ePMP 4500L vs LTU

This is just a quick field test of a 4500L AP to 4525 SM link vs a LTU Rocket to LTU Pro link. Comparisions will not be 100% equal as LTU is fixed-frame only and ePMP 4500L is WLR (flexible) only. Both sets of equipment do support 1024 QAM.

Both used the same AP antenna (Ubnt AM5G-16) on the same tower and both SMs are installed on the same pole on the roof of our main building. Power levels should be similar except LTU will report higher RSSI as they use auto power which changes the power levels in the background as needed for the MCS rate, but only report the MCS 1 RSSI value.

Distance between AP and SM is around 600ft

Both APs were on basically the same channel but I forgot I had the LTU AP on 5840 while the 4500L is on 5835. Both are 20mhz and the channel is clear on both ends.

Both SMs are in NAT router mode. Also both SMs during the tests were connected to the same mikrotik hap AC2 for TCP performance testing.

In a clear channel both LTU and ePMP 4500L perform very similar with the LTU having a slight performance advantage most likely due to it being a more mature platform at the moment. I am not able to test interference rejection as to do that I would have to interfere with my own 5ghz equipment, but from what I have seen LTU is not very good with interference and is in most cases worse than AirmaxAC for fresnel issues and interference.

Performance photos will be in following posts.


ePMP 4525 SM

1 Like

LTU Performance. LTU internal speedtest is Duplex test vs single directional for the ePMP Link Test.

1 Like

If you’re comparing PtP links, then you should try using ePTP mode on the 4500, it should provide much better performance over WLR mode, which is intended for PtMP.

Interesting tests though. I’m still surprised that you don’t see any difference between using 1 connection and 5-10 connections.

I’ve heard from quite a few operators that LTU is very sensitive to noise and Ubiquiti’s AC platform is far more forgiving. We’ve been testing a 4500 8x8 AP at an extremely noisy tower site with many 5GHz operators and so far it’s worked amazingly well. Quite some time ago we pulled down all our old 5GHz equipment and converted to 3GHz due to so many noise issues at this site.


My intent is to not compare PtP links. The LTU while it only has 1 CPE is running in AP mode. You have to use a whole different FW to put LTU in PtP mode. Also the LTU was my test envirment where I could test settings and new FW without impacting any clients. Right now I have the 4500L setup in that capacity. This is for me to test PtMP with the ePMP and new FW before I deploy it to a customer site. Once I get a 4500 AP I do plan to put that on a tower and connect real clients to it.

When you use the Mtik bandwidth test without a defined number of connections it defaults to using 20 connections. You can verify that by doing a Torch on the Btest Server side. For some reason on the Client side Torch will only show one connection.

LTU is very sensitive to noise and that is why I really did not deploy it for PtMP. I am using it for PtP links though as they tend to have less noise than my PtMP clients.

For example I have a short range (3mi) PtP link using LTU-Pro to LTU-LR. It is running with GPS sync and 75/25 fixed-framing. Total cost of the link is $309 and I get 182/45Mbps out of the link.


The real beauty of the 4500 is yet to come as the firmware matures - 8x8 MU-MIMO. We see remarkable efficiency on some of our 3000 APs, and the 4000 series will be even more remarkable, once the firmware features are fully implemented. :crossed_fingers:

1 Like

Note the 3000 series took 3 years and still MIMO for 80Mhz is not working at full power.

Well I was just saying that MU-MIMO is one of the key differentiators between some of the technologies. Comparing speeds on a single SM is one thing (and it is obviously appropriate for a PTP link) but on a PtMP system, MU-MIMO can dramatically tilt the balance. You can potentially get 2x - 3x - 4x throughput for ‘free’ spectrum, and the more heavily loaded the AP is, the more beneficial MU-MIMO becomes

So, each to their own experiences. With our 3000 we often average 50% - 75% extra “free” throughout.

Wanted to add some info and clarify a bit.

LTU does fall apart in noise, badly. I’ve done extensive testing of AF5xHD (LTU PTP) vs 425 and 400C. They perform similarly in perfectly clean channels but as the noise increases, the LTU product degrades rapidly. Also, LTU doesn’t scale up in channel width well no matter how clean the environment is while 400/425 is nearly linear scaling. My primary test rig is 11.2km away and the LTU w/ 30dB dish does 300Mbps aggragate at 240x60 and cannot get past 60 on the uplink side because of noise. 425+range extenders does 400Mbps aggregate on the same channel but can do ~300 on the uplink where LTU falls down.

Also, 425 ptp links are tollerant of fresnel issues, LTU is not.

LTU AP to LTU CPE with a single client actually does a different set of timings that are more like a PTP shot. Ie, that 1:1 scenario doesn’t reflect PtMP. You need to hook up a second LTU client to get into the PtMP scheduler.

That said, I dont have 4500 to test in PtMP mode. I have Eric’s posts on performance to compare another WiFi6 product to which lines up similarly and have done side-by-side against LTU and there’s no comparison at all.

Also note that if a competitor is on LTU and you are on WiFi6 4500 etc on the same channels, your WiFi6 radio will certainly hold up better and their LTU will fall on it’s face. You will be their destructive interference.