Well, you can fit a pretty beefy heatsink on the QCA9558 the c7 v2 uses. Remove the lid, add the heatsink with some quality thermal paste, add a low current 5v fan running off the USB port if needed. Your cooling problem is now solved, and the SoC will now run cooler than it ever did at stock settings with the lid on, and the extra heat from overclocking is not a problem. Hell, I don't overclock my routers yet I always do this to them, cool electronics are happy electronics under any load.
This is safe overclocking since no one is messing with voltages, just increasing clock speeds. You can overclock things this way because they're designed with a safety margin. One can get closer to the edge of stability at stock voltage, obviously stress testing things to verify while running at a higher clock speed, what's the issue? The worst that can happen is instability, and that can be easily solved by lowering clock speeds.
I've seen people here run their QCA9558s at 1GHz from 750MHz stock (+33%), that combined with the latest improvements like fast path should result in a pretty powerful router for home conditions.
Go and get better Hardware for 100ā¬/$ more.... Please dontĀ“t try to compare old fashioned Hardware with modern SOCĀ“s.
Just buy a second or a third device to maximise performance
... and I started overclocking with 68060 and PPC 603e
This is factually inaccurate, as overclocking is overclocking (how exactly do you believe the clock speed gets to where it is without voltage & impedence)?
Overclocking a router SoC is not the same as overclocking a PC/server CPU... they are two fundamentally different things engineered for entirely different purposes. Overclocking is never "safe" in and of itself, as the risk is always relative.
.
This is also factually inaccurate. Please see my above post, or use your search browser of choice.
By all means, but the QCA9558 & QCA9563 are SoCs that include the radios on the chips themselves, which is yet another thing that differentiates router SoCs from PC/Server CPUs.
Has anyone contemplating overclocking bothered to look into how overclocking affects the radios on the SoC?
Overclocking will cause heat that have to be transported... playing many years with it -> AMIGA1200 / 68060/603e PPC | Thermoelectric cooling (using the Peltier effect).... in that case we have a problem with humidity
And IĀ“am sure some Overclocking can be done by replacing some chips on the PCB (replacing some silicon with higher clocks) .... is it worth? Go and get a second or a third device
I actually underclocked my C7 v2 to 680 Mhz because it being warm to the touch kinda irks me and also I use it only as a dumb AP. While it still feels warm, I'm sure it has a lesser power draw. I should invest in a laser thermometer to measure it, but then that minuscule power bill saving will be moot. ::sarcasm::
I believe that we are more or less in the same page, except that while you state " the likely course of events is the router will fail or exhibit instability due to physical damage" I believe that nothing wrong will happen to it during it's expected working life because of being run on an overclocked state.
I have no doubts that overclocking it will reduce its life, just as undercloking it will extend it. The less power the cpu consumes and the less heat it is subject to, the longer it will take for degradation due to electromigration to take it's tool. I just think that the router will not have a service life where I will see it fail because of an overclock.
I also don't believe that the extra heat will cause any damage to the PCB or solder joints. Extra heat does affect electrolytic capacitors but I think that the C5/C7 capacitors are solid state and so, much more heat tolerant.
The C7 is an overclocking beast, to the point that I am considering selling the NBG6817 that I ordered.
I overclocked it from 720MHZ to 1480MHZ, along with ram fron 600 to 760MHZ and bus from 200 to 740MHZ and I have been running it at these speeds for one week now without any problem, reboot or hang. absolutely amazing!!
no idea, but as the CPU clock more than doubled, the performance increase should be linear, so I would expect nat performance, or the performance of anything else that is CPU bound to more than double.
