Used gpu
I have found results from50000+products and20000+suppliers aboutUsed gpu
Nvidia H200 Nvl 141GB GPU High-Performance Computing Graphics Card Pn: 900-21010-0040-000
Nvidia GPU A100 40g/80g/96g H100 80g/96g Pcie H200 H100 94G Nvlai/Data Analysis/High Performance Computing Accelerator Card
New Rtx A2000 A4000 A5000 A6000 Rtx16GB GPU Network Card Video Card GPU
Original New Nv A30 24GB Tensor Core GPU Accelerator 165W Ampere Pcie Passiv Gddr6 PCI Express 3.0 X16 A10 24GB
Original New Nvidia Tesla T10 16GB GPU H100 80GB Tesla H100 A100 96GB Advanced GPU Pcie New Used Workstation/Desktop/Gaming Video Card 16GB Gddr6X Video
Rtx PRO 5000 Blackwell Server GPU for Workstation
Nvidia Rtx A4000 16GB Gddr6 Ecc Professional Workstation GPU Graphics Card
Nvidia A10 24GB Gddr6 Ecc Data Center GPU Accelerator Card for Server
Nvidia Rtx PRO 6000d 84GB Gddr7 Ecc 624 Tensor Core GPU Card
High-Performance Rtx A6000 48GB GPU Server for Professionals
High-Performance Nvidia A40 48GB GPU for Visual New Ai Training Inference Server Graphics Card Host
Nvidia H200 Nvl 141GB Tensor Core GPU High-Performance GPU
Nvidia Blackwell Rtx PRO 4500 32GB PCI Express 5.0 Fan-Cooled GPU Graphics Card
Nvidia Rtx PRO 6000d Blackwell 84GB Gddr7 Ecc Ai Server GPU Accelerator
High-Performance Blackwell GPU Rtx PRO 6000 Graphics Card for Sever or Workstation
Nvidia Rtx PRO 4000 Blackwell 24GB GPU Blackwell Generation Professional Graphics Card
Professional Level A100 80g GPU for Ai Inference Training 80GB
Data Center Computing Solution Premium Quality Nvidia H200 141g Ai Computing GPU Trade Supply with Warranty Service
The New Spot Suiyuan S60 48GB GPU Computing Graphics Card Replaces The L20 Graphics Card Ai Inference Acceleration Card
New Computing Artificial Intelligence Domestic GPU Graphics Card Tianshu Zhixin Tiangai 150 64G Corresponds to Tesla A100
Brand New Stock Rtx5090 32GB GPU Public Graphics Card Artificial Intelligence Analog Computing Graphics Card
New Stock Original Ai Inference Training Compute Acceleration GPU Graphics Card a-100 40GB
100% Original Nvidia GPU Tesla H100 80GB Graphics Card for High-Performance Computing
Sourcing guidance for Used Gpu
What are the key technical specifications to verify when sourcing used GPUs for commercial use?
When purchasing used GPUs, you must prioritize VRAM capacity (GB) and Memory Type (GDDR6/GDDR6X) to ensure compatibility with modern software. For AI or rendering tasks, verify the CUDA core count (NVIDIA) or Stream Processors (AMD). It is critical to request the BIOS version to ensure the card hasn't been flashed with a custom mining BIOS, which can affect stability in standard workstations.
How can a buyer assess the physical and functional health of a used GPU remotely?
Request a FurMark or 3DMark stress test video from the supplier, ensuring the Serial Number (S/N) on the card matches the one in the software interface. Check for thermal performance; temperatures should not exceed 85°C under load. Ask for high-resolution photos of the PCB (Printed Circuit Board) to look for signs of overheating, such as discoloration (yellowing) around the memory chips or oil leakage from thermal pads.
What compliance and safety standards apply to used electronic components in cross-border trade?
Ensure the products comply with RoHS (Restriction of Hazardous Substances) to facilitate smooth customs clearance in the EU and North America. Even for used goods, CE or FCC markings are often required. Additionally, verify that the supplier follows UN38.3 standards if the GPUs are shipped as part of a system containing lithium batteries, though standalone GPUs primarily require Anti-Static Packaging (ESD bags).
What are the typical usage scenarios for bulk-purchased used GPUs?
Used GPUs are highly sought after for Crypto Mining farms, AI Model Training, Video Rendering clusters, and Internet Cafe upgrades. For enterprise-level deployment, focus on Data Center GPUs (e.g., NVIDIA Tesla/A-series), whereas for retail or gaming builds, Consumer-grade cards (e.g., RTX 30-series) offer the best price-to-performance ratio.
Cross-Border Procurement Risks and Strategies for Used Electronics
What are the primary risks when importing used GPUs and how can they be mitigated?
The biggest risk is 'Mining Fatigue', where cards have been run 24/7 under high heat. Mitigate this by sourcing from Verified Suppliers on Made-in-China.com who offer a functional warranty (typically 30-90 days). Avoid 'as-is' listings without buyer protection. Use Escrow payment services to ensure funds are only released after the goods pass your initial quality inspection.
How should buyers negotiate with suppliers regarding bulk orders of used components?
Negotiate based on Batch Consistency. Request that all units come from the same original equipment manufacturer (OEM) like ASUS, MSI, or Gigabyte to simplify driver management. Aim for a 5-10% discount on orders exceeding 100 units, and always negotiate for replacement units (buffer stock) to be included in the shipment to cover potential Dead-on-Arrival (DOA) cases.
What are the logistics and customs precautions for shipping used GPUs to international markets?
Accurate HS Code classification (typically 8473.30) is essential to avoid fines. Be aware that some countries have strict e-waste regulations regarding the import of 'used' electronics; ensure the commercial invoice clearly states the items are 'Refurbished' or 'Working Used Components' rather than scrap. For shipping, insist on double-walled boxes with individual foam slots to prevent physical impact damage during transit.





























