Old GPUs Can Still Sell, Stream, Encode, and Run AI Work

After an upgrade, an old graphics card can still sell, stream, encode, capture, or run local AI workloads. Here is what operators should know.

What happened

Engadget reported on October 19, 2026 that throwing away an old graphics card is a bad idea, especially as GPU prices rise. The article explains that an old GPU has options for a second life, from adding more power to a current machine to building a purpose-built second one. Every three to five years is a common cadence for GPU upgrades, so the source argues it is important to plan what happens after the next cycle.

The most popular card in the latest Steam hardware survey is the Nvidia GeForce RTX 5070, installed in 6.15 percent of surveyed systems. It launched at $549 and can now cost more than $800 because of memory and silicon price inflation. Higher prices mean more people will use their GPUs for longer, which may stretch the typical upgrade cycle.

Selling remains an option. An eBay search at the time of writing shows some used RTX 3060s selling for almost as much as they originally cost. Even the 10-year-old $249 GTX 1060 is typically listed at $50 to $100 in good condition, though these are always subject to fluctuation.

Beyond resale, an old GPU can go into a home media server for hardware-accelerated video encoding and decoding with Nvidia NVENC or AMD AMF. Both Jellyfin and Plex support hardware-accelerated streaming, so the old card can convert video codecs in real time and offload tasks from the processor. It can also run legacy code libraries or DirectX 9-era games, or add extra HDMI/DisplayPort connections.

Key facts

• The most popular Steam hardware survey card is the Nvidia GeForce RTX 5070, installed in 6.15 percent of surveyed systems.
• The RTX 5070 launched at $549 and can now cost more than $800 because of memory and silicon price inflation.
• Used RTX 3060s sold for almost their original cost; the 10-year-old $249 GTX 1060 is typically listed at $50 to $100 in good condition.
• Old GPUs can handle hardware-accelerated encoding and decoding via Nvidia NVENC or AMD AMF in Jellyfin or Plex, and can support DaVinci Resolve, Blender, video capture, and local AI workloads.
• Lossless Scaling costs $6.99 on Steam and can let an old card act as support for a new one.

Our analysis

For social media managers, streamers, and small video teams, this is not just PC-building trivia. A spare GPU can offload encoding work from the main processor, which likely reduces render bottlenecks when producing platform-ready video. The source specifically points to Plex, Jellyfin, DaVinci Resolve, and Blender as practical uses, meaning an old card can become a low-cost production asset rather than waste.

The reported resale prices suggest that older hardware still has meaningful value. Teams that refresh workstations can recover part of an upgrade cost or keep a backup unit. The hardware inflation described in the source may make this behavior more common, since the every three to five years upgrade pattern becomes harder to justify when new cards cost more.

There is a clear caveat: putting a second GPU in the same main PC requires enough power, ventilation, and motherboard support. If the processor uses lane splitting, a second card can degrade the primary GPU. The source advises checking the motherboard manual before committing resources, which is practical guidance for anyone attempting a dual-GPU setup.

What it means for operators

• After an upgrade, do not discard the old card; first sell it if you have no immediate use. Used RTX 3060s and the GTX 1060 show resale value, but prices fluctuate.
• Build a media server or capture PC with the old GPU to handle hardware-accelerated streaming in Plex or Jellyfin, offloading encoding from your main editing machine.
• For video teams, use the old card for capture or as a second encoder for DaVinci Resolve or Blender. If adding it to the main PC, confirm the power supply, motherboard dual-GPU support, and PCIe lane splitting first.
• If the card is still working but unused, store it in an antistatic bag as a spare, or run Lossless Scaling on it as support. If it is dead, use e-waste recycling rather than trash to recover gold, copper, and aluminum.