Anyone tested the new L009?

Anyone tested the new L009? Or compared it to a Hex?
https://mikrotik.com/product/l009uigs_rm
https://mikrotik.com/product/RB750Gr3

The Hex have 2 core at 880Mhz, the L009 have 2 core at 800Mhz
When looking at the test result only with no fw rules it is faster.

I can see it have some nice features compared to the Hex, but is it faster?

Price is the double, so i get two Hex for one L009.
If i want a fast one it is the RB5009 but the price on the L009 is better for me…

I know all the compare is against the RB2011.

(edit because of error)

hEX does NOT have 4 cores, really are 2 cores presented as 4 Threads

but according to Published Test Results hEX can be somewhat faster than L009 in CPU demanding scenarios

L009 is a very new device, performance test results can improve because of optimizations

Advantage of L009 in memory, storage and other features makes it much more versatile than hEX

i think If you are considering hEX you must not be comparing it with L009

in terms of performance and form factor a more relevant comparison will be with hAP ac² or hAP ax lite or even hAP ax² ignoring the wireless section

L009 is only nice looking switch. CPU is garbitch.

i had the same misconception when L009 was anounced

The purpose of L009 is to provide that features at lowest possible cost, is not focused on performance, if you need more performance there already other devices for that

Key features of L009:

RouterOS license 5
Size of RAM 512 MB
Storage size 128 MB
Number of DC inputs 2 (DC jack, PoE-IN)
DC jack input Voltage 24-56 V
PoE in 802.3af/at
PoE in input Voltage 24-56 V
PoE-out ports Ether8
PoE out Passive PoE
SFP ports 1 (2.5G supported)
Serial console port RJ45
Number of USB ports 1
USB Power Reset Yes
USB slot type USB 3.0 type A
Max USB current (A) 1.5
Mode button Yes

This make it a powerful laboratory/testing device, and more important, make it a very viable drop in replacement for sites where rb2011 currently is deployed in providing almost the same features and performance, at similar cost

as i said it is only switch…who cares about RAM etc. when CPU is garbitch

it is 1000 times better to buy AX2 than this garbitch

AX2 is not a bad recommendation but I prefer the ax3 for best bang for buck.
Check out the throughput!

Routing 512 byte 25 filter rules.

ax2- 912.9
ax3- 1145mbps ( great for a 1gig ISP connection - very common )

is a interesting option, if budget is available, around 33% improvement in performance in CPU intensive test for 40% more in price
ax2-vs-ax3.png

It depends on your needs, as explained by chechito.
If you need more then 5 ether ports, you can’t do much with ax2 or you would have to add another switch.
USB, SFP ? ditto.
If MT says 2011 is one of their top sellers, it makes perfect sense to provide L009 as drop-in replacement (and I’m sure they will have some cost optimalisations along the line with the case etc.).

Similar discussion when AX Lite was released. It has it’s purpose.
I use that one very frequently since I have it for situations where I before used to go with mAP/mAP Lite. And more ether ports too.
I would have liked it’s package to be smaller (more hAP or Hex formwise) but it is what it is.

in most scenarios you need router and switch. So you have router AX2 or AX3 and switch L009.

I’m running an L009AX for a couple of months now, and it’s quite capable. The WiFi sees very little use, and mostly the ETH is working. It sits in the basement of our building and is routing traffic with two ISP uplinks, routing and bridging to several VLANs for Video Surveillance, IoT, home & guest nets, and has several severs; AC2s & an AX3 for my home attached to it. It also runs WireGuard and a PyHole container. And frankly, copes well with all this, without getting too high of CPU usage.

Compared to the AX2/3 and the AC2/3, L009 has HW offloaded switching on the bridge with VLAN filtering, and this is an advantage for me. The IPQ switch in the AX series is yet to get a driver (if at all) to support HW offloading for VLAN filtering, so the filtering is done by the AX3’s CPU, and not the switch chip, as is L009. The AX3 is very powerful, and gracefully handles a gigabit of traffic with up to 20-30-40% CPU load, but not enough ports for my use case and the software based VLAN is not ideal IMO.

I still don’t have a good use of the 2.5g port on the L009, as it’s an SFP, but if I run out of eth ports, I may purchase a transceiver to connect the AX3 (or the NAS!) over 2.5gb.

After setting it all up, I think the RB5009 would have been a better choice for me, due to the 2.5g port (for the AX3 link), and the 10g sfp, that I may use for attaching the NAS. But that’s more cash, with a questionable benefit, as the end-point network will remain 1Gbit with a slightly faster backbone. It is more to my RGAS (rapid gear acquisition syndrome) and appetite/greed to want more, than actually needing and benefiting from it. The L009 will work great for me for years to come. An RB5009 would improve my net a little bit, but is the more € justified, I’m not sure.

All in all, the L009 is very capable and practical. And it’s beautiful and of very good quality, IMO. As an owner, I recommend it.

Garbitch?
:laughing:
Do you mean garbage!?

Urban slang but the word does exist … Google it.

jargon voor afval,


Sorry Loop, disagree!
The 1009 2.5 port is a mystery to me as its real world WAN throughput is 300-400Mbps whereas the old hex will get you 400-500 Mbps.
Both have two cores…
The AX3 will get you over 1Gbps and has 4 cores and double the RAM of the L1009, its no contest, and yes the AX3 also has a 2.5g port, probably
to attach to upstream fiber modems but again, if your WAN throughput is not going to optimize that, strange design.

Now to the bogus statement:
Compared to the AX2/3 and the AC2/3, L009 has HW offloaded switching on the bridge with VLAN filtering,

Not true, the AX3, has HW offload…

truth.jpg

I think you are using the wrong column in your screenshot. The “Hardware Offload” column shows the value of the user-selectable setting checkbox. You have to make the “Hw. Offload” column visible instead, to see the real actual hardware acceleration state of the port:

hwoffload.png
Even without that column, you can see in your screenshot that there is no “H” flag on the 2nd column, compared to my screenshot.

If we look at this page https://help.mikrotik.com/docs/display/ROS/Switch+Chip+Features, we’ll see that the hAP ax³ has the Switch-chip classified as “IPQ-PPE” for the 5 ethernet ports. And if we look at the Bridge Hardware Offloading table https://help.mikrotik.com/docs/display/ROS/Bridging+and+Switching#BridgingandSwitching-BridgeHardwareOffloading we’ll see that IPQ-PPE has no hardware supports for Bridge VLAN Filtering:

bridge-hw.png
The number 6 in the table even has this remark attached to it:


Currently, HW offloaded bridge support for the IPQ-PPE switch chip is still a work in progress. We recommend using, the default, non-HW offloaded bridge (enabled RSTP).

So @loop has been right with his “software based VLAN” comment..

I stand corrected! Thanks for that and I owe Loop an apology!!
Interestingly I have no problems with CPU usage or performance to date.

As for the SFP port, if we keep ether1 out of the main bridge (https://i.mt.lv/cdn/product_files/L009UiGS-RM_230555.png), the bridge with the rest of the ports can work wonderfully as a smart switch, with hardware acceleration for VLAN filtering, DHCP snooping, IGMP snooping, RSTP, etc… And if you connect the SFP port to a 2.5Gbps-capable NAS for instance, the NAS will be able to simultaneously serve 2-3 clients connected to the GbE ports and saturate at least two of the ports, the CPU is not needed at all. Not everything needs to go to the WAN side and uses the CPU.

I think the CPU is still powerful enough and probably has no problems moving data between two ports at wire speed, even with all the extra VLAN tagging/untagging/matching calculations that it has to make. It might still be able to do that with 4 ports (two pairs of ports simultaneously), or not. But the L009 can switch with VLAN support on 8 ports. The table here https://mikrotik.com/product/l009uigs_rm#fndtn-testresults shows lower numbers, but it’s the table for the IPQ-5018 CPU, not the separate 88E6190 switch chip.

you feel the same after few months? do ax2/3’s are managed locally or by capsman? I got one unit, eth only, and consider jumping from hex to it. USB 3.0 ARM and 8 ports are tempting, but after reading some reviews I consider returning it and getting RB5009UG+S+IN, I still have few more days to decide.