NVIDIA GeForce RTX 2080 Ti modded with 22GB memory hits eBay for $500, ready for AI workloads

NVIDIA released its GeForce RTX 2080 Ti graphics card five years ago, launching with 11GB of VRAM, but now modders are doubling that to 22GB of VRAM and selling them on eBay as AI GPUs.

3

NVIDIA GeForce RTX 2080 Ti with 22GB VRAM (source: eBay)

The US sanctions on China have stopped high-powered graphics cards from entering the country, with the likes of the GeForce RTX 4090 banned, which is why we saw the introduction of the modified GeForce RTX 4090 D sent to China. We've seen modders in the country taking apart gaming graphics cards and turning them into AI GPUs, and now we've got modded RTX 2080 Ti cards hitting the market. Oh boy.

The modded consumer-based gaming-ready GeForce RTX 2080 Ti has been modded, rocking 22GB of VRAM, which is far better than 11GB for AI workloads. The VRAM acts as a high-speed buffer for those huge AI datasets, complex models, and calculations that are needed for AI workloads. Interesting move, modders.

• Read more: NVIDIA GeForce RTX 2080 Ti Review
• Read more: Check out this GeForce RTX 2070, modded from 8GB to 16GB of GDDR6

A seller on eBay is flogging off the modified GeForce RTX 2080 Ti graphics cards with 22GB of VRAM for $499 which isn't too bad. The modders themselves have tested the modified GPUs, finding that it generates a detailed picture with a resolution of 512 x 768 in 5.4 seconds, not too shabby for a five-year-old GPU.

3

The modders are using a blower design in dual-slot form factors as the modded RTX 2080 Ti graphics cards, which is perfect for servers and systems that are being used for AI. You don't want huge 3-slot or gigantic 4-slot cards, as you want more than one of them installed for your AI workloads.

The eBay description on the modified RTX 2080 Ti with 22GB VRAM reads:

• Best budget alternative to RTX 3090 RTX 4090
• More information and videos at: https://2080ti22g.com
• Fully supported by NVIDIA's official driver. No need to use the customized driver.
• Perfect for deep learning training! Only half $ compared with 3090 with almost the same GPU memory.
• Budget choice for deep learning students & startups.
• Tested for Stable Diffusion. Large language model. LLAMA2.
• Compatible with personal PCs or data center servers.
• No need to modify your BIOS. It works out of the box.
• Official driver supported. No need to install additional software. nvidia-smi
• Support Ubuntu / Centos / Debian / Windows / Pytorch / Jax / Tensorflow
• Support Regular Gaming Use. Display ports and HDMI ports.
• Shipping Time: Allow 10 business days to receive your order. We need time to upgrade the GPU and fully test it before we ship it to you