I have tested and tested and tested, and see no difference between this new firmware and the old firmware.
The Balance One product has 99% CPU utilization at 350-400Mbps of download, and around 250-300Mbps of upload. I can’t even get the UI to respond appropriately, and VPN connections continue to be disconnected during this time.
I bought this product because it was advertised as capable of gigabit throughput. I am not using the AP on the product, but an AP One AC Mini because it was recommended to get the throughput.
I have only one wired machine on the Balance One, so nothing else is eating up CPU cycles. In fact, I’ve tested with three different machines directly wired, and I have the same results.
It clearly seems like this product cannot do a gigabit throughput in all conditions, and is severely limited in this particular condition.
Sorry for the delayed reply, Steve, and thanks for the information.
Firmware 6.2.1s022 has quite significant improvement for me and I can get > 900 Mbps DL (compared to ~ 800 Mbps using RC) in my environment (a 4 years old Asus motherboard running Windows 8.1).
What is the model year of your MacBook Pro and Mac Mini? I hope we can reproduce this in our testing lab to get a closer look, I’ll try to get one with the exact same model as you to do the test.
Hi Steve, could you please help to run the following command on your MacBook Pro, then test again the throughput with Balance One (6.2.1s022) and let us know the result?
Open Spotlight, search “terminal”, and run it.
In the Terminal, type:
sudo sysctl -w net.inet.tcp.delayed_ack=3
then press Enter. Please copy the output and share with us here. The above configuration will be restored to system default after you reboot the MBP, so don’t worry about it.
After running the command above, test the throughput again. Thank you very much for your help.
The MacBook Pro is late 2013, the Mac Mini is 2014. Both very modern.
And AGAIN to be clear - two other competitors products work perfectly fine. No changes required on the MBP or Mini. Clearly something with your TCP optimizations on the wan side, detection of MTU, or something similar is causing the congestion. Or the code is just simply not optimized, and the CPU is being maxed out.
I did adjust the TCP delayed ACK per your request, and I was able to get around 600Mbps download, 200Mbps upload. So it increased the download performance (still short by about 150-200Mbps compared to competitors) and reduced upload by 100Mbps. The latter is sometimes expected when adjusting this sort of TCP behavior.
Still saw 100% CPU and all PepVPN and PPTP VPN connections were immediately disconnected. UI unreachable, etc. No other competitors have this behavior - perhaps you should consider setting aside 5% CPU for management and user processes? Not sure…
(I have no dog in this race, but I just wanted to comment that I really like the quality of support that Peplink is offering here - being straightforward about issues, suggesting beta builds of software that may fix it, etc. Nice to see, and I hope they find a fix! )
We have tested Apple MacBook Pro / Air (Directly connected, or with Thunderbolt-to-Ethernet adapter) that manufactured in different years in our testing lab. Unfotunately looks like we have overlooked this part and the best download throughput we got is only about 700 Mbps, some of the older MacBook Pro is even lower than that. Compared to other Windows 7/8 PC in the testing lab, we can get 900 Mbps DL via Balance One with firmware 6.2.1s022. This performance difference is quite disappointing (even for us, the engineers), and we are still trying to find out the root cause.
In the mean time, we will revise our marketing message, will not claim it to be a 1Gbps router, but setting the expected throughput to be 600 Mbps, until we can figure out the problem and get it fixed.
Sorry for the inconvenience caused, I’ll of course keep this forum post updated if we get any breakthrough about this matter.
Thank you so much for keeping me and the rest of the forum updated. I think I can agree with the other folks who have posted not only on this thread, but many others, about the transparency and willingness Peplink and your teams have to working on issues and updating the community. It is much appreciated, and is a true competitive differentiator.
I’m obviously disappointed in the performance as well, but happy that you guys are still looking into it, and have identified that it’s not just my setup! I’ll keep testing myself and see if there is anything I can learn from Apple or others as it seems like something they’re doing with their TCP stack.
I recently had AT&T Fiber installed and am having the same speed issue. I am testing with a brand new Dell E7450. When connected through the Peplink One, I average 480 down / 280 up. When connected directly to AT&T’s gateway, I average 910 down / 900 up. I have not yet tried firmware 6.2.1s022, but that is next on the list.
Installing firmware 6.2.1s022 had no noticeable difference in performance with my existing configuration (480/280).
Installing firmware 6.2.1s022 with a factory reset increased performance to 510/440.
Enabling IPv6 support decreased performance to 420/420.
Running an IPv6 speedtest through Comcast nets me a 910/640, as my understanding is the Peplink simply passes through the IPv6 traffic.
This is specifically for LAN-to-WAN. On the one hand I would have expected that to be simpler as there is usually less inspection on outgoing traffic. But clearly something is going on to cause that limitation if its able to handle incoming traffic at 600Mb rates.
My main site has a gigabit ethernet connection, and I was the original reporter of the issue. I still regularly top out at 300Mbps download and upload from a 1Gbps ethernet direct attached Mac Mini, and via WiFi from several other Macs. If I remove the Balance One completely, I’m able to get almost 900Mbps bidirectionally through the connection. I think it was traced down to TCP behavior and potentially MTUs on top of this, and doing the conversion from one side to the other.
We have opened a ticket with support but we have 2 balance 580’s in the field recently defaulted and running firmware 6.3.3 and 7.0.0
they are experiencing the same problem where on a gigbit circuit they are only achieving 450 mmbps download and 990 MBPS upload.
We are using a cat6 patch chord, and these are from two different ISP. When using other devices we receive the full circuit. These devices are also spiking 100% while performing test. We have tested beoth devices with both LAN users (Several) and Lan users(just 1) and achieve same results.
We really need someone with peplink to look into this issue as the 580 is rated for 1.5 throughput and 1000 users and we are struggling.
Was this ever resolved? We are having the same issue… two 1gb Wan connections and not the same speeds from WAN to LAN on peplink. But directly to the WAN connection we get the speed as advertised…
