• "Signal Quality" readings were not available
• Testing was done with a Linksys WPC54G Cardbus client card in a WinXP Home Dell Inspiron 4100 laptop
I used the WRT54G extensively in the examples in my 802.11g NeedToKnow, and will just repeat some of that information here for your convenience. Although as I was writing this review, new firmware was released for the WAP54G that incorporated changes mandated by version 6.1 of the 802.11g draft standard, Linksys had not yet released a similar version for the WRT54G. So the results you see here reflect the performance provided by 1.01.4 router firmware and original WPC54G client drivers.
I also won't be spending any time discussing the ins and outs of draft-802.11g throughput testing. Again, if you want the background, read the NeedToKnow. I'll only say that to get an accurate picture of what draft-802.11g products will deliver in the throughput department, you need to test under three scenarios:
Without any 802.11b clients active
With an 802.11b client associated to the same draft-802.11g access point or router, but idle
- With an 802.11b client active, i.e. transmitting data simultaneously with 802.11g data
Scenario 1 - No 11b active
Figure 8 shows the first scenario and with WEP disabled and with the router set to Mixed wireless mode.
Figure 8: Four Condition Throughput test
The "warmup" effect that I found in my WAP54G testing appeared to be solved by the newer firmware and the 21Mbps average throughput under Condition 1 is the best of the draft-11g products that I've tested so far. I didn't try to find the limit of the range of my test pair, since I was seeing signs of being close to the range limit similar to those that I see when testing 802.11b product. It's worth noting, though, that my worst-case Condition 4 throughput of about 4.8Mbps is about equal to the best throughput that typical 802.11b equipment provides.
Also in Linksys' favor is that enabling WEP doesn't appreciably affect throughput, as shown in Figure 9.
Figure 9: Test Condition 1 - No WEP vs. 128 bit WEP
The summary for Scenario 1 then, is that the WRT54G does pretty well with no 11b competition around and with a Linksys client card.
Scenario 2 - 11b associated, but idle
As I've discovered while testing both Broadcom and Intersil-based draft-802.11g products, the trend is definitely toward implementing automatic and what I call "adaptive" 802.11b protection. This method of protection yields higher throughput in what appears to be a protection "standby" or "sniffing" mode, when there are no 802.11b stations in range. Throughput drops noticeably, though, once an 11b client is detected and associates with the AP. Figure 10 provides two examples of this behavior.
Figure 10: WAB501 vs. ORiNOCO Temporary Association
The screen shot shows the effect from the temporary association of two different 802.11b clients on a throughput test running between the WRT54G and a WPC54G client. During each Chariot run, I plugged an 802.11b client into a second laptop sitting close by at about the 20 second point in the test run. I then shut down and removed the 11b card at about the 50 second mark.
The bluish plot shows the throughput effect from the temporary association of a NETGEAR WAB501 Dual-Band CardBus card (Atheros 5100X-based), and the reddish plot the effect from an ORiNOCO Gold client. Although both cause a throughput drop at about the same time, the effect from the WAB501's association goes away seconds after I shut down and remove the card. But throughput doesn't recover from the ORiNOCO's association until about a minute after I remove the card.
The main points here are:
The throughput of your draft-11g WLAN will drop when even one 802.11b client comes into range and associates with your access point or wireless router
The lowered throughput can remain in effect for minutes after 11b clients leave your WLAN.
I expect that manufacturers will be tuning their algorithms to minimize both effects, but you should expect lower throughput when 802.11b clients are present in your draft-802.11g WLAN, at least until manufacturers stop tweaking their protection algorithms.