Poor wifi speed on new B1300 - would MediaTek be better?

#1

I received my GL.Inet GL-B1300 today. I first tested the latest firmware before flashing vanilla Openwrt. Unfortunately, I obtained what I consider to be fairly ordinary wireless speeds across the LAN with both firmware builds.

I was getting 31MB/s write and 36MB/s read with the standard 700MB file I use for testing between my laptop (with Intel AC7265) and NAS. In comparison, my R7000 and AC68U with dd-wrt offer at least 50/50MB/s.

Are my speeds typical of Qualcomm devices due to the opensource wireless drivers? Would a MediaTek device offer superior wireless speeds with Openwrt, and if so, what's a good option? Or is Broadcom always going to perform better due to the proprietry drivers?

#2

I have a potato router (mt7621at+ mt7603 + mt7612) and it got about 28MB/s but it doesn't have heatsink so I don't think it will better.

#3

The ipq40xx chips are mid-range and dual-stream only (so you only see them pairing at 866Mbps max theoretical). The higher end devices like those based on ipq80xx are 4-stream.

#4

As just one point of reference, I've seen 500 Mbps throughput using iperf3 (50-60 MB/s) with an IPQ4019-based device (Linksys EA3800) being ported on my bench and a MacBook Pro.

Am I correct in assuming you're looking at straight throughput through the router, not that you're reading or writing on the router itself?

Have you tested with the GL.iNet firmware?

MediaTek vs. Qualcomm is a debatable point, closer now that the open-source MTK drivers are, as I understand it, significantly better than they were a year or so ago. Broadcom wireless is a non-starter for open-source firmware. My experience with various Atheros/Qualcomm wireless over the past several years under OpenWrt has been very good.

#5

Yet most clients are just two stream; so a three or four stream router doesn't help there. Most laptops have two antennas at best. Cheap laptops even have one stream AC cards onboard (433 Mbps yay).

(I know streams and antennas don't necessarily correlate but in this basic comparison they do.)

#6

Thanks for the reply. Your speed of 50-60MB/s was using OpenWrt? I suppose I'm wondering if these speeds are obtainable on devices running OpenWrt or whether I need to adjust my expectations. It could be that the B1300 doesn't have especially good wireless compared to say, the EA6350v3 or R7800 (neither of which I've tried).

Yes, the tests were done copying between my laptop and my NAS, a few metres away from the B1300. The GL.iNet firmware obtained the same speeds as the OpenWrt stable and snapshot builds. Also, the speeds varied significantly. Copying the same 700MB file would sometimes give me 40MB/s in one direction and 25MB/s on subsequent tests. I tried different 5GHz channels, and they were all clear anyway.

I'd certainly like to make the move over to OpenWrt. However, I'm not willing to take a huge performance hit if dd-wrt is performing better on cheaper devices. I can easily return the B1300 to Amazon and grab a second hand EA6350v3 or R7800 if they're going to offer superior wireless speeds, but maybe they won't.

Any competition for TPLink Archer C7 in low cost space?
#7

Yes, I have been getting several hundred mbps on 5 GHz with the older generation Atheros chips in the TP-Link Archer C7 v2 units, likely limited by the single-core MIPS-based SoC in those older units. I haven't pushed either of the GL.iNet AR300M-Lite or AR750S hard, but I would expect similar performance from those MIPS-based devices as well. This is all under OpenWrt/LEDE, built of "master of the moment", with the Archer C7s running for around five years now.

The EA8300 (my typo) is IPQ4019 based and I'm in the process or responding to comment on the PR for it. I mentioned it has a similar wireless SoC to the B1300 and I've done a bunch of throughput and range testing to confirm that I had the cal files set up properly. Testing was done with iperf3 on the MacBook Pro and running on my Debian-based build machine with GigE switches and cabling between (no bonded interfaces).

#8

If you can obtain 50MB/s read and write, for example, on a slower Archer C7, then I'm thinking that the chip isn't the issue with the B1300.

I've read that the EA8500 can be very difficult to flash, so I probably wouldn't go with that option. The EA6350v3 is interesting, but doesn't seem to be that popular here, so support might not be great if I ran into issues. I'm thinking that the best option would be a D7800, as the R7800 isn't available here in Australia, and I don't know that I want to order one from overseas.

#9

By the way, is the Archer C7 on kernel 4.14 in the snapshot builds?

#10

Linux <redacted> 4.14.72 #0 Mon Feb 4 03:24:08 2019 mips GNU/Linux on the Archer C7v2 units I have in everyday service, yes (built from master locally).

#11

Yup. I was pointing out that if one gets > 500mbps out of a theoretical max 866mbps, then its not so bad.

I have a B1300, and running iperf3 in server mode on the router itself, with client running on a samsung phone, I was able to break 500mbps in short peaks about 2 meters away from router. Repeated tests show speeds around 430mbps or so.

In some runs I had irqbalance running, but I'm not quite certain if it makes a difference. Some people in other threads say it does, but I couldn't be sure. I also saw similar numbers in openwrt vs stock GL.inet (qsdk/qca wifi drivers).. but it could be my clients themselves.

1 Like
#12

At what rate did your notebook connect to the GL-B1300? What was the distance between AP and STA?

I need to replace one of my APs and was considering the GL-B1300, but since its quad-core CPU can't be an issue here it must be its WiFi range I guess. So have you conducted any further tests?

#13

I can't remember.

In close proximity.

No. I ended up returning it to Amazon. I wouldn't recommend it. I'd go with the Linksys EA6350v3, Asus RT-AC58U or TP-Link Archer C7 before the B1300. It was a very disappointing device.

#14

Please drop the ASUS RT-AC58U from that list, it doesn't have enough RAM for reliable operations (devices with two ath10k radios need more than 128 MB RAM).

#15

was both the laptop and NAS on wifi or the NAS was wired?

#16

I didn't know you could get NAS models with wireless. It was wired. The laptop was on wifi.

I've since switched to dd-wrt on my router (Netgear R7000) and dd-wrt on my WAP (Netgear R6400v2), and I get 50MB/s read and write from my laptop, and even faster using my mobile (Pocophone F1), so I'm happy.

#17

Thank you for sharing your final verdict.

Given its WAF, had you told me that the GL-B1300 shows a decent WiFi performance, I would have paid the higher price due to its four CPU cores and large memory, and given it additional tasks.

But I'll buy another Archer C7 v5 which is a very good managed VLAN AP + switch in my view. The EA6350 and RT-AC58U don't have the right profile for my use case (less WiFi, more CPU).

#18

EA6350v3 and rt-ac58u have significantly better wifi than the c7 (but the rt-ac58u should be avoided due to its RAM shortage, for the EA6350v3 it's elementary to get the v3 h/w revision) - for about the same amount of money as the c7.

#19

Could you define "significantly better" for me? Are you referring to beamforming, MU-MIMO, and not the bandwidth?

What is your opinion on the WiFi range of these devices, not considering beamforming?

Btw, the C7 is cheaper at my end, but YMMV of course.

#20

Have you tested and compared all three devices? In my experience, comparing specs and determining what should be better doesn't necessarily equate to what is better.

1 Like