Network Performance
So far, everything about the WMAG01 has been pretty good. Installation of the driver and software have been simple, as well as the configuration of the various profiles to be used in our tests. Next, we'll take a look at the performance of the WMAG01. We'll go through our usual benchmark tests with SANDRA, IPerf, SpeedTest, and some real application usage.The two test machines used were configured as follows:
| Laptop | |
| Make: | Dell Inspiron 8500 |
| OS: | Windows XP SP1 and all patches applied |
| RAM: | 512 MB |
| CPU: | Intel P4M 2.2 |
| Wireless NIC: | Gigabyte GN-WMAG01 and built-in Dell TrueMobile 1300 WLAN mini-PCI card. Approx. 10 feet apart and line of sight with AP. |
| Wireless AP: | Gigabyte IEEE 802.11 g Wireless Router (GN-B49G). Used for wireless connectivity. |
| Test PC | |
| OS: | Windows XP SP1 and all patches applied |
| RAM: | 1024 MB |
| CPU: | Intel 2.8 HT enabled |
| Motherboard: | MSI 865 PE |
| NIC: | Onboard Intel Realtek RTL8169/8110 gigabit Ethernet |
| Cabling: | 25 feet CAT5 |
| Wireless AP: | Gigabyte IEEE 802.11 g Wireless Router (GN-B49G). Used for wired connectivity. |
The onboard TrueMobile 1300 is an 802.11b card that helps give a baseline to the performance increases going to an 802.11g Turbo card. With both IPerf and SANDRA, we ran the following tests:
Test 1: Open system, WMAG01 and TrueMobile communicating to wired PC
Test 2: 64 bit WEP, WMAG01 and TrueMobile communicating to wired PC
Test 3: 128 bit WPA-PSK, WMAG01 communication to wired PC
Test 4: Open system, no Turbo G, WMAG01 communicating to wired PC
Test 5: 64 WEP, no Turbo G, WMAG01 communicating to wired PC
Test 6: 128 bit WPA-PSK, no Turbo G, WMAG01 communicating to wired PC
Test 7: Real world file copy tests
Here are the results:
| SiSoft SANDRA RESULTS | ||
| Gigabyte WMAG01 | Dell TrueMobile | |
| Test 1 Open system, TurboG | 44.6 | 18 |
| Test 2 64 bit WEP, TurboG | 42.3 | 18 |
| Test 3 128 bit WPA-PSK, TurboG | 41.3 | N/A |
| Test 4 Open no TurobG | 21 | 16.4 |
| Test 5 64 bit WEP no TurboG | 21 | 15.7 |
| Test 6 128 bit WPA-PSK no TurobG | 20.9 | N/A |
| IPerf RESULTS | ||
| Gigabyte WMAG01 | Dell TrueMobile | |
| Test 1 Default window size | 26.2 | 13.6 |
| Test 1 25K window size | 34.8 | 13.7 |
| Test 1 50K window size | 38.1 | 16.5 |
| Test 1 100K window size | 40.2 | 16.7 |
| Test 2 Default window size | 23.7 | 12.9 |
| Test 2 25K window size | 32.7 | 13.1 |
| Test 2 50K window size | 35.7 | 13.3 |
| Test 2 100K window size | 37.7 | 13.4 |
| Test 3 Default window size | 23.2 | N/A |
| Test 3 25K window size | 30.6 | N/A |
| Test 3 50K window size | 33.9 | N/A |
| Test 3 100K window size | 36.1 | N/A |
| Test 4 Default window size | 20.9 | 13.5 |
| Test 4 25K window size | 21.1 | 14.0 |
| Test 4 50K window size | 21.3 | 15.2 |
| Test 4 100K window size | 21.8 | 16.8 |
| Test 5 Default window size | 20.1 | 13.7 |
| Test 5 25K window size | 20.9 | 14.2 |
| Test 5 50K window size | 21.6 | 14.9 |
| Test 5 100K window size | 21.6 | 16.1 |
| Test 6 Default window size | 20.4 | N/A |
| Test 6 25K window size | 20.6 | N/A |
| Test 6 50K window size | 19.8 | N/A |
| Test 6 100K window size | 21.7 | N/A |
While SANDRA and IPerf tests are synthetic, they do a good job of comparing the two cards. With all this talk about Turbo G, we wanted to see how much difference the newer Turbo G card and AP makes when compared to standard 802.11g. We repeated 3 of the tests with the Turbo G option turned off on both the AP and the client. As shown in the results, having a Turbo G card does make a difference in available bandwidth.
With the earlier 802.11x cards, you would see performance drop as you increased security. With the WMAG01 operating in non-Turbo G mode, we didn't notice any noticeable performance degradation when increasing the security settings.
Next, we have some real-life application testing. We used SpeedTest to measure the time it took to copy a 412 MB file from the PC to the laptop. We tried this with an open system, 64 and 128 bit WEP. The times that we obtained were within less than 15 seconds of each other. The three copy times were 4:46, 4:60, and 4:50 respectively.
Click to enlarge.
Next, we did a comparison of file copy times using DU Meter in different scenarios. The first scenario was such that the PC and laptop were connected via the AP's 10/100 Ethernet ports, to get a baseline. The next scenario involved the WMAG01 to PC, and finally, we measured the time it took the laptop to send and receive data.
The file used was a 2 gigabyte ghost image file and the wireless network was configured for 128 bit WEP. With that, we obtained the following file copy times:
| PC to laptop connected via AP's Ethernet port: | 3:30 |
| PC to laptop via wireless: | 11:56 |
| PC to laptop, and laptop to PC: | 20:214 |
As the copy times show, wireless networking still has a long way to go before even approaching networked 100Mbs. Even though Turbo G is marketed at 108 Mbs, it is misleading. The 108 Mbs is achieved by adding the individual transmit and receive speeds together.
Finally, we tried some other real world usage such as online/LAN gaming and watching DIVX movies. Playing LAN Diablo II and UT2004 for about an hour did not reveal any noticeable lag. The same held true for the online MMORPG City of Heroes. Watching movies on a remote file share went without any problems. The only noticeable differences are the initial load time, and manually "seeking" the movie had taken a little longer. Finally, we tried watching movies while performing a file copy from the laptop to PC, and there were no issues in the movie as well.
Facebook
Twitter
RSS
7 Comments
View All Comments
BrianNg - Tuesday, September 21, 2004 - link
#2 Unfortunately I'm not an LINUX expert and I don't think Kris has the time to do LINUX networking. There's to many variables, packages, configurations that you can do in a LINUX environment.#4 Since AT is new to networking reviews, vendors are hesitant to send a large selection of cards to test. When I do get additional cards, I do a quick retest of the old versus new cards to see which performs better.
Also, the WMAG01 is an Atheros AR5213+AR2112.
#5 I'll add that in my next wireless review.
Thanks,
Brian
mindless1 - Thursday, September 16, 2004 - link
Performance for a G card isn't something I really care about (assuming it's not crippled somehow) but range is. Please include comparisons of attainable range from similar offerings, of course with a breakdown of what the loss is at certain ranges.overclockingoodness - Wednesday, September 15, 2004 - link
I agree with #4. I think AnandTech should've done a huge round-up of these cards and then by creating a base for these reviews, they should've reviewed single products. They have done quite the opposite this time. :)3Suns - Wednesday, September 15, 2004 - link
What is the real point of Anandtech reviewing one wireless card in a sea of largely identical cards? Unless they are going to start comparing lots of wireless cards, this review only says "Yup, they work. Not as fast as wireless though", which is pretty pointless.Also, I agree with #2 and would go on to say that Anandtech should report on the Linux compatibility of everything they review within reason (yeah we know the monitor's gonna work =P). There is a distinct lack of Linux-related hardware reviews today, mostly because only big review sites like Anandtech, TomsHardware, etc. see a large variety of hardware come their way. It'd be really nice to see Anandtech load up Linux when they test the machine, and maybe do a performance test as well. And please not the "works by default in Fedora Core 2" compatibility reviews that one sees everywhere.
At least they should mention the chipset/hardware family. Is this a Aironet-based card? Intersil? Orinoco? Intel? No mention is made what underlying chipset it is, or if it's something completely new.
overclockingoodness - Wednesday, September 15, 2004 - link
I doubt AnandTech is going to do Linux Video Card reviews simply because it is not a matter of plugging it in, firing it up, and doing a quick tests. If they say "Brand A works great in Linux," people will want more and more evidence. They will want more benchmarks and by the time you know it, AnandTech will either have to conduct seperate video cards reviews for Windows and Linux or they will completely stop the Linux testing.I don't think it is a wise idea. I strongly suggest that you visit some Linux/Hardware forums and ask the members. They will better help you decide and answer you since they are regular users, I think they will be your best bet.
darksarin - Wednesday, September 15, 2004 - link
This may be obvious, but I would REALLY love to see information about how well these wireless cards work under linux. Same in video card reviews. I know that some wireless chipsets don't work too hot, so when a new card comes to you guys, it would be cool to see that mentioned.Second, for video cards I have had the unfortunate experience of some brands not working (even though there is chipset support (nvidia) in linux). So if you could just fire it up, test it real quick, and comment on that (works? good) in your reviews, that would be great!
Thanks
SpaceRanger - Wednesday, September 15, 2004 - link
Couple of suggestions:1) When you have your tables, I presume that the larger the # the better the performance? The article doesn't state this.
2) I have the Inspiron 8500 w/ the TrueMobile 1300 MiniPCI card in it. It's capable of doing 802.11g (using it right now infact), and would like to have seen its performance compared to the Gigabyte card.