Joined: 30 May 2007 Posts: 94 Location: Arizona, USA
Posted: Sun Nov 07, 2021 20:31 Post subject: What wireless speeds should I expect on WNDR3700v3?
I checked out this article (again), that is suggested reading in a sticky here, but I want to confirm if I'm understanding correctly.
It says:
Quote:
a single high-quality client device will generally get somewhere between one-third to two-thirds of the PHY rate for the channel it's connected to [...]
Now, my manual for the WNDR3700v3 claims "up to" 300 Mbps at 5 Ghz. Is this 300 number the PHY rate? I can't find a proper spec sheet. So, do I take this number and apply the 1/3-2/3 rule and conclude that I should expect to get something like 100 to 200 Mbps download speeds (for 1 device) under ideal conditions? _________________ Not all those who wander are lost
300Mbit + 300Mbit (the maximum your router supports)
and no I do not have a WNDR3700v3
this was tested with a R7800 with adjusted parameters
the R7800 supports of course even more like 256-QAM, 4x4MIMO, 160Mhz channel width, IEEE 802.11ac (800 + 1733Mbit)
But I did NOT test this as you can see on the screenshot but 40Mhz IEEE 802.11n 2x2 (300Mbit)
and i think that answers your question, this is the throughput you can expect in good conditions for the wifi standard
the actual throughput for your device depends of course on other factors like CPU performance, driver and chip quality, the used client and WLAN interference
Joined: 30 May 2007 Posts: 94 Location: Arizona, USA
Posted: Sat Nov 13, 2021 7:07 Post subject:
ho1Aetoo wrote:
Yes, this is a throughput measurement from WLAN to LAN.
[...]
the actual throughput for your device depends of course on other factors like CPU performance, driver and chip quality, the used client and WLAN interference
I see, thanks.
From you mentioning the CPU, and from other threads, I think we can assume that the old electronics of my device are my main limiting factor. However, I'm still struggling to get a definitive answer to the following two questions.
1.) Should I expect the speed on DD-WRT to always be substantially slower compared to stock firmware? And if so, roughly how much? Is 50% slower a normal outcome?
2.) Should I expect wifi download speed to always be substantially slower than the wired download speed, even under optimum conditions? Roughly how close to the wired speed can the wifi speed get? Can I expect 90%, 80%, or what...?
Regarding question #1, I've searched the forums, and it looks like I'm seeing this pattern from a few threads, for example here, here, and here. In particular, in that last link, the person got exactly 50% improvement in wired download speed by reverting to stock fw: went from 80 to 120 Mbps. Can you confirm: should I be unsurprised to see a similar dropoff in speed for my device relative to stock firmware?
As for question #2, I've identified in my other thread that I'm getting about 120 Mbps on wired. So, from that baseline, should I be happy with ~80 Mbps on wireless on the 5 GHz radio? Or should I expect it to perform closer to the wired throughput? _________________ Not all those who wander are lost
to 1.) I don't know if you can make a generally valid statement about this.
But with dd-wrt the routers are usually much slower than with the stock firmware.
But there are for example Broadcom routers that support CTF+FA and with the acceleration the throughput is comparable to the stock firmware.
the acceleration has as said also its disadvantages
to 2.) you can probably not make a generally valid statement, it also depends somewhat on how you measure (WLAN <-> LAN or WLAN <-> WAN)
and it also depends on whether you are testing download or upload
Upload to the router (RX) needs less CPU than download from the router (TX)
for example I have:
600Mbit 2x2 TX 65% CPU load and 900Mbit 4x4 RX 60% CPU load
So my R7800 manages ~1GBit WLAN to LAN with ~60% CPU load.
Reversed (from LAN to WLAN), the CPU load is significantly higher and the processor is the limiting factor.
With an OpenWRT build that uses the NSS cores (network accelerator) it looks like in the attached pictures.
1Gbit WLAN download = 70% CPU load
1Gbit WLAN upload = 30% CPU load
As you can see WLAN itself needs some (alot of) CPU power (especially TX / download)
Therefore WAN --> WLAN should always be slower than WAN --> LAN
to title this general valid for all routers with a certain percentage is i think not possible
in conclusion, I would say that CPU power cannot be replaced by anything except perhaps by even more CPU power.
Joined: 30 May 2007 Posts: 94 Location: Arizona, USA
Posted: Sat Nov 13, 2021 20:27 Post subject:
I see, thank you for that explanation!
Also, I just read the "Hardware" wiki from following your link--very interesting summary/history!
ho1Aetoo wrote:
But with dd-wrt the routers are usually much slower than with the stock firmware.
But there are for example Broadcom routers that support CTF+FA and with the acceleration the throughput is comparable to the stock firmware.
the acceleration has as said also its disadvantages
I wish this info about the pros and cons was presented in either the stickies or in the installation wiki, which is the first wiki that any new(ish) user would read. It might make some users feel betrayed if they flashed dd-wrt for the first time and then find that they get slower speeds than on stock firmware (not counting those devices that support CTF+FA). I am okay with getting slower speeds in exchange for the better web GUI and more features (and maybe better stability?) but it's better if these tradeoffs are made clear front and center. Also, I wasted quite some time trying to figure out why I seem to be getting lower throughput numbers, which could've been avoided if I had known from the start that I can expect a speed drop-off. I see now that this is indirectly stated in the "Hardware" wiki but I don't think it's an obvious place to look. I never thought to check that page until you linked it.
ho1Aetoo wrote:
to 2.) you can probably not make a generally valid statement,
[...]
As you can see WLAN itself needs some (alot of) CPU power (especially TX / download)
Therefore WAN --> WLAN should always be slower than WAN --> LAN
to title this general valid for all routers with a certain percentage is i think not possible
Got it, thanks for the details! _________________ Not all those who wander are lost
