Routing throughput was measured running v4.2.1.02 firmware, using our router test process. Table 4 summarizes and compares the RV042G and RV042 v3's routing throughput. Cisco is a bit optimistic in its throughput specs for both models, specifying 100 Mbps NAT throughput for the v3 and 800 Mbps for the G.
|Test Description||RV042G||RV042 v3|
|WAN - LAN||609||91|
|LAN - WAN||492||90|
|Maximum Simultaneous Connections||24061||34925|
Table 2: Routing throughput
The IxChariot composite plot below shows downlink throughput stayed mostly at around 600 Mbps, but with frequent spikes up to as high as 933 Mbps. I saw this behavior both when I ran downlink alone and in the simultaneous up/downlink test. Uplink throughput was very steady when run by itself to produce the 492 Mbps average and dropped to an average of only 126 Mbps during the simultaneous up/downlink test.
RV042G routing throughput
Bandwidth management was not engaged for any of the routing throughput tests, but SPI and DoS features were enabled, which could be the cause of the high variation.
Note that the G's simultaneous connection test did not max out at the 34925 test limit that the v3 hit. I ran the test three times and it sometimes stalled before restarting and then stopping at the 24K or so connections that I recorded as the maximum.
Cisco rates the RV042 v3 IPsec throughput at 59 Mbps and the RV042G at 75 Mbps. Doug's iperf-based tests of the RV042 v3 achieved only a best case of 48 Mbps through a 3DES encrypted tunnel, pairing the RV042 v3 with a NETGEAR SRX5308.
Table 3 shows I did a bit better using a pair of Win 7 PCs connected via Gigabit Ethernet to test gateway-to-gateway IPsec tunnels between two RV042Gs using IxChariot's throughput script with TCP/IP. Although I didn't hit the 75 Mbps Cisco spec, 60 Mbps through an AES 256 encrypted tunnel ain't too shabby. Note that throughput in both directions was pretty evenly matched as shown in Table 3.
Table 3: Gateway-to-gateway IPsec throughpt
I also ran client-to-Gateway tests connecting another Win 7 machine via Cisco's QuickVPN utility and measured a nice steady 54 Mbps. QuickVPN uses 3DES and MD5 in its one-way IPsec tunnel (traffic must be initiated by the client) and I had my usual struggles with it.
Even though the latest 18.104.22.168 version QuickVPN client now offers the hint (when the connection attempt fails) that Windows Firewall must be enabled, I still could not connect when I switched my machine's network profile from Home to Public to comply with this requirement.
After many attempts, I finally succeeded when I stopped trying to use a security certificate exported from the RV042G. I just refused to quit when the missing certificate message popped up (twice) and finally was able to connect and run the tests.
I also checked the RV042G's PPTP server performance using the Win 7 built in client. Gateway-to-client throughput of only 9 Mbps and client-to-gateway of only 12 Mbps (not shown) showed that PPTP performance isn't a priority for Cisco. These results closely match what Doug measured on the RV042 v3 for PPTP.
The IxChariot plot below of PPTP and IPsec tests shows nice steady IPsec throughput, but cyclical throughput variation for PPTP.
RV042G VPN throughput
Keep in mind that the PPTP measurement is client-to-gateway using Win 7's PPTP client and the IPsec runs are all gateway-to-gateway using a pair of RV042Gs.
Cisco has kept all the RV042 v3's features and added the goodness of Gigabit ports. My tests show that the higher bandwidth connectivity provides a significant boost in routing throughput and a smaller goose to IPsec performance. The main negative for the G is the $50 premium (~35%) Cisco wants for the faster ports. Those little Gigabit switch chips sure must be expensive...