The most expensive-looking cheap cable in the building is the one that passes every casual test and fails every real one. Cable quality faults are distance faults: perfect on a 2 metre bench test, guilty at 60 metres in the wall - which is why they survive years of troubleshooting.
Ethernet distances are electrical budgets. Gigabit signalling over twisted pair is specified for a 100 metre channel - roughly 90 metres of solid-core installed cable plus patch leads - using proper copper conductors of the right category. Every shortcut spends that budget faster: aluminium-core cable spends it on resistance, flat cable spends it on crosstalk, and sloppy terminations spend it at both ends. A link can be inside the length limit and still be outside the electrical one.
Example only: if the same PC is perfect at the switch and faulty at the desk, the PC is innocent and the installed run - cable, jacks, patch panel - owns the fault.
Much budget cable is CCA: copper-clad aluminium. An aluminium core with a copper skin conducts worse than solid copper - higher resistance on every metre - and it is brittle, work-hardens at bends, and oxidises at terminations. At 3 metres you will never know. At 60 metres it can be the difference between gigabit and fallback, and under PoE it heats more and drops more voltage than copper, which is why cameras on CCA runs reboot on cold nights. It is also sold with honest-looking category labels. Price and weight give it away; a magnet will not, but experienced buyers cut and look at the silver core.
Flat under-carpet cables trade the twisted-pair geometry that cancels interference for convenience - fine for a short run in a quiet room, marginal beside power cables. Tight kinks and crushed sections change a pair's spacing permanently; a chair leg can create a fault that appears only when someone sits down. And routes matter: a run strapped alongside lift motor feeds, big UPS outputs or fluorescent ballasts drinks interference for its whole life. If drops correlate with the lift or the AC compressor starting, stop testing software and go look at the route.
Cat5e, honestly installed, carries gigabit to the full 100 metres - that is the specification, not optimism. Cat6 buys margin and future speed, not magic, and no category rescues a bad termination. Conversely, ancient genuine Cat5 or a run re-purposed from telephone wiring may negotiate 100 Mbps happily and never hold gigabit. Judge the installed result - link speed, error counters, length - instead of the printing on the jacket, which describes the day the cable was made, not the decade since.
See the video case - a reminder that the cable is one suspect among several, and it was cleared by test before the real culprit was found.
Watch Full VideoFor a short desktop lead, you may never notice it. The problems grow with length and load: aluminium has higher resistance, so long CCA runs lose signal margin, run hotter under PoE and age badly at terminations. For installed runs and gigabit at distance, solid copper is the specification, not a luxury.
Yes - gigabit Ethernet was designed for exactly that: 1000 Mbps over a 100 metre channel of Cat5e or better, properly terminated. Most real-world failures at length are termination quality, CCA conductors, detours that push the true length past the limit, or interference on the route - not the category itself.
Gigabit stresses the channel far harder: all four pairs, both directions, tighter signal margins. A run with barely enough margin passes at 100 and fails at gigabit, especially under full-speed traffic. Shorten the path with a mid-point switch or replace the run - marginal physics will not heal.