How Much Throughput Can You Really Get From An AC Router?

[theoak]

Happy New Year!

NETGEAR used to have a version of its firmware with an "Airtime Fairness" option. The general idea being for a mixed network of fast and slow wireless devices, the router would try to ensure all devices got the same amount of the router's time.

I wonder if airtime fairness would have any impact in this specific test case too? We will have to wait until NETGEAR adds the option back in to see of course.

Interesting results!

 

[thiggins]

I don't know how you can get any more fair than the sharing shown with capped transmit rates. Remember that there is virtually no application that can blast data at the rate that the IxChariot throughput script can. There is virtually no protocol overhead.

At any rate, I would think the "Airtime Fairness" would apply to mixed standard devices, not same standard, different streams.

 

[sinshiva]

hey, i never used the IxChariot program, but my former testing methodology usually involved DataRAM software RAMdisks. Appears they've seen been bought by AMD or something. Also used tmpfs, of course.

Originally, i had 12gb of ram so I would use 4gb ramdisks with good stability while using the ramdisks, however the machine's overall performance seemed to degrade over time whereas without the ramdisks running, not at all. When i lost 2 sticks of ram (wouldn't OC high enough with these particular sticks and with the particular mobo as populated as it was), I dropped down to 8gb total mem and continued to play with 1 and 2gb ramdisks, until i got annoyed with the instability. (this was all back on i7 920 6x2gb triple channel. very populated. anybody keeping track, i've decided to go haswell.)

In the beginning, I used the RAMdisks for basic file copies to see 'real world' read/write performance of lots of different kinds of drives. The results were always extremely interesting, having such a universal, yet simple methodology for testing all kinds of things like partition formats (reiserfs and xfs were interesting before ext4), single drive I/O, 4x single 500gb platter Raid0 I/O, encryption cipher I/O, file conversion, you name it. I've also seen interesting results running different benchmark suites directly off the RAMdisks.

Has there been any comparisons here between RAMdisks/tmpfs and the benchmark suite? I intend to play with ramdisks again, myself, soon.

 

[thiggins]

Filecopy still has more overhead than IxChariot throughput.scr. I don't use filecopies for wireless testing.

 

[armia741]

When i lost 2 sticks of ram (wouldn't OC high enough with these particular sticks and with the particular mobo as populated as it was), I dropped down to 8gb total mem and continued to play with 1 and 2gb ramdisks, until i got annoyed with the instability. (this was all back on i7 920 6x2gb triple channel. very populated. anybody keeping track, i've decided to go haswell.)