[GCC 7.2 BUILD] Optimized TP-Link Archer C7 V2 AC1750 LEDE Firmware


#877

Idk how many times we’ve gone over this, @phinn just can’t read


#878

Thanks I can read. But if your internet connection didn't change and it went up by 20Mbits from that change I figured it still wasn't full saturating. Was just trying to show this router is capable of 800ish. My internet is capped at 150Mbits but regularly hit 165ish down. It just makes no sense that an overclock would give more bandwidth to me for something this low, it's not like we're pushing 500+ here.


#879

Those benchmarks on smallnetbuilder all use the stock firmware, and hence hardware NAT is enabled. Hardware NAT is unsupported by Openwrt/Lede and hence without fastpath 200-300mbit is all you'll be able to reliably get from the Archer C7. Fastpath could push that higher, but it won't be close to gigabit speeds since fastpath is still a software solution.


#880

Thanks for the explanation, that's really discouraging that we take such a massive performance hit going to OpenWrt. I thought FastPath solved that problem and was fully implemented everywhere at this point. Might just go back to stock firmware now.


#881

Fastpath was never upstreamed and has only been added to some custom builds by community members. So it is hardly implemented everywhere.

Flow offloading is a similar approach that has been upstreamed and is included in all recent master builds, regardless of architecture. While not as fast as HW NAT, it is much faster and comparable to fastpath speed wise.

Flow offloading also supports HW NAT, which currently only offloads IPv4 connections on mt7621 chips when enabled and should perform at maximum performance. It should support IPv6 soon and will also support other architectures as people add support, but this is a slow progress because each implementation is chip specific and there are many many different chips being utilized across all routers.

Stock hardware NAT will never be included in Openwrt/Lede, because the code is not opensourced.


#882

Okay understood. FastPath/SFE has been included in DD-WRT for sometime and I just (falsely) assumed it has been in OpenWrt/LEDE too. Sounds like with Flow Offloading coming then, hopefully in OpenWrt 2018, will be a big improvement. They did call it a "major release" on the main page. Looking forward to upgrading my Wrt3200ACM with it too.


#883

The software implementation is already in the snapshots / master branch and should perform similarly to fastpath. It can be used today. No need to wait for anything :slight_smile:


#884

I think this has been discussed before in this same thread, flow offloading is available on LEDE Kernel 4.14 and Archer C7 v2 is 4.9 based.


#885

Good point. I completely forgot about that. Then flow offload is indeed not available for the Archer C7.


#886

Hi. After use this firmware during a long time, I discover some trouble. I make some iperf3 test between this firmware and Vanilla and I found some differences. Router is configured as NAT with PPoE

Router b00t firmware --> IMAC LAN
[ ID] Interval           Transfer     Bandwidth
[  4]   0.00-10.00  sec   318 MBytes   267 Mbits/sec                  sender
[  4]   0.00-10.00  sec   317 MBytes   266 Mbits/sec                  receiver

Router b00t firmware --> IMAC WIFI 5Ghz
[ ID] Interval           Transfer     Bandwidth
[  4]   0.00-10.00  sec   103 MBytes  86.8 Mbits/sec                  sender
[  4]   0.00-10.00  sec   103 MBytes  86.3 Mbits/sec                  receiver

Router b00t firmware --> IMAC WIFI 2Ghz
[ ID] Interval           Transfer     Bandwidth
[  4]   0.00-10.00  sec  68.7 MBytes  57.6 Mbits/sec                  sender
[  4]   0.00-10.00  sec  68.1 MBytes  57.1 Mbits/sec                  receiver

Router VANILLA LEDE --> IMAC LAN
[ ID] Interval           Transfer     Bandwidth
[  4]   0.00-10.00  sec   364 MBytes   305 Mbits/sec                  sender
[  4]   0.00-10.00  sec   363 MBytes   305 Mbits/sec                  receiver

Router VANILLA LEDE --> IMAC WIFI 5Ghz
[ ID] Interval           Transfer     Bandwidth
[  4]   0.00-10.00  sec   219 MBytes   183 Mbits/sec                  sender
[  4]   0.00-10.00  sec   219 MBytes   183 Mbits/sec                  receiver

Why so difference?


#887

Same here!
@sycohexor4d builds have same problem.


#888

That's weird, I just ran a throughput test between 2 PC's in the same LAN (NO NAT TEST) and I got 650-700Mb/Sec. (ran 3 times, same results)
image
Ran a NAT throughput test, I got 600Mb/sec avg (the pic is from the last test):
image

If someone wants to test:
https://wiki.openwrt.org/doc/howto/benchmark.nat
And
https://www.tamos.com/products/throughput-test/


#889

Distance 2 feet NAT Enabled

Windows hardwired laptop to Android samsung s4 running iperf3 server on 5ghz

iperf

Windows hardwired laptop to Android samsung s4 running iperf3 server on 2.4ghz

iperf1


#890

wish i got better speeds thats for sure, guess i should just stick to the original firmware with just fast path and bbr patch.


#892

I dont know why, but throughput using vanilla or your build at 5Gb wifi is inestable.
Can you repeat you test using latest lede build?
Regards,


#893

what do you mean? just with the basic lede snapshot? or an updated snapshot with bbr patch and fast class?


#894

how did you disable nat? and what image are you using? Also whats your wifi performance look like, ill try iperf on another gigabit adapter.


#895

raspiiperf
windiperf


#896

A post was split to a new topic: SQM still beneficial?


#897

any ideas?