ChipFlip
$214.49 separates a used RTX 3080 from a used RTX 4070, $185.50 against $399.99. That's more than double the price for what's technically a lower tier card. Buying the previous generation's flagship over the newer generation's mid-range part is one of the oldest debates in used hardware, and this pairing puts real numbers on it.
Raw performance still favors the 3080
Nvidia's Ampere flagship was built with a wide memory bus and a genuinely large core count, and none of that stopped mattering just because a newer architecture exists. In pure rasterized performance, the 3080 holds its own against the 4070 and beats it outright in several workloads. Old flagship silicon is still fast silicon.
But the 4070 wins everywhere else
Lower power draw, better ray tracing performance per watt, and access to Nvidia's newer DLSS frame generation, which the 3080's older tensor cores can't run. If your build cares about efficiency or you actually use upscaling, the 4070 closes the raw-performance gap and then some.
Age is the real risk with the 3080
A 3080 on the used market has had years to accumulate wear, mining history, or thermal paste that's never been reapplied. None of that shows up in a spec sheet. It's the actual reason a $214 discount exists: you're not just buying older architecture, you're buying more uncertainty about the specific card's history.
Bottom line
Buy the 4070 if efficiency, DLSS frame generation, or peace of mind about hardware age matters to you. Buy the 3080 if raw rasterized performance per dollar is the only thing you're optimizing for and you're comfortable with the risk that comes with older used silicon.