router capable of processing 1000mbps in NAT mode, with a slot speed of 1000mbps and up to 5 slots.
Welcome @bin, on the product page you can find the test results.
It is not exact science, but if you choose ethernet test results and check the Routing, 25 ip filter rules, 512 bytes (Mbps) you will get a good indication. I.e. the RB5009 is more than capable.
The typical response is that for the lowest cost, your best bet to achieve a 1 Gig internet connection is to go for the AX3 wifi router. If your budget permits, the RB5009 router is very good value.
If you plan on going to 2.5Gig in the near future ( < 5 years ) then get the RB5009 now.
Be advised these are not plugNpray routers, they take work but its worthwhile!!
2 Likes
The problem with the C6U is that it only has 128MB of RAM total. A conntrack table entry takes up about 1kB in RAM (a sort of hand-wavy but good enough estimate.) A significant part of memory is taken up by the system and wifi drivers and buffers, so the number of connections for such a device are in the low tens of thousands.
Another usual problem is that some firmwares tend to leak memory. OpenWRT is pretty good in this regard, but a lot of what is put forward as OpenWRT contains additional modules and customizations which don't live up to its standards. The number of connections is also a configuration parameter (sysctl) for the kernel networking subsystem with a common default value of 10k. This is way too small for a router but some manufacturers fail to configure it properly, and it's only noticed by customers. (Most benchmarks only test with a limited number of connections.)
All of the usual devices can handle 1Gbps NAT-ing for usual traffic patterns. The hEX Refresh and hEX S are the basic models. If you need WiFi or have a more demanding traffic pattern (small packets, high rate of new connections) then the ax2 is a good choice. The ax3 has a bit more CPU power, it can mostly handle up to 2.5Gbps.
The rb5009, as suggested, is a very safe choice. It's a but pricier but you will not have to worry about its performance in basically any situation.
The hEX variations have 512MB of memory which is more than enough to handle several hundred thousand connections. The others are equipped with 1GB RAM, which is good for almost a million.
1 Like
There could also be the newish Hap Ax S as a competitor in the lower throughtput category.
Usual price comparison, model, price, 512 bytes routing with 25 firewall rules, memory,
| Model | List price | Speed | Memory |
|---|---|---|---|
| Hex Refresh | $60 | 498.1 Mbps | 512 Mb |
| Hex S (2025) | $69 | 498.1 Mbps | 512 Mb |
| hAP Ax S | $79 | 498.1 Mbps | 512 Mb |
| hAP ax² | $99 | 912.9 Mbpsb | 1Gb |
| hAP ax³ | $139 | 1145.2 Mbps | 1Gb |
| RB5009 | $219 | 3096.2 Mbps | 1Gb |
If we add a BftB column (Bang for the Buck 
) column, where BftB is expressed as (speed x memory) / (price x 1024) we have:
| Model | List price | Speed | Memory | BftB |
|---|---|---|---|---|
| Hex Refresh | $60 | 498.1 Mbps | 512 Mb | 4,15 |
| Hex S (2025) | $69 | 498.1 Mbps | 512 Mb | 3,61 |
| hAP Ax S | $79 | 498.1 Mbps | 512 Mb | 3,15 |
| hAP ax² | $99 | 912.9 Mbpsb | 1Gb | 9,22 |
| hAP ax³ | $139 | 1145.2 Mbps | 1Gb | 8,24 |
| RB5009 | $219 | 3096.2 Mbps | 1Gb | 14,14 |
Interesting BfB model but even though I would agree the hapax2 has been a decent model in many regards, it fails a basic requirement 1gig throughput in real world conditions and thus the ax3 is still my recommendation for a low cost option and the RB5009, the better long term value. One should note that the ax3 wifi came out of the gate with "old wifi" and will look really old when Mikrotik comes out with axe4 with wifi7 in 2030. 
1 Like
NAT is the requirement and that means fasttrack. All of them saturate 1Gbps readily.
@jaclaz: I still consider the hap ax somewhat of a vapor in terms if global availability. I'm really looking forward to reports about its wifi performance - this is going to be the first with a MediaTek chipset.
No matter what, if 5009 is affordable and WiFi is not needed, then I would choose overpowered device to sleep better. If WiFi is a must I would consider installing external AP, not focusing on built-in.
1 Like
@lurker888
It is available here in Italy from the major dealers/distributors.
No idea on real world performance of the wifi, but - and I may be wrong - even if it has the lower BftB score, it isn't a bad idea as a router, the "pure router" speed is the same as the hex refresh and it seems like a very flexible device can be re-used as AP at next network upgrade.
The advantage (IMHO) over the hex refresh (and also over hap Ax 2 and 3) is the SFP cage, from what I can understand it is a hex S with optional wi-fi for 10 bucks more.
(I am editing my previous post adding the hex S(2025)), and the USB port could always come useful.
Glad to hear. I'm really waiting for the feedback on wifi. Has Mikrotik managed to pull off a well orchestrated release or will there be incoming bugs?
I really hoped they would retain the 48V support but we can't have everything 
[ lurker888 ] - NAT is the requirement and that means fasttrack. All of them saturate 1Gbps readily
Can you point to where Mikrotik states they dont use fasttrack for any of their testing?
They don't use fasttrack for any of the tests. They do use fastpath, but only where stated. The two throughput numbers (fastpath and fasttrack) are quite close (when full hw support is available.)
You specifically asked for sources for this statement. I'm not aware of any. (That's not exactly atypical.) All I can point to is that they've published results in line with this even before fasttrack was introduced, and the numbers are consistent.
I've done a fair bit of testing and I've found the statement to be accurate.
The 512 byte / 25 filter rules numbers are frequently quoted. I do so as well, and I rely on them also. However my applications usually involve features that are incompatible with fasttrack: queues, mangling, multiple routing tables, not rule based policy routing, vrfs, etc.
My more limited experienced is with fastrack on and they do equate well with the posted numbers for 25 filter rules. I always assumed they tested with fasttrack especially when quoting the number of filter rules and and thus thought of them as useful for gauging the real world throughput. I shall ask directly.
It has been discussed on the forum several times to ask Mikrotik to publish fasttrack test results. For a lot if first time buyers, the 512 byte / fasttrack number would probably be the single most useful. You may want to enquire about this as well...
Thank you for pointing out the AX3 as a somewhat capable home router for 1 gbps use. My current hEX refresh can barely cope and I have an AX3 next to my hEX. When I looked at the specs, it might be wise to make the AX3 also the main router. The RP5009 looks like overkill.
It will need some effort to create a migration plan, but i will get there.
Good plan, the axe3 is quite good.
Put "make a current backup and export" at the top of the list. Always good to not burn the bridge in case you have to retreat.
1 Like
Hi Lurker, I did see on the CCR2216? ethernet results that included, the standard 25 IP filter rules and then one for when fasttrack is enabled. That certainly support your notion that all of the previous results were indeed with fasttrack off. Will see if they can confirm that. I suppose at least the results are good for when fasstrack is disabled, such as for mangling and queuing etc..
"fasttrack L3HW". Not exactly the same as only "fasttrack". And I guess they included "fasttrack" in description of this entry to indicate that it's about L3HW offloaded firewalling (rather than "plain" routing).
Yes, the "routing - 25 ip filter rules" is likely with fasttrack disabled. These numbers are generally lower than numbers for "routing - 25 simple queues" and we know that queues don't work with fasttrack enabled. So it is likely that router would perform better with fasttrack (and many filter rules) than without fasttrack (due to queues).