You don't need hardware acceleration. The device is powerful enough to pass 1Gb/s through NAT with Openwrt software flow offloading.
Hardware acceleration is supported on Mediatek, but this IPQ4018 is better.
Thanks Pilot6 , I wasn't sure if openwrt completely removed their software or if openwrt was a edited version of their firmware that kept some of the code intact.
Sorry for the newbie posts but i wanted to be sure before i buy the EA6350 and try openwrt for the first time.
Guide: don't do it outside of testing sandbox environment. If you ask for guides then you mostly are not someone who should do it.
The possible performance gain is near to irrelevant for real time applications but the stability problems for your network service and reduced device longevity is there all the time.
Like i said above I was curious. I am curious how the performance would be with Cake enabled.
I have to disagree with you on this. During my CompSci degree this was the best way for me of learning a new language is looking a guides and examples.
I figured since I have a backup pfsense box I'll be prepared for the worst.
At least when you screw up a computer program, generally all that happens is that it doesn't run.
If you screw up overclocking, your SoC is dead. Permanently. Or your memory fails. Permanently.
These devices are designed from silicon, packaging, and device to meet specs, not exceed them, and at the lowest possible cost. Higher clock rates mean higher power dissipation, which needs to be extracted from the silicon, and even from within the silicon. Higher clock rates often require higher voltages -- no BIOS to change the supply rails. Higher voltages will, at some point, break down the junctions or insulating layers in the chip. These are devices clocked at rates where electron transition time becomes critical. "Wires" aren't wires at 1 GHz. Signal propagation often isn't linear with clock rate. This ain't your over-built P90 that you can clock to 120 Mhz.
Overclocking can, at best, gain you a few percent. What you see in actual performance gain is usually significantly less than the increase in clock rate.
Yes, this really is a "if you have to ask, you shouldn't try" kind of thing, unless the $75 or more you'd spend is "throw-away" money for you.
Let's assume you can reach 20% overclock with reasonable stability. Let's also assume the OpenVPN task fits in the CPU cache and scales 66% with clock speed increases.
Then you might get a performance of 30 Mbit/s stock and 34 Mbit/s overclocked OpenVPN throughput. But at the same time you could see stability problems.
I would suggest something else. IF you are able to route 100% of line speed through your device: no overclocking necessary. Don't do it as it gets you nothing but it will produce problems. IF not: buy a more powerful router which is able to reach your line speed.
Overclocking could be useful, if you earn money out of more performance and you are already working with the fastest available stock speed devices. But not if you run a low end device and just want to play, without a reason to run into stability problems.
There is another performance dimension: latency. But that's a different story and overclocking cheap routers would also be one of the last things I would consider to get a lower latency.
Another thing would also be the case if you wouldn't do real time processing, but you would do computation tasks that take minutes or hours to complete. Let's assume for example your task takes one hour to complete on stock clock speed and again your task is heavily CPU speed bound and benefits 66% of more CPU clock speed. Then it could finish not in 60 minutes, but in 53 minutes.
The price for this benefit would be that you would need to check a lot more for consistency, as you could not be sure the result is correct on 20% overclock. I guess it might often be the case that it is not worth the increase in performance.
Also: I would have expected you to get to this conclusion by yourself if you own a CompSci degree.