2x Speed "Plateau Shift" with RT-AC66U

[civuck]

After some reading, I seem to be getting the same speed "plateau shifting" that Tom's Hardware recently reported on in their Gigabit Wireless article and router roundup. (I have the RT-AC66U with .270 firmware and the PCE-AC66 adapter)

This is best described as a sudden jump (up to 2x) in throughput speed during a file transfer. Unfortunately I've only observed this desirable "speed boost" on a few occasions, despite testing transfers with the exact same file on my NAS under the exact same configurations.

I've been trying to find the magic ingredients that allow for this but I haven't been able to recreate it.

Here's the Tom's Hardware graph illustrating this phenomena which is near identical to what I have seen:

So this begs the question - is 300 Mb/s+ transfers what we should be expecting under real-world conditions for Wireless-AC?

And is anyone else seeing this as well? The router's built-in traffic monitor tool will show it clearly during transfers.

 

[thiggins]

What are you doing getting your information from that rag...?
http://www.smallnetbuilder.com/wire...art-wi-fi-router-ac-1750-hd-video-pro?start=3
http://www.smallnetbuilder.com/wire...1750-gigabit-router-reviewed?showall=&start=3

 

[civuck]

What are you doing getting your information from that rag...?
http://www.smallnetbuilder.com/wire...art-wi-fi-router-ac-1750-hd-video-pro?start=3
http://www.smallnetbuilder.com/wire...1750-gigabit-router-reviewed?showall=&start=3

If I'm interpreting this graph correctly, it seems it is clearly location that allows for the sustained 200mbps+ uplink throughputs with the Linksys.

But the graph for the ASUS seems to indicate location has less of an affect.

I'm still left trying to find exactly what conditions/settings lead to these "upshifts", as they dramatically increase throughput with Wireless-AC.

I was able to reproduce them time after time one evening and then not at all the next.

I'm wondering if it is related to heat in the router at all.

 

[thiggins]

It's really still early days in draft 11ac and drivers are still being tuned. I wouldn't drive myself crazy trying to figure it out.

You can't really compare 11ac router test results apples-to-apples yet either. I use a standard test client with 11n routers. But for 11ac, I'm still using a "matching" client provided by the manufacturer.

So, too many variables to draw any sort of rational conclusion.

 

[civuck]

It's really still early days in draft 11ac and drivers are still being tuned. I wouldn't drive myself crazy trying to figure it out.

You can't really compare 11ac router test results apples-to-apples yet either. I use a standard test client with 11n routers. But for 11ac, I'm still using a "matching" client provided by the manufacturer.

So, too many variables to draw any sort of rational conclusion.

I appreciate your follow-up and insight in to this.

As it is now, these random/unpredictable throughput "shift ups" really seem to mean the difference between wireless-ac being a minor upgrade over wireless-n and a major one. Hopefully, as you indicate, mature drivers will bring more consistency.

 

[thiggins]

And keep in mind, you are looking at strong-signal performance. Throughput can drop rapidly as signal level decreases. And range /coverage will be reduced due to higher signal attenuation for 5 GHz.

 

[civuck]

I captured this screenshot from the AC66u's built-in traffic monitoring tool earlier this week. It's the last time I saw it go above ~22MB/s during an 8GB file transfer from my Synology NAS to my PC:

Despite all conditions/configurations being the same, I have not seen it go above 22MB/s since then, despite numerous...numerous.. attempts.

Again, I'm wondering if this could be heat related and perhaps the firmware limits speeds when the unit running too warm. My unit is currently sitting on the floor without the included stand which may be causing it to heat up more. I understand the custom "Merlin" firmware can report temperatures; perhaps there's a correlation to review.

 

[RMerlin]

Out of curiosity, can you try this beta firmware (http://forums.smallnetbuilder.com/showpost.php?p=62090&postcount=165)?

That firmware is based on the new wireless driver that Asus is currently working on.

 

[civuck]

I've loaded that Beta firmware into my AC66U. Throughput is still not going beyond ~22MB/s like it did in my graph posted above, despite all conditions/settings remaining the exact same.

My PCE-AC66 adapter is reporting a link rate of 1053 Mbps and the inSSIDer app is reporting no other wireless signals in the 5ghz range other than my own; basically near perfect conditions.

Looking at the throughput graph during a transfer, it is very flat at 22MB/s with no major dips - but it just will not "upshift" into the 30 to 40 MB/s+ range like was doing repeatedly earlier this week.

 

[RogerSC]

What are you doing getting your information from that rag...?
http://www.smallnetbuilder.com/wire...art-wi-fi-router-ac-1750-hd-video-pro?start=3
http://www.smallnetbuilder.com/wire...1750-gigabit-router-reviewed?showall=&start=3

*smile* Good one, I like that.

 

[civuck]

So as reported by the AC66U built-in traffic monitor, here's what happened when I started 5 simultaneous downloads from my Synology NAS with the Firefox web browser:

This seems to indicate that, if nothing else, my wireless connection is good as I saw a burst up to 63 MB/s ! I believe the bottleneck at this point is my NAS itself as that's equal to about the best speed I've seen from it going over a wired connection to an SSD.

However, what still isn't clear is why one browser file download uses only ~22MB/s while five use between 40 and 50MB/s. I would expect the maximum throughput to be identical regardless of how many transfers were occurring.

Other notes:

• My 5ghz radio temp was up to around 64 celsius according to the 3.0.0.4.270.24 Merlin firmware
• My Core i7 PC was running at about 25% CPU usage while these transfers were occurring
• The data rate as reported by my ASUS PCE-AC66 adapter utility fluctuated between 877 and 1053 Mbps, channel was 149, channel bandwidth set to 20/40/80, Tx power was 80mW and RSSI of -42 according to inSSIDer app

 

[p71]

You mentioned that you used Firefox for the transfer test. Does your NAS provide a file browser interface or I misunderstand the situation?

 

[civuck]

You mentioned that you used Firefox for the transfer test. Does your NAS provide a file browser interface or I misunderstand the situation?

Yes, my Synology NAS has a web-based interface called "DSM".

Here's a live-demo on the Synology website if you are interested in seeing what it looks like.

 

[civuck]

Out of curiosity, can you try this beta firmware (http://forums.smallnetbuilder.com/showpost.php?p=62090&postcount=165)?

That firmware is based on the new wireless driver that Asus is currently working on.

When I still had that beta firmware installed (I've since moved over to Merlin .270.24) there was a significant degradation of my 2.4ghz connection. Normally my Nexus 7 tablet reports the maximum 65Mbps connection rate. But with that firmware, it was down to 5Mbps consistently. My first clue was video streams from my NAS were stuttering.

 

[luke01]

After some reading, I seem to be getting the same speed "plateau shifting" that Tom's Hardware recently reported on in their Gigabit Wireless article and router roundup. (I have the RT-AC66U with .270 firmware and the PCE-AC66 adapter)

<snip>

About Tom's Hardware graph illustrating this phenomena that you mentioned,
it seems regarding Windows System. If you use Win XP to test, the throughput speed will be stable at the beginning, no sudden jump phenomenon. but the throughput has sudden jump in Win 7. Please refer to attachements. In Win XP, the TCP WindowSize can be set to a fixed max value. but Win 7 uses auto-tuning mechanism for TCP receive window. I think this is why the sundden jump happen in throughput speed.

 

[civuck]

In Win XP, the TCP WindowSize can be set to a fixed max value. but Win 7 uses auto-tuning mechanism for TCP receive window. I think this is why the sundden jump happen in throughput speed.

Interesting post. Thanks,

So the obvious question is: is there a tweak for Windows 7 to give equal performance as WinXP?

 

[civuck]

Wow! It seems I found the magic ingredient based on the last poster's comment.

After finding this page, I made a tweak and now I am consistently getting 40MB/s speed when downloading from my Synology NAS via web browser over Wireless-AC. Previously I was hitting a limit of around 21 MB/s.

My Samba copy speeds from my NAS also went from about 12 MB/s to about 19 MB/s.

My Windows 7 "Receive Window Auto-Tuning Level" was initially set to Disabled. I changed it to Normal to get these speed boosts:

The speed is all there immediately when the file copy starts - I haven't observed any of the "plateauing" effect yet where it starts slow and then jumps up (as shown in some graphs).