Router temperatures - 2023

[Tech9]

It's not a thermal issue.

Total failure is not thermal related issue, but this router is also known for failing thermal transfer pads and this is a thermal issue. Everyone seeing an RT-AC86U running at 90C at room temperature has a router with failing thermal transfer pads. Replacing the pads lowers the temperature with about 15C and restores the original condition.

 

[RMerlin]

Total failure is not thermal related issue, but this router is also known for failing thermal transfer pads and this is a thermal issue. Everyone seeing an RT-AC86U running at 90C at room temperature has a router with failing thermal transfer pads. Replacing the pads lowers the temperature with about 15C and restores the original condition.

My RT-AC86U failed after having spent weeks unplugged, that's why I'm saying the cause of failures aren't thermal related.

 

[coxhaus]

For a SW based router, clock speed matters - e.g. if I have a CPU that is 2GHz, in ideal conditions, it can do 2 gigabits per second in perfect conditions - that being said, NAT and other things take away from that...

But that's not the whole story is it?

What is more important is packets per second - and there, one has to consider packet sizes and traffic use cases...

Should also note that unlike Linux and Netfilter, BSD based distro's like PFSense/OpnSense do not have the fast path in SW - Linux/Netfilter does...

Yes, packet per second based on size of packet. Cisco posts that on their equipment. Consumer equipment is afraid to do that. They give you a small packet size and a large packet size.

I have read a little about fast path in the later versions. Something they wrote into the kernel. Does it work in all cases even if you are doing traffic shaping and filtering?

 

[sfx2000]

Yes, packet per second based on size of packet. Cisco posts that on their equipment. Consumer equipment is afraid to do that. They give you a small packet size and a large packet size.

I have read a little about fast path in the later versions. Something they wrote into the kernel. Does it work in all cases even if you are doing traffic shaping and filtering?

that's probably a discussion for a new thread...

one quick note - fast path precludes shaping - fast path assumes no packets can be dropped, whereas shaping permits packets to be shed to maintain QoS levels for the different flows

 

[coxhaus]

one quick note - fast path precludes shaping - fast path assumes no packets can be dropped, whereas shaping permits packets to be shed to maintain QoS levels for the different flows

Thats kind of what I thought. It doesn't work so well with routers, maybe small scale. I will stick to high clock rates and low heat.