Hi,

do you have any idea how is it possible that AX2 has only 874 MHz and the AX3 has 1800MHz but Mikrotik says that they are able to route almost the same?? Im wondering if it is good idea to replace AX2 with AX3 due CPU power benefit...

Thank you

published test cover some different scenarios:
one of its is fast-path scenarios with big packets where you reach the limit of CPU to Switch interface

the other scenarios are cpu taxing tasks where cpu becomes the limit

this small CPU's have a narrow memory bus to be cheap and power efficient

but this narrow bus become saturated when 4 cores are pushing through, limiting the achievable max performance

i think hap AX3 can shine in specific scenarios where you need CPU processing power adn the taskt mostly use a single core and the other 3 cores are not used too much, but not when maxing out all cores as on published performance test

Well, official test results indicate 25% difference in real life (e.g. routing 25 filter rules, 512 byte packets: 1145Mbps ax3 VS 912Mbps ax2). Surely that's a lot less than the difference in CPU clock. But then routing (in v7 specially) can be memory-intensive and it's possible that RAM types (ena memory bus in general) is very similar in both devices.

But then, IPsec tests are identical ... as if somebody copied test results of one of devices to the product page of the other device...

i think ipsec tests reach the imposed limit of encryption engine before reaching cpu limit because of that the limits are the same

thanks for answers but I still dont get it...

published test cover some different scenarios:
one of its is fast-path scenarios with big packets where you reach the limit of CPU to Switch interface

so here should AX3 shine....

this small CPU's have a narrow memory bus to be cheap and power efficient

do you mean that the procesor is bad designed and can use only 4x874??? I dont think so

i think hap AX3 can shine in specific scenarios where you need CPU processing power adn the taskt mostly use a single core and the other 3 cores are not used too much, but not when maxing out all cores as on published performance test

Do you have any example?

Well, official test results indicate 25% difference in real life (e.g. routing 25 filter rules, 512 byte packets: 1145Mbps ax3 VS 912Mbps ax2). Surely that's a lot less than the difference in CPU clock. But then routing (in v7 specially) can be memory-intensive and it's possible that RAM types (ena memory bus in general) is very similar in both devices.

But then, IPsec tests are identical ... as if somebody copied test results of one of devices to the product page of the other device...

Let's accompany this with some crude calculations

% of advantage of ax3 over ax2
where you see single digit advantage is because is reaching CPU to SWITCH interface limit not cpu processing limit

so here should AX3 shine.... Ans: Just the opposite

this small CPU's have a narrow memory bus to be cheap and power efficient

do you mean that the procesor is bad designed and can use only 4x874??? I dont think so Ans: bad designed no, goodfully market segmented

i think hap AX3 can shine in specific scenarios where you need CPU processing power adn the taskt mostly use a single core and the other 3 cores are not used too much, but not when maxing out all cores as on published performance test

Do you have any example? Ans: When you use a VPN or a tunnel to pass most of your traffic, frecuently in that scenario i have seen mostly only one cpu core being used

ok, here I see the diffrences...but still 4x874 vs 4x1800 ??? it is only 30% in avrage more powerfull? doubled MHz...

Hi, I will get 300/300 and im using wireguad between 3 location. Id no not if to change AX2 for AX3 because of CPU for wireguard or to change wireguad for hardware decode IPSEC????????????????

Not sure where you're going with this thread. Doubling clock speed doesn't double processing power because other factors impact performance and become the bottleneck in any computer system. ARM is designed for specific tasks and low power requirements. There's documentation online explaining the differences between architectures and why RISC chips are used. A SoC is a cheap device, end of story. Apple Silicon is in another class. It doesn't mean that any of them are bad, and ax2 is down clocked because of the space/heat limitations.

If you require a lot more performance, MikroTik offers more powerful devices based on more powerful chip designs. hAP has always been about value when compared to similar devices from other manufacturers, with the added benefit of using ROS for free.

Well, official test results indicate 25% difference in real life (e.g. routing 25 filter rules, 512 byte packets: 1145Mbps ax3 VS 912Mbps ax2). Surely that's a lot less than the difference in CPU clock. But then routing (in v7 specially) can be memory-intensive and it's possible that RAM types (ena memory bus in general) is very similar in both devices.

But then, IPsec tests are identical ... as if somebody copied test results of one of devices to the product page of the other device...

Let's accompany this with some crude calculations

% of advantage of ax3 over ax2

ax2-vs-ax3.png

where you see single digit advantage is because is reaching CPU to SWITCH interface limit not cpu processing limit

Could you please create the same conparison for AC3 vs AX3 and post it here? ????

Let's accompany this with some crude calculations

% of advantage of ax3 over ax2

ax2-vs-ax3.png

where you see single digit advantage is because is reaching CPU to SWITCH interface limit not cpu processing limit

Could you please create the same conparison for AC3 vs AX3 and post it here? ????