I need a new router, and the CRS418-8P-8G-2S+RM looks perfect for me, but when comparing it to the CCR2004-16G-2S+PC, I was puzzled by the performance values, especially since both devices are priced similarly.
|
CRS418-8P-8G-2S+RM |
CCR2004-16G-2S+PC |
| CPU |
4×2.2 GHz ARM 64bit |
4×1.2 GHz ARM 64bit |
| Max power w/o attachments |
28 W |
30 W |
| Cooling |
4 fans |
passive |
| RAM |
1 GB |
4 GB |
I understand that the CRS’s advantage for switching is its single switch chip, and I understand that the CCR has twice the throughput, namely 40 Gbit/s, to the CPU, and let's assume I don't need PoE: Should I opt for the CCR?
Why are the performance figures for the CRS so significantly worse? Am I reading the test results incorrectly? I understand that the CCR has an Annapurna Labs CPU, but the CRS has 1 GHz more – is it still worse than the Annapurna Labs?
|
CRS418-8P-8G-2S+RM |
CCR2004-16G-2S+PC |
| Routing (fast path) |
4934.8 Mbps |
31253.4 Mbps |
| IPsec (256 tunnels, AES-256-CBC+SHA256) |
603.7 Mbps |
2601.8 Mbps |
I would be grateful if someone could shed some light on this for me 
CRS418 is and remains at first a switch.
It can route far better then a lot of other switches but it remains a switch.
It is no match to a modern device designed as router.
That doesn't mean it can not have value in some contexts/ environments but if you are already wondering now about routing performance, maybe you should look at a router.
RB5009, maybe ?
This comment is mainly to avoid someone does get this device and then complains about routing performance compared to other routers.
That's comparing apples to oranges.
My personal comment with CRS418 is that it only has 1Gb ports.
As far as I understand however there is still a huge market for it.
CRS are SWITCHES focusing on switch + moderate abilties to route as a bonus.
CCR are ROUTERS focusing on routing stuff with some better or worse results in simple switching.
As https://mikrotik.com/product/crs418_8p_8g_2s_rm says:
You wanted a switch that could do more.
Cloud Router Switch = CRS hybrid of a switch + router
Cloud Smart Switch = CSS = mainly switch
Cloud Core Router = mainly router
Sometimes the difference is so small that you do not know how to classify a device.
I understand that, but I was wondering more from a technical point of view: How can the 4×2.2 GHz CPU be slower than the 4×1.2 GHz CPU? I thought that maybe I was misinterpreting the test results. Because the block diagram would not have led me to expect this difference in performance.
AL32400 is quad core A57 arm.
Ipq-8072 is quad core A53 arm.
A57 is more performance oriented.
A53 is more power efficiency oriented.
So obviously they perform quite different when compared to each other ( not even taking into account clock frequency difference)
Want more ?
All info is available if you search a bit ...
The CPU in the CRS418, IPQ8072 is a quad-core ARM Cortex-A53 chip. The CPU in the CCR2004 is an Annapurna Labs Alpine v2 AL32400, which is a quad-core ARM Cortex-A57 chip.
Cortext-A53 only has in-order execution, while the Cortext-A57 supports out-of-order execution. It depends on the workload, but clock-for-clock, an out-of-order processor should be a few times faster than an in-order processor. And then you also have differences like cache sizes and memory bandwidth.
Those A53 cores are usually used as the low power cores (LITTLE), while the A57 would be the performance cores (big) in those CPUs with big.LITTLE cores from 10 years ago like the Snapdragon 810. Of course, the A57 is old and no longer used in flagship phones, but the A53 cores are still used as low-power cores even now.
And the IPsec numbers also depend on whether IPsec hardware acceleration is available. And the IPsec numbers for the CCR2004-16G-2S+PC is probably wrong, because they seem to be copied from the numbers of the CCR2004-16G-2S+ (with fan) which is clocked at 1.7GHz. You should divide the CCR2004 PC IPsec numbers by about 1.4x.
Thank you guys for your answers. I ordered my CRS418-8P-8G-2S+RM today. It will definitely be more than fast enough for me.
Thanks for pointing that out. I understand the technical differences.
Nevertheless, I still can't quite believe the test results. The CRS CPU is connected to the switch chip at 2× 10 Gbit/s full duplex. And the maximum throughput through the CPU is supposed to be only 5 Gbit/s? I don't know, I just don't think that can be right.
