Modified GeForce RTX 4080 cards with double the VRAM are turning up on Goofish, China's secondhand marketplace, according to VideoCardz. These are stock RTX 4080 units that started life with 16GB of GDDR6X memory and have been upgraded to 32GB, with sellers asking more than 10,000 yuan, or roughly $1,480 for these cards.

We've seen a similar trend recently with the RTX 2080 Ti cards bumped to 22GB, which sell for around $500 on eBay. Of course, the newest RTX 4080 flavor of this conversion costs almost three times that, but you are getting more GPU for that price premium as well.
The RTX 4080 still runs a 256-bit memory bus and 9,728 CUDA cores, so doubling the VRAM doesn't touch bandwidth or gaming performance. The upgrade only really matters for AI inference, where fitting a larger model into local memory tends to matter more than raw throughput.

It isn't just RTX 4080s that are being given the AI-ready treatment. We've tracked similar mods on the RTX 4080 SUPER, which showed up with 32GB last year for around 9,200 yuan, and on the RTX 4090, which has been pushed to 48GB and even teased at a wild 96GB. The scale here seems to be growing too. It should be noted that these RTX 4080 conversions look less like one-off repair shop projects and more like a proper segment of the secondhand GPU market at this point.
The modding process reportedly involves swapping in higher-capacity memory chips and adjusting the board so the GPU actually recognizes the extra VRAM. It's the same basic trick used on the 2080 Ti and 4080 SUPER conversions, and it seems to work well enough that these GPUs now have their own dedicated market segment.

For AI hobbyists or small operations that are priced out of workstation-class hardware (like the refurbished RTX 4090D cards Chinese data centers have been offloading), a modded 32GB 4080 probably looks tempting next to official options that cost several times more.
Of course, the risk with these cards is that these aren't NVIDIA-sanctioned builds, so there's no factory warranty, no guarantee the memory chips are running within spec long-term, and no way to know exactly who did the soldering. Stability under sustained AI workloads is anyone's guess until more of these cards get real-world mileage.


Frequently Asked Questions
TweakBot answers common questions about this news using TweakTown's own coverage from this page and related content from our archive. All answers are generated from TweakTown content and not outside sources. Tap a question to reveal the answer, or type your own below related to this content.
How were these RTX 4080 cards modified to reach 32GB VRAM?
Does doubling VRAM on a 4080 improve gaming performance or memory bandwidth?
Why would AI hobbyists prefer a modded 32GB RTX 4080 over official workstation cards?
What warranty and reliability risks are associated with these non‑NVIDIA‑sanctioned mods?
Have a question about this content not listed here? Ask below and TweakBot will answer it.
It's worth watching whether NVIDIA or its partners try to crack down on this practice, given how normalized it's become across the RTX 40 series. For now, it looks like the modding scene in China isn't slowing down.






