Let's talk about a successor of the RB4011(+WiFi)

- The AL21400 chipset is powerful. You can actually keep it. Perhaps there's a new hardware revision available with higher clock and/or less power.
- passive cooling: keep it.
- ports: I like the SFP+, but would it possible to add a RJ45 port with "Multi-Gig" 1/2.5/5 Gbit/s capabilities with support of 802.3at/802.3bt or passive PoE for powering it?
- Wireless LAN: Well, Reinis is currently working on MU-MIMO support. If he adds Uplink MU-MIMO aswell we have the first step done for IEEE 802.11ax support.
- antennas: 8x for 5 GHz, 4x for 2,4 GHz, as this makes sense with MU-MIMO
- 802.11ax + Tri-Radio support: You can either use 1x 5GHz with all your 8 antennas or alternatively split 5 GHz in two parts so you can use lower frequencies with 4 antennas on 5 GHz block A and the other 4 antennas on 5 GHz block B.
- OFDMA is great but needs time to implement. So support it in a future release
- pricing: 349$

Let's talk about a successor of the RB4011(+WiFi)

- The AL21400 chipset is powerful. You can actually keep it. Perhaps there's a new hardware revision available with higher clock and/or less power.
- passive cooling: keep it.
- ports: I like the SFP+, but would it possible to add a RJ45 port with "Multi-Gig" 1/2.5/5 Gbit/s capabilities with support of 802.3at/802.3bt or passive PoE for powering it?
- Wireless LAN: Well, Reinis is currently working on MU-MIMO support. If he adds Uplink MU-MIMO aswell we have the first step done for IEEE 802.11ax support.
- antennas: 8x for 5 GHz, 4x for 2,4 GHz, as this makes sense with MU-MIMO
- 802.11ax + Tri-Radio support: You can either use 1x 5GHz with all your 8 antennas or alternatively split 5 GHz in two parts so you can use lower frequencies with 4 antennas on 5 GHz block A and the other 4 antennas on 5 GHz block B.
- OFDMA is great but needs time to implement. So support it in a future release
- pricing: 349$

I take, Wrap one please.

i would buy one

make it more about proper implementation of existing features that needs to be fixed in hardware:

• SFP+ slot that's not picky and supports SFP/SFP+/passive DAC/GPON modules without any issues
• Switch chip that supports at least 8 ports at 0.01/0.1/1/2.5/5Gbit/s with hardware VLAN filtering and other features you would expect from $300+ router
• Extra miniPCI-E x1 slot for modems,LoRA,etc.
• USB ports, CPU have quite a few of them, so let's have some available to user
• metal (alu) case, that spreads the heat from cpu and SFP+ cage. For $300+ I'd expect metal case with proper heat dissipation design, not some overheating cheap plastic box
• 802.11ax would be nice, but based on current state of WAVE2/MU-MIMO, ax will be realistically fully working on Mikrotik somewhere around year 2021/2022, even if hardware is available today...
• One could dream: all features actually working on a first day of sale

As a software developer I would rather prefer continuous delivery of new features. Let Mikrotik focus on rock-solid basic feature sets and enabling step by step more sophisticated features.

And please, bring back the variants with touch LCD displays (yes, I actually use the LCD).
I avoid upgrading all my 3011s just because there is nothing newer with LCD.

Extra miniPCI-E x1 slot for modems,LoRA,etc.

+1
this is tempting features, espeically with the new LTE6 mpic-e

make it more about proper implementation of existing features that needs to be fixed in hardware:

• SFP+ slot that's not picky and supports SFP/SFP+/passive DAC/GPON modules without any issues
• Switch chip that supports at least 8 ports at 0.01/0.1/1/2.5/5Gbit/s with hardware VLAN filtering and other features you would expect from $300+ router
• Extra miniPCI-E x1 slot for modems,LoRA,etc.
• USB ports, CPU have quite a few of them, so let's have some available to user
• metal (alu) case, that spreads the heat from cpu and SFP+ cage. For $300+ I'd expect metal case with proper heat dissipation design, not some overheating cheap plastic box
• 802.11ax would be nice, but based on current state of WAVE2/MU-MIMO, ax will be realistically fully working on Mikrotik somewhere around year 2021/2022, even if hardware is available today...
• One could dream: all features actually working on a first day of sale

+10,000
@r00t -- -- -- -- -- -- -- -- -- AMEN -- a true Dream Machine !!!

+6Ghz (Wifi 6e) support

Just no.

If you need powerful WiFi, get a proper AP and place it properly.
If you need powerful switch or router, get a proper switch or router.

If you don't care but like fancy numbers and lot of antennas, get a Netgear or D-Link.

+2
Make two case options, a proper rack-mount, and a nice desktop version.

this is tempting features, espeically with the new LTE6 mpic-e

I am looking for a RB4011s sick performance, but with a built in 4G/5G modem. This would be awesome to do a selfbuilt SD-wan solution with dynamic ipsec tunnels....
+1 for LTE /mini pci-e slot with simcard support!
Then I could build my own "sdwan" solution out of just one single high performance box.

I would add a secondary SFP+ and a switch chip to the specs already mentioned


Sent from my iPhone using Tapatalk

a new 64bit ARM with lower heat output would be my #1 requested feature, closely followed by a better switch chipset e.g. Marvell Linkstreet series (the Realtek's suck)

i dont need WiFi or many RJ45.
What i need is a powerfull router with min. 2x SFP+

let the router be a router and nothing more

i dont need WiFi or many RJ45.
What i need is a powerfull router with min. 2x SFP+

let the router be a router and nothing more

Agreed. There is room between the RB4011iGS+RM and the CCR2004-1G-12S+2XS for a router with a small number of 10G interfaces. Something like a CCR version of the CRS305.

I agree that something is missing between the RB4011 and CCR2004. The RB4011 is a bit more tilted against Home-usage. CCR2004 is a datacenter device.

Or, your could make two variants of the same device. One with most Ethernets and DC-jack, one with mostly SFPs and telecom-power.

Devices with everything built in tends to be not really good on anything.

I agree that something is missing between the RB4011 and CCR2004. The RB4011 is a bit more tilted against Home-usage. CCR2004 is a datacenter device.

I've been thinking about the MT routing & switching lineup in general. IMHO, it's due for a cleanup.
What I would do, were I MikroTik:

1) Ditch the RB2011 line. Probably the RB3011 as well, although lowering the price to RB2011 levels is also an option.
2) Build a RB4011v2 with USB port and better switch chips.
3) Ditch the RB1100. For a Dude Edition people can use an USB stick in a RB4011v2
4) Build a CCR2004-8G-4S+ for something like $400,=
5) Ditch the CCR1009s, CCR1016s and CCR1036-12G
6) Build a CCR2016 (which they probably are going to do, from the info we have)
7) Ditch the rest of the TILE boxes. ARM64 is where it's at.

Idem switching:

Lose the CRS1xx line & RB260s. Concentrate on the CRS3xx line and make a low end CRS305-5G or CRS308-7G-1C or something like that.

They need to focus and concentrate on ARM64 & PresteraDX. EOL the rest.

Yes, the RB1100:s are strange, no SFP/Combo:s at all but -48V supply. And RB2011 is just old.

However, I can see understand why the RB4011 has such a dumb switch-chip. Is not supposed to be used as a switch, only as a port-expander. But then the device is somewhat unsuitable for SoHo-usage.

The CCR1009-7G-1C-1S+ is very versatile so I hope that remains in some form.

However, I can see understand why the RB4011 has such a dumb switch-chip. Is not supposed to be used as a switch, only as a port-expander. But then the device is somewhat unsuitable for SoHo-usage.

Obviously Mikrotik is placing RB4011 in SoHo segment ... variant with WiFi is listed under wireless for home and office and in such environment it's customary to live with all-in-one device (some people try to use CRS line of devices as all-in-one and get bitten by their shitty routing performance). But then it seems MT doesn't consider VLANs to be SoHo feature as the built-in switch chip is just fine as long as one doesn't use VLANs.

However, I can see understand why the RB4011 has such a dumb switch-chip. Is not supposed to be used as a switch, only as a port-expander. But then the device is somewhat unsuitable for SoHo-usage.

Exactly. It would be better at what it is if it had a somewhat more capable switch chip, one capable of VLAN handling in hardware at least.

The CCR1009-7G-1C-1S+ is very versatile so I hope that remains in some form.

Why? A CCR2004-8G-4S+ would be as versatile (remember that SFPs work in SFP+ cages), more powerful, and roughly the same price. Nope, TILE is dead.

But then it seems MT doesn't consider VLANs to be SoHo feature as the built-in switch chip is just fine as long as one doesn't use VLANs.

True, it seems that way. But I beg to differ, and probably a lot of SOHO networkers along with me. The 4011 is an awesome little box for the price (I even bought a WiFi one because is was the cheapest mGig-capable AP I could find!), but with a real switch chip it would be better still. And a USB port for storage or LTE backup.

Speaking for myself as a mid-power-user with RB4011 bought as an all-in-one for home, keep in mind the CPU is really beefy.
You can and should do vlans and stuff in software, no problem. It's meant to handle that quite easily up to reasonable home / small office levels.

WAN-to-LAN and WIFI-to-wired communication always goes through the CPU anyway, and in such home situations most users are interested about "how fast they get their internet" more than device-to-device transfers onsite.

Personally, with stuff like queues enabled, so limited fasttrack (I used a trick posted here on the forum to partially fasttrack the biggest group of "default" clients not needing a packet mark added). It runs my home, homelab, and 500Mbps uplink, and stays under 15% CPU most of the time.

A VLAN-enabled switch chip (well, more accurately, two-layer-VLAN-enabled so the upper layer is available to your use) would sure help if you have a big internal network that you'd like to see switched at line speed; that's definitely not the intended use case for the 4011.
But I guess you could pair it with a CRS via the SFP+ port, and let the CRS handle in-vlan traffic.

That said, I do agree with most of the recent posts above regarding clean-up of the product lines. Y'all have good ideas.

edit: One day I'll learn to proofread before posting...

+1.

I'm relatively new to MikroTik and the RB4011 was my first MT device. Before reading this thread, I was not aware it indeed lost some features compared to the previous 20xx/30xx boxes. So my two cents might be a bit different.

We run several branch offices in different sized and started to use RB4011 as branch routers. In larger branch offices, we have HPE/Cisco core switches and use the RB4011 as router-on-a-stick with a single 10G/SFP+ connection. The RB4011 has a lot of CPU power and the 10G SFP+ is directly connected to the SoC/CPU. In our experience, it routes >7GBit/s between SFP+ VLANs with usual FW rules and NAT. Without NAT, it constantly delivers full 10GB wire speed routing. It is by far the cheapest device I know about capable of doing this.

In small branch offices, we use cheap GS1200/GS1900 1G "smart switches" doing all the VLAN tagging/trunking/switching and providing active PoE for the WiFi APs and passive PoE for the RB4011. On the RB4011, no VLANs are used. All routing goes through 2-4 untagged individual physical ethernet ports patched to the switch with parallel short cables. This approach makes configuration and debugging easier and everything runs from the switch power supply reducing cable clutter. The RB4011 and the switches are cheap enough to allow stocking ready to run preconfigured replacement devices.

The convenient/safe location for a branch router is usually bad radio-wise for WiFi APs. Even in my home network, the RB4011 is in the basement were the ISP fiber enters the house. There is also a fast pace in WiFi evolution, so integrated WiFi tends to get outdated faster than the router part. In my conclusion, I don’t see a need for integrated WiFi.

The VLAN feature of the Realtek Switch chips is obviously used to expose all 10 ports as individual interfaces to ROS. To add HW based user configurable VLAN switching on top, a switch chip providing VLAN stacking would be required. I see this would be beneficial for power SOHO users. But I also think this is not a big market, as most SOHO users are not capable of configuring "real routers" and are better off with consumer boxes from one if the usual suspects. Most SOHOs applications demand high WAN/LAN (going through the CPU anyway) transfer rates much more than high bandwidth/low latency between LAN hosts.

If full VLAN switching is added to a post RB4011 device, there should also be proper active PoE support on all Ethernet ports. Nowadays, there is PoE in use for most WiFi APs and Voip phones also in small office setups. So an external switch is required anyway to avoid a battery of clumsy PoE injectors.

What are your thoughts on these design ideas ?

Design 1



Design 2


These designs bring the following improvements:

Switching
- Hardware based QOS
- Hardware ACL's
- Hardware based Ingress rate Limiting (Storm Protection)
- Hardware based egress rate shaping
- TCAM allows for L3 Cut-Through/CPU Offload (FastPath)

General
- Higher throughput than existing platforms (hEX PoE Pro/hEX S/RB1100AHx4/RB4011)
- Lower heat output than existing platforms (RB1100AHx4/RB4011)
- 64bit ARM Architecture
- Makes use of SOC's that Mikrotik are already using in other products (nRay)

I also think the hAP ac2 and cAP AC should be upgraded to the 88F3720 or IPQ6xxx for more performance/less heat

I like your Design 2, would buy it, if price was right...
Only thing I would add is at least one mini-pcie slot, it's about $1 in BOM cost and it would make it much more versatile (could add WIFI, LTE, LoRA, ...)
Probably worth even for cheaper Design 1, cost is minimal... and then it can easily be sold with wifi or LTE card, without any design/PCB changes.

3) Ditch the RB1100. For a Dude Edition people can use an USB stick in a RB4011v2

I have been experimenting with Dude server on various routerboards, both MicroSDHC (fastest type) and very fast USB3 sticks, and it does not work.
Bottleneck appears to be read/write to Dude DB, performance is terrible with MicroSD and USB3.
The only solution I have managed to get to work is the RB1100AHX4 Dude with MSata SSD.

We have much improved disk read/write speeds in the new v7 beta that will be coming out soon.

We have much improved disk read/write speeds in the new v7 beta that will be coming out soon.

Thank you Normis,
You expect that it will be possible to run Dude DB on MicroSDHC or USB3 under ROS7? Has this been tested and confirmed?

That said, after swapping the MSata SSD that came with my RB1100AHx4 for a new one (It broke after a few weeks), operation of Dude on this HW has worked flawlessly for years now

In addition to above
- SFP+ports x 2 that can accommodate all speeds / sync so that ISP SFP modules can sync at 2.5Gbps to give full speeds to allow traffic in and route out to a 10G switch Support 10G/5G/2.5G/1G syncing and all SFP port types. I’m only considering RB4011 now because GPON modules were just removed from exemption and passive DAC modules still remain
- multiple 10G / 5G / 2.5G / 1G copper ports not just 1G ports as you want to get the 10G out of the router and not these combo ports - dedicated 2x SFP+ ports
- switch chip that supports VLAN HW offloading
- bridging (if possible) and PPPoE and other hardware offloading as Ubiquiti Edgerouter can do most HW offloading except bridging
- direct port to CPU If possible instead of switch port of combination of switch chip and directly wired ports like Edgerouter 12/12P
- Better SoC that is not limited or compromising and bring features of various CPU and switchchip features together
- desktop with proper rackears (for home users) and rack mount version
- OpenVPN UDP and Wireguard HW offloading along with good ole IPSec
- support for PoE standards such as 802af/at to support UniFi devices (same goes for a Edgerouter devices on my wish list of options as they only support 24v PoE) on all ports

What are your thoughts on these design ideas ?

These designs bring the following improvements:

Switching
- Hardware based QOS
- Hardware ACL's
- Hardware based Ingress rate Limiting (Storm Protection)
- Hardware based egress rate shaping
- TCAM allows for L3 Cut-Through/CPU Offload (FastPath)

General
- Higher throughput than existing platforms (hEX PoE Pro/hEX S/RB1100AHx4/RB4011)
- Lower heat output than existing platforms (RB1100AHx4/RB4011)
- 64bit ARM Architecture
- Makes use of SOC's that Mikrotik are already using in other products (nRay)

I also think the hAP ac2 and cAP AC should be upgraded to the 88F3720 or IPQ6xxx for more performance/less heat

These are good but more switch ports and faster 4 core cpu.

I think Design 1 is quite limited for nowadays - only 1G ports won't be very interesting, when almost every enthusiast/prosumer is lookig at >1Gbit. However, with correct agressive pricing, it might be still a great lower-mid range router. You could also drop the memory and flash to lower values to achieve better price. 128MB flash should be enough for any basic usage including partitioning. RAM seems also too luxurious for lowcost 8 port router

Design2 is excellent and with a right cost, I would grab bunch of them instantly. Personally, I would prefer two SFP+ ports (to avoid router-on-a-stick for >1G WAN installation) but I understand that the switch chip can't do that. Choice between two internal 10Gbit lines or two SFP+ ports and only one 10Gbit line to the CPU is clear - internal lines win!

Idea of single proper switch chip seems much better than current trend of two very limited chips. (After some thoughts, I realised those Realteks are rather port-extenders than switch chips).

I agree with the points you all have raised. However they would require much higher cost SOC's e.g. Marvell Octeon TX2 (CN9132) that have more 10Gigabit ports, and would use more power and generate more heat.

I tried to make a suggestion that is realistic based on low cost SoC's and will be a nice upgrade to the existing devices.

The SOC's I suggested can be configured in different ways, e.g. 2x 10gig direct to the SOC them multiple 2.5Gbit into the switch ASIC, but this would then limit bandwidth to the switch ASIC.

I agree with the points you all have raised. However they would require much higher cost SOC's e.g. Marvell Octeon TX2 (CN9132) that have more 10Gigabit ports, and would use more power and generate more heat.

I tried to make a suggestion that is realistic based on low cost SoC's and will be a nice upgrade to the existing devices.

The SOC's I suggested can be configured in different ways, e.g. 2x 10gig direct to the SOC them multiple 2.5Gbit into the switch ASIC, but this would then limit bandwidth to the switch ASIC.

I really agree for your suggestion. But how about your design 2 but marvell 88E6390X + AL21400? Then there would be nice ipsec power too.

I really agree for your suggestion. But how about your design 2 but marvell 88E6390X + AL21400? Then there would be nice ipsec power too.

The Marvell SOC's have hardware crypto too... The reason I didn't suggest the AL21400 is that it's 32bit, puts out a lot of heat, and all of Mikrotiks Annapurna Labs based devices have widespread complaints about stability.

Marvell is a safe choice. 64bit, low cost, low TDP, excellent vendor support, proven stability, HW crypto, familiarity for Mikrotik developers and engineers.

The Marvell Cavium Octeon TX2 CN9132 is the best choice for Design2. My only concern was the cost of it compared to the Armada 8040.

I know Mikrotik have been very "cost sensitive" on this tier of products in the past. But maybe it's time to spend a few extra $ to get a much more powerful SOC that delivers better performance and has less bottlenecks. I don't think price is the main concern for customers with this level of product.

I had no intention to say that it needs to be modified. I meant that "I wish for it, but I understand that it is not possible". After all, every device is perfect for particular task and it depends how we balance performance/price. I don't think that getting second SFP would be good justification for increased cost.

The Marvell Cavium Octeon TX2 CN9132 is the best choice for Design2. My only concern was the cost of it compared to the Armada 8040.

I know Mikrotik have been very "cost sensitive" on this tier of products in the past. But maybe it's time to spend a few extra $ to get a much more powerful SOC that delivers better performance and has less bottlenecks. I don't think price is the main concern for customers with this level of product.

Oh yes. You are absolutely right. Those cores are much more powerful than AL21400.

btw. CSS326-24G-2S+RM is quite fine product. I hope that Mikrotik will look other Marvell chips too.

I like your Design 2, would buy it, if price was right...
Only thing I would add is at least one mini-pcie slot, it's about $1 in BOM cost and it would make it much more versatile (could add WIFI, LTE, LoRA, ...)
Probably worth even for cheaper Design 1, cost is minimal... and then it can easily be sold with wifi or LTE card, without any design/PCB changes.

I agree, this is a must, and this would fit perfectly as a backup solution for small office/home office customers.
I am eagerly waiting for such device with RB4011 performance and PCI-e/LTE card support!

+++
Mikrotik I want RB5011!!! Sell me two: in desktop case with wifi and for 19" with LCD, double power supply, M2 nwe for dude :-)