Benchmark Results: GaNE, 2.4 GHz
Qualcomm's Killer Gaming Network Efficiency benchmark is likely the test that gamers will pore over, as it delves into wireless ping times and jitter. Thus, we’ll emphasize including the raw charts here. The tool works by linking two wireless clients and running data between them. Because of the nature of the configuration, one client would need to stay present to help test the other two and act as a baseline. We selected the Killer 1103 system for this, and as a result we obtained two sets of GaNE Killer data for every one of the Cisco or Intel. In the following results, we averaged the two Killer numbers. Note that for each test, GaNE collected 300 ping data points for each system.
Is it surprising that our first results show Qualcomm handily trouncing the competition with their own benchmarking tool? Probably not. We all expect Qualcomm to win the day on ping times. However, it’s intriguing that the company is able to do this largely through its drivers, since the 1103 adapter itself centers on a standard radio in the company's portfolio, completely lacking the offload capabilities found on its accelerated desktop cards. Clearly, there’s more than hot air to Qualcomm’s claims of having proprietary intellectual property in play here.
Would anyone have guessed that Cisco’s USB adapter would yield lower latency results than Intel’s 6300 flagship? We should would have predicted the opposite. Also observe that there seems to be very little correlation between ping time and distance.
In these GaNE graphs, the Killer 1103 is the red line in all cases. On the left, black shows Cisco and on the right Intel. Do you see that weird spike in latency in the Location 1 Intel results? We were seeing those all the time from Qualcomm before our driver update. That’s how we know that much of the 1103’s performance advantage here is coming from software.
One advantage these graphics provide is that they allow us to see how two adapters are performing in the same time and environment. Any ambient interference affecting one should be affecting the other as both operate within two feet of each other across all locations. At the very least, it seems that Qualcomm is somehow filtering out system overhead or packet transfer limitations that Cisco and Intel are still battling.