Hello.
Need some help with router.
Old HEX RB750GR3 - ETH 1 - WAN, ETH 5 - LAN , default config - speed from speed test ok - 1000Mbit/250Mbit
New HEX E50UG - ETH 1 - WAN, ETH 5 - LAN , default config - speed 150-220Mbit
What change or what is wrong with new HEX ?
I dont use any PPPoe etc.
Look at block diagram of Hex refresh and then look at your config if WAN really has to be on port 1.
You don’t give any more info so that’s the only thing I can respond from my side.
hEX refresh E50UG
RB750Gr3 Block Diagram with disabled switching
RB750Gr3 Block Diagram with enabled switching
Hex refresh E50UG clearly has a real problem with port 1. There are numerous other threads complaining about it and one doesn’t have to search very hard, ex:
http://forum.mikrotik.com/t/hex-refresh-download-speed/182770/20
http://forum.mikrotik.com/t/hex-refresh-slow-hw-offloading/182271/3
http://forum.mikrotik.com/t/hex-refresh-e50ug-router-for-gigabit-internet/180511/31
http://forum.mikrotik.com/t/rb951ui-2hnd-replacement-hex-refresh/183581/2
http://forum.mikrotik.com/t/hex-refresh-e50ug-router-for-gigabit-internet/180511/31
This Reddit post is also good reading:
https://www.reddit.com/r/mikrotik/comments/1hn5isz/brand_new_e50ug_hex_refresh_horrible_upload_and/
As it is affordable, I bought one myself to play with, and experience the same throughput problems many others are reporting with port 1. Various people keep replying with block diagrams as if they don’t very clearly show port 1 having a dedicated 1Gb/s line to the cpu and yet not reaching anywhere near that speed for many.
Whether MikroTik can/will do anything about it is the real question. One has to wonder if they even tested it properly seeing as the port has “internet” labeled right above it - implying that’s where the WAN connection should go.
“And uhm, I don’t know about you guys, but when I buy a router, I really want to use it as a router, not as a switch. But also, I really want to squeeze all the speed out of those ISP cables.”
https://youtu.be/Zrzq_zPWoQ4?si=7CANxHI36xN78M-T&t=172
“Depending on the config, we’ve seen performance even twice as fast”…. Meanwhile Szczepan’s RB750Gr3 has no problem apparently by comparison. Hmm.
thank you for your message, yes indeed ethernet port 1 of the new hex has problems, I also encountered this using it as a WAN,
I tried to create a ticket in support, but I no longer have enough strength and I do not understand what they want from me, I tried everything, and sent all possible support.rif: it took me 2 months of correspondence with them, where from the first month they just showed me a block diagram and asked to show the % of CPU load, this is completely pointless, since this is a clean default configuration and the % of CPU load is no more than 15-20%, now I checked on the following configuration, Mikrotik CHR on VM - PC - switch - hex, created a pppoe server and tried to connect to it, my idea was to exclude the possible influence of incompatibility of network cards for this I used an intermediate switch, and also to make the configuration cleaner and independent of the provider’s equipment, but I got exactly the same result, the same problem on port 1. the idea of ​​incompatibility does not fit into my practice, because if you use any port 2-5 except 1, then everything works fine. and the only difference is the different hex ports.
I also note a sharp overall drop in the speed of the new hex in any scenario where ipsec is used, approximately up to 250 Mbps of total bandwidth of the part without ipsec, which is probably a reflection of the processor speed, something similar but not so pronounced I observed on the old hex.
at the moment, I just decommissioned this new router and intend to sell it
I have neither the strength nor the desire to work with this device. of course, describing the problems above, I understand that this is a relatively cheap device and has good speed for its price. but I do not find alternatives, the old hex provided only 65-250 Mbps in my configuration, which was not enough. also for testing I bought a new AX2 and I had problems with VLAN and WiFi, so I do not see the point in buying a more expensive AX3. perhaps 5009 would be suitable for my configuration, but its price is unreasonably high, a PC hypervisor with an inexpensive VLAN switch gives a better result for a better price.
Right, double the peformance?? If you take it out of the box and hook it up per MikroTik’s own instructions here:
https://help.mikrotik.com/docs/spaces/QG/pages/144212023/Quick+Guide+-+hEX
“Connect the ISP cable to the Ethernet1 port”,
then it’s more like half or one quarter of the performance, using their own factory default configuration.
The new “refresh” model has its ether1 interface attached directly and not through the switch. I think this was done by the manufacturer because the main/usual application for this sort of chip uses this port as a WAN port, so it’s in a routed configuration. Also, I’ve seen variants from other manufacturers where this port has some other way of being attached, e.g. some have a 2.5GbE port (which MT’s accelerators can handle, but Mikrotik doesn’t use them) or often a DOCSIS modem or similar is attached to this port.
Even though it would be interesting to have a look at the issue in lab, I don’t have a device available to do that. What often helps in these situations is to use a proper interface queue for this port. For the other ports the switch’s internal memory handles buffering fine, however raw MACs usually have much less memory on board, so using only-hardware-queue for these is not optimal (or inadequate if you want to go there.) And I think this is the default configuration out of the box.
Could one of you with this device do a test with a queue type pfifo or mq_pfifo with about 500 packets as the limit on ether1? Does it improve the situation?
Thanks for answer.
One qestion - what to do with that problem ? ETH 1 is dedicated to WAN - ETH2,3,4,5 will be use in futer for other site.
Is any solution ?
Disable HW offload - nothing change
Disable Fasttrack - nothing change
Disable Firewall - nothing change
1 Like
As I said, add a large mq_pfifo or pfifo queue. Like this:
/queue/type/add name=eth1test kind=mq-pfifo pfifo-limit=500
/queue/interface/set [ find interface=ether1] queue=eth1test
Again, this is not something I have tested and would be a sure solution. But if you have the time, it’s something worth checking out. (If it does work, a size of 100 would probably be more appropriate.)
EDIT: Interface queues such as added above are compatible with fasttrack.
Have any MT staff responded to OR discussed the issues of ether1 on this model of hex?
Not to my knowledge, unfortunately.
However support has this as their first suggestion when a separate MAC is used, and has been the go-to solution for several models.
has any of you here emailed mikrotik support at all?
Not sure Normis, I know in another thread that is exactly what I recommended to the OP. ( The op had manipulated the mac address on ether1 which clouded the issue ).
http://forum.mikrotik.com/t/hex-refresh-download-speed/182770/19
http://forum.mikrotik.com/t/hex-refresh-download-speed/182770/19
and
http://forum.mikrotik.com/t/internet-speed/182962/1
I had a quick look, I only see one report dealing with PPPoE, so please, anyone that has any issues with this product with default config from LAN to WAN, make a supout.rif file, preferably during some speed test, and email support@mikrotik.com
OK. I prepare and send file to support @.
Normis,
Thank you for taking a direct interest in this. Please don’t misconstrue my earlier statements - I am a huge fan of MikroTik and want you to succeed, more than any other current network device company. But when a problem like this occurs and is reported across numerous platforms, continuously, it would be wise of your own design/engineering team to take some kind of notice.
If you take the time to read through the threads regarding this on your own forum here, or reddit, or anywhere else this problem is complained about, you’ll find people simply dumping the E50UG and going with a competitor’s product instead, which is unfortunate. If I can take your product right out of the box, which claims up to double the performance of the previous generation, plug it up in exactly the same way as the previous device using your own factory configuration and instructions, and it fails miserably, that’s bad for MikroTik, bad for your customers, bad for everyone…
If I cannot trust MT to get a simple 5-port router right, how can I ever trust an RDS2216, or any more “serious” product to also meet whatever claims are made about it?
Thank you again.
Tested on second E50UG - just buy - same 200Mbit only …
Mine happily pushes 540 mbps down, 440 mbps up through WAN with a QoS fq_codel queue tree. Cpu is not maxed - reaching like 80%.
That’s roughly in line with what I would expect.
The answer by @Szczepan doesn’t make it clear if they added a queue or not. Did you happen to test the device with only-hardware queues? Does yours give different results?
Edited my first post here to add additional relevant forum post links.