11b/g clients will never have a chance at airtime.
The 11g radio has the same set of controls, with a different set of Multicast Rate settings and an additional 802.11g Protection checkbox. Kudos to Buffalo for including the Privacy Separator control that blocks wireless client-to-client connection. I just wish Buffalo would also add a control to keep wireless and wired clients separated.
Figure 13: 11a advanced radio settings
For a real set of incomprehensible controls, the WMM Settings take the prize. The online help’s notation that "It isn’t necessary to change this value ordinary." (sic) sounds like good advice.
Figure 14: WMM controls
Finally, it’s worth noting that while you can enable MAC address filtering for each radio separately, the same MAC address list is applied to both radios. You do get a handy list of currently-associated clients, however, to ease the task of creating the list.
Wireless Performance
But you’re really reading this review to see how draft 11n works in the 5 GHz band, aren’t you? So, let’s begin. My last round of draft 11n testing included throughput vs. range testing done using Azimuth’s W-Series Test System. But just as I was about to start this review, Azimuth asked me to change over to using its ACE 400NB Channel Emulator.
The ACE is internally a pretty complex box, performing bi-directional emulation of up to 32-channel 4×4 MIMO systems. You can choose from six standard usage models as well as two special modes: Bypass and Butler. Figure 15 shows a block diagram of the ACE, so that you can get an idea of what’s inside.
Figure 15: Azimuth ACE NB Block Diagram
I can’t begin to scratch the surface of what the ACE can do and how it does it in this review, but the basic idea is that the ACE digitizes the incoming signal from up to four channels and applies some pretty hefty digital signal processing to it to simulate multipath, delay, fading and the other stuff that happens to the RF signals in the real world. After all the crunching, the signal is converted back to analog and sent out the second set of ports.
Figure 16 shows pretty much what my test setup looked like. (I’m using the diagram until I can get my setup looking a bit more photogenic.) The router was taken out of its enclosure and put into one of Azimuth’s RF-tight enclosures, while my notebook containing the Buffalo nFiniti companion card went into another. With the help of my trusty tools, I got the router out of its case and the plastic top off the card’s antenna area so that both were directly cabled to the ACE.
Figure 16: Azimuth ACE Test Setup
The Buffalo WLI-CB-AG300N Cardbus card was inserted into a Fujitsu P7120 Lifebook (1.2 GHz Intel Pentium M, 504 MB) notebook running WinXP Pro SP2 with all the latest updates. I installed the 3.0.0.10 driver and version 1.2.6 of Buffalo’s Client Manger 3 utility. The router ran version 1.46 firmware and I left all factory default settings in place except as noted. When testing the 2.4 GHz radio, I used the defaults of primary Channel 6 and "extension" channel 10. For 11a, I used primary Channel 36 and extension channel 40.
To start, I did some open-air IxChariot runs to look at maximum performance and throughput variation. This also gave me some baselines to check Azimuth results against. Figure 17 show simultaneous up and downlink throughput for the 2.4 GHz band using 40 MHz (channel bonding) mode. This plot was made with the router and card about three feet apart in open air sitting on my lab desk with no other networks in range. Aside from the short rise up to the steady-state speed, throughput is well-behaved. 11a close-range open-air performance is very similar, so I’m not showing that plot.
Figure 17: 2.4GHz up/dn throughput – 11n 40MHz mode
The nFiniti supports a full range of non-enterprise grade wireless security options including WEP 64/128, WPA-PSK (TKIP and AES), WPA2-PSK and a mode that allows a mix of WPA/WPA2 TKIP and AES clients. While I did see some degradation in throughput when using security, the performance loss was nowhere near the severe reductions I found in the Belkin N1. Instead I saw losses of around 6% and 10% for WPA AES and WEP128 respectively.
As I found in the Belkin N1, WPA TKIP was once again the worst performer with around a 22% throughput loss. Actually, I had such a hard time getting WPA-PSK TKIP to even work at all that I would recommend avoiding it if at all possible.
Wireless Performance – Throughput vs. range
The latest word I have is that we won’t be seeing draft 11n products that incorporate the protection mechanisms in Draft 1.10/2.0 for legacy 2.4 GHz gear until the May/June time frame. So I did not do any interoperability or "bad neighbor" testing. Instead, I used the Azimuth to compare 2.4 and 5 GHz band performance. I also ran comparisons with a dual band a/b/g router and card.
Azimuth advised starting with the ACE set to Butler mode. Butler mode emulates a unity gain, non-fading channel with a fixed set of antenna-to-antenna phase relationships. According to Azimuth, Butler mode provides a channel with a "high degree of linear independence". In simpler terms, Butler mode essentially makes the ACE disappear and so can be used to look at best-case performance.
I also spent some time trying to correlate results between the Azimuth W Series MIMO test and the ACE. After some back-and-forth with Azimuth, I was able to get the shape of the performance vs. range "waterfall" curve to be pretty similar, but found that gain differences in the ACE’s algorithms shifted the curves along the path loss axis with respect to each other. So if you compare these results to those in my earlier tests, keep that in mind.
Let’s start out by looking at the throughput vs. range curves. Figure 18 shows 2.4 GHz band results in both 40 MHz (channel bonding) and 20 MHz modes. Figure 19 shows the same thing for the 5 GHz band.
Figure 18: 2.4 GHz band throughput vs. range
It’s immediately obvious that with the current nFiniti firmware, you have your choice of high speed and shorter range or lower speed and longer range. The main reason appears to be that the range setting algorithms don’t allow downshifting from 40 MHz to 20 MHz mode when the signal gets too low to sustain connection in 40 MHz mode. In other words, there is no auto 20/40 MHz mode.
The other observation is that uplink speed is greater than downlink. This is something that is consistent throughout my test results, so it’s possible that it could be due to TCP/IP or other settings that I don’t futz with.
Figure 19: 5 GHz band throughput vs. range
Wireless Performance – 2.4 vs. 5 GHz
The nFiniti is on the leading edge of what I think will be a trend back to dual-band WLAN gear. Once the 2.4 GHz legacy protection mechanisms are implemented, most users will find it difficult to achieve maximum throughput in the overcrowded 2.4 GHz band. So the move to 5 GHz will be necessary for many users.
However, 5 GHz wireless gear has been forever haunted by the ghost of first generation 11a gear, which had lower throughput vs. range performance than 11b/g products. 802.11a performance was long ago improved in Atheros’ second generation chipset. But given the physics of RF propagation, everthing else being equal, the throughput vs. range curve for 5 GHz operation should still be under that for 2.4 GHz band operatoin.
Figures 20 and 21 show throughput vs. range for both bands on one plot for both 40 and 20 MHz modes. Please note the difference in Y axis scale between the two plots, as the Azimuth Studio application that I used to generate these plots doesn’t currently have the ability to control the plot axes.
Figure 20: 5 GHz vs. 2.4 GHz throughput vs. range – 40 MHz mode
What is interesting is that the performance is so close between the two bands and in Figure 20, 5 GHz band uplink throughput actually exceeds that of 2.4 GHz in some portions of the curve. 20 MHz mode performance shown in Figure 21 shows a picture more like what I expected, with 5 GHz throughput rolling off before 2.4 GHz.
Figure 21: 5 GHz vs. 2.4 GHz throughput vs. range – 20 MHz mode
The conclusion you would reach from these plots is that you would expect similar performance from using either band—or at minimum, be able to remain connected in the same locations with either band. However, a quick walk around my home with the nFiniti card in my notebook linked to the nFiniti router showed that the Azimuth results don’t accurately reflect the difference in real-world operation.
At one low-signal spot, the Windows Wireless Zero Configuration utility showed 4 bars of signal from the 2.4 GHz radio, but only one bar from the 5 GHz radio. But I’ve found signal level indications (and link rate especially) to not be accurate for other products, so I’m not basing my judgement solely on that.
The more telling test was done using IxChariot in my "black-hole-of-wireless" test Location 4. When connected to the 2.4 GHz radio, I was able to continue a test—although as much lower throughput—that I started at a location where both radios could connect. But when I connected to the 5 GHz radio and moved to the black hole spot, I promptly disassociated from the router.
I discussed these results with Azimuth before posting this review and they confirmed that the ACE’s models do not incorporate signal attenuation differences between the 2.4 and 5 GHz bands. Azimuth and I will be discussing ways to have the ACE more accurately reflect "real world" 2.4 vs. 5 GHz operation in future reviews. But for now, these results can only be used for comparing performance of products within a given band.
Wireless Performance – Draft 11n vs. Super AG, 11a, 11g
The last question I’ll explore is whether draft 11n provides a performance advantage over legacy dual-band gear. I loaded a Netgear FWAG114 wireless router and WAG511 a/b/g cardbus card into the Azimuth to attempt to answer this question. Because this is SISO (Single Input, Single Output) gear, I used the ACE’s simpler Bypass model and connected only one of the two antenna outputs on both card and router. (The second antenna is for diversity purposes, not another channel, so isn’t needed for the test.)
Since the Netgear products are based on Atheros’ Super AG / G technology, which also employs channel bonding for higher thoughput, I ran tests with the pair set to both "108 Mbps" (channel bonding) mode and straight 802.11a and 11g modes.
Figure 22 shows a 2.4 GHz band test that compares the Buffalo gear set to 40MHz mode vs. the Netgear set to "108 only" mode. The results show that the 11n gear delivers up to 4 times the speed of Super G, but fails to deliver the same range due to the Buffalo gear’s inability to downshift out of channel bonding at lower signal levels.
Figure 22: 11n 40 MHz mode vs. Super G throughput vs. range – 2.4 GHz band
Figure 23 switches the comparison to the 5 GHz band, which shows similar results.
Figure 23: 11n 40 MHz mode vs. Super AG throughput vs. range – 5 GHz band
Switching the Buffalo gear to 20 MHz mode looks like the way to go for either 2.4 (Figure 24) or 5 GHz bands (Figure 25).
Figure 24: 11n 20 MHz mode vs. 11g throughput vs. range – 2.4 GHz band
You don’t sacrifice that much throughput by switching to 20 MHz mode, but gain "friendliness" with your 2.4 GHz neighbors and much improved range.
Figure 25: 11n 20 MHz mode vs. 11a throughput vs. range – 5 GHz band
Closing Thoughts
The Buffalo nFiniti is the first simultaneous dual-band draft 11n router on the market and with two radios, its ~$250 street price will be an obstacle for some. But as I said earlier, if you’re looking to be able to reach channel-bonded speeds, you’re going to need the option of moving to the 5 GHz band, since you’re unlikely to be the only person in your neighborhood with a wireless LAN.
That being said, since the nFiniti doesn’t have the ability to downshift from 40 to 20 Mhz modes at lower signal levels, you’re probably better off using the default 20 MHz mode setting in the 2.4 GHz band and switching the 5 GHz radio also to 20 MHz mode if you use it. My tests show you’ll sacrifice little throughput, but gain a lot in range.
However, despite the wireless performance and decent, although not outstanding, routing features and performance, I’m still maintaining my "do not buy" stance on draft 11n products, so the nFiniti falls under that umbrella. While some of my reviewing brethren are saying soothing things about taking the step to buying draft 11n products, I maintain that there still are plenty of bumps in the road ahead.
Implementation of the new protection modes mandated by Draft 1.10/2.0 is sure to be fun and will take a lot of work (and time) to get right. And tuning performance for mixed client networks is sure to be a ton of fun for vendors—not to mention frustrating for buyers who will wonder why their notebook with built-in 11g radio runs so slowly when connected to their whizzy new 11n router.
So while the Buffalo nFiniti is a significant step forward in the march toward draft 11n, it’s not time to take the leap yet. My advice is to wait until after the Wi-Fi certification program is in place and Draft 2.0 compliant products start appearing. In the meantime, your options will only increase and prices will start to come down. And the Back to School sales are a good a time as any to treat yourself to a long awaited WLAN upgrade.
Check out the slideshow for a router admin interface tour.