Used graphics cards
Previously owned GPUs, including cards removed from workstations and servers. Prior use alone does not mean a card has been refurbished or tested.
SECOND-HAND GPU MARKETPLACE
Find local GPU sellers.
Buy second-hand in bulk.
GPU.OOO connects buyers with independent sellers of second-hand NVIDIA graphics cards. The marketplace focuses on wholesale lots of 10 or more GPUs, not new retail hardware.
91GPU models
FILTER BY MEMORY
VRAM capacity
Modified / High-VRAM
7 resultsUnofficial board-level modifications for AI and memory-heavy workloads. These configurations are not supported by NVIDIA and require additional seller verification.
ModifiedModified / High-VRAM
GeForce RTX 2080 Ti 22GB (Modified)
Turing
- VRAM
- 22 GB
- CUDA
- 4,352
- Bandwidth
- 616 GB/s
ModifiedModified / High-VRAM
GeForce RTX 3070 16GB (Modified)
Ampere
- VRAM
- 16 GB
- CUDA
- 5,888
- Bandwidth
- 448 GB/s
ModifiedModified / High-VRAM
GeForce RTX 3080 20GB (Modified)
Ampere
- VRAM
- 20 GB
- CUDA
- 8,704
- Bandwidth
- 760 GB/s
ModifiedModified / High-VRAM
GeForce RTX 3090 48GB (Modified)
Ampere
- VRAM
- 48 GB
- CUDA
- 10,496
- Bandwidth
- 936 GB/s
ModifiedModified / High-VRAM
GeForce RTX 4080 SUPER 32GB (Modified)
Ada Lovelace
- VRAM
- 32 GB
- CUDA
- 10,240
- Bandwidth
- 736 GB/s
ModifiedModified / High-VRAM
GeForce RTX 4090 48GB (Modified)
Ada Lovelace
- VRAM
- 48 GB
- CUDA
- 16,384
- Bandwidth
- 1,010 GB/s
ModifiedModified / High-VRAM
GeForce RTX 5080 32GB (Modified)
Blackwell 2.0
- VRAM
- 32 GB
- CUDA
- 10,752
- Bandwidth
- 960 GB/s
Official NVIDIA models
84 resultsGamingNVIDIA TITAN
TITAN RTX
Turing
- VRAM
- 24 GB
- CUDA
- 4,608
- Bandwidth
- 672 GB/s
GamingGeForce RTX 50 Series
GeForce RTX 5050
Blackwell 2.0
- VRAM
- 8 GB
- CUDA
- 2,560
- Bandwidth
- 320 GB/s
GamingGeForce RTX 50 Series
GeForce RTX 5060
Blackwell 2.0
- VRAM
- 8 GB
- CUDA
- 3,840
- Bandwidth
- 448 GB/s
GamingGeForce RTX 50 Series
GeForce RTX 5060 Ti
Blackwell 2.0
- VRAM
- 8–16 GB
- CUDA
- 4,608
- Bandwidth
- 448 GB/s
GamingGeForce RTX 50 Series
GeForce RTX 5070
Blackwell 2.0
- VRAM
- 12 GB
- CUDA
- 6,144
- Bandwidth
- 672 GB/s
GamingGeForce RTX 50 Series
GeForce RTX 5070 Ti
Blackwell 2.0
- VRAM
- 16 GB
- CUDA
- 8,960
- Bandwidth
- 896 GB/s
GamingGeForce RTX 50 Series
GeForce RTX 5080
Blackwell 2.0
- VRAM
- 16 GB
- CUDA
- 10,752
- Bandwidth
- 960 GB/s
GamingGeForce RTX 50 Series
GeForce RTX 5090
Blackwell 2.0
- VRAM
- 32 GB
- CUDA
- 21,760
- Bandwidth
- 1,790 GB/s
GamingGeForce RTX 50 Series
GeForce RTX 5090 D
Blackwell 2.0
- VRAM
- 32 GB
- CUDA
- 21,760
- Bandwidth
- 1,790 GB/s
GamingGeForce RTX 50 Series
GeForce RTX 5090 D V2
Blackwell 2.0
- VRAM
- 24 GB
- CUDA
- 21,760
- Bandwidth
- 1,340 GB/s
GamingGeForce RTX 40 Series
GeForce RTX 4060
Ada Lovelace
- VRAM
- 8 GB
- CUDA
- 3,072
- Bandwidth
- 272 GB/s
GamingGeForce RTX 40 Series
GeForce RTX 4060 Ti
Ada Lovelace
- VRAM
- 8–16 GB
- CUDA
- 4,352
- Bandwidth
- 288 GB/s
GamingGeForce RTX 40 Series
GeForce RTX 4070
Ada Lovelace
- VRAM
- 12 GB
- CUDA
- 5,888
- Bandwidth
- 504 GB/s
GamingGeForce RTX 40 Series
GeForce RTX 4070 SUPER
Ada Lovelace
- VRAM
- 12 GB
- CUDA
- 7,168
- Bandwidth
- 504 GB/s
GamingGeForce RTX 40 Series
GeForce RTX 4070 Ti
Ada Lovelace
- VRAM
- 12 GB
- CUDA
- 7,680
- Bandwidth
- 504 GB/s
GamingGeForce RTX 40 Series
GeForce RTX 4070 Ti SUPER
Ada Lovelace
- VRAM
- 16 GB
- CUDA
- 8,448
- Bandwidth
- 672 GB/s
GamingGeForce RTX 40 Series
GeForce RTX 4080
Ada Lovelace
- VRAM
- 16 GB
- CUDA
- 9,728
- Bandwidth
- 717 GB/s
GamingGeForce RTX 40 Series
GeForce RTX 4080 SUPER
Ada Lovelace
- VRAM
- 16 GB
- CUDA
- 10,240
- Bandwidth
- 736 GB/s
GamingGeForce RTX 40 Series
GeForce RTX 4090
Ada Lovelace
- VRAM
- 24 GB
- CUDA
- 16,384
- Bandwidth
- 1,010 GB/s
GamingGeForce RTX 40 Series
GeForce RTX 4090 D
Ada Lovelace
- VRAM
- 24 GB
- CUDA
- 14,592
- Bandwidth
- 1,010 GB/s
GamingGeForce RTX 30 Series
GeForce RTX 3050
Ampere
- VRAM
- 6–8 GB
- CUDA
- 2,560
- Bandwidth
- 224 GB/s
GamingGeForce RTX 30 Series
GeForce RTX 3060
Ampere
- VRAM
- 8–12 GB
- CUDA
- 3,584
- Bandwidth
- 360 GB/s
GamingGeForce RTX 30 Series
GeForce RTX 3060 Ti
Ampere
- VRAM
- 8 GB
- CUDA
- 4,864
- Bandwidth
- 608 GB/s
GamingGeForce RTX 30 Series
GeForce RTX 3070
Ampere
- VRAM
- 8 GB
- CUDA
- 5,888
- Bandwidth
- 448 GB/s
GamingGeForce RTX 30 Series
GeForce RTX 3070 Ti
Ampere
- VRAM
- 8 GB
- CUDA
- 6,144
- Bandwidth
- 608 GB/s
GamingGeForce RTX 30 Series
GeForce RTX 3080
Ampere
- VRAM
- 10–12 GB
- CUDA
- 8,960
- Bandwidth
- 912 GB/s
GamingGeForce RTX 30 Series
GeForce RTX 3080 Ti
Ampere
- VRAM
- 12 GB
- CUDA
- 10,240
- Bandwidth
- 912 GB/s
GamingGeForce RTX 30 Series
GeForce RTX 3090
Ampere
- VRAM
- 24 GB
- CUDA
- 10,496
- Bandwidth
- 936 GB/s
GamingGeForce RTX 30 Series
GeForce RTX 3090 Ti
Ampere
- VRAM
- 24 GB
- CUDA
- 10,752
- Bandwidth
- 1,010 GB/s
GamingGeForce RTX 20 Series
GeForce RTX 2060
Turing
- VRAM
- 6–12 GB
- CUDA
- 2,176
- Bandwidth
- 336 GB/s
GamingGeForce RTX 20 Series
GeForce RTX 2060 SUPER
Turing
- VRAM
- 8 GB
- CUDA
- 2,176
- Bandwidth
- 448 GB/s
GamingGeForce RTX 20 Series
GeForce RTX 2070
Turing
- VRAM
- 8 GB
- CUDA
- 2,304
- Bandwidth
- 448 GB/s
GamingGeForce RTX 20 Series
GeForce RTX 2070 SUPER
Turing
- VRAM
- 8 GB
- CUDA
- 2,560
- Bandwidth
- 448 GB/s
GamingGeForce RTX 20 Series
GeForce RTX 2080
Turing
- VRAM
- 8 GB
- CUDA
- 2,944
- Bandwidth
- 448 GB/s
GamingGeForce RTX 20 Series
GeForce RTX 2080 SUPER
Turing
- VRAM
- 8 GB
- CUDA
- 3,072
- Bandwidth
- 496 GB/s
GamingGeForce RTX 20 Series
GeForce RTX 2080 Ti
Turing
- VRAM
- 11 GB
- CUDA
- 4,352
- Bandwidth
- 616 GB/s
WorkstationRTX PRO Blackwell
RTX PRO 2000 Blackwell
Blackwell 2.0
- VRAM
- 16 GB
- CUDA
- 4,352
- Bandwidth
- 288 GB/s
WorkstationRTX PRO Blackwell
RTX PRO 4000 Blackwell
Blackwell 2.0
- VRAM
- 24 GB
- CUDA
- 8,960
- Bandwidth
- 672 GB/s
WorkstationRTX PRO Blackwell
RTX PRO 4000 Blackwell SFF
Blackwell 2.0
- VRAM
- 24 GB
- CUDA
- 8,960
- Bandwidth
- 432 GB/s
WorkstationRTX PRO Blackwell
RTX PRO 4500 Blackwell
Blackwell 2.0
- VRAM
- 32 GB
- CUDA
- 10,496
- Bandwidth
- 896 GB/s
WorkstationRTX PRO Blackwell
RTX PRO 5000 Blackwell
Blackwell 2.0
- VRAM
- 48–72 GB
- CUDA
- 14,080
- Bandwidth
- 1,340 GB/s
WorkstationRTX PRO Blackwell
RTX PRO 6000 Blackwell
Blackwell 2.0
- VRAM
- 96 GB
- CUDA
- 24,064
- Bandwidth
- 1,790 GB/s
WorkstationRTX PRO Blackwell
RTX PRO 6000 Blackwell Max-Q
Blackwell 2.0
- VRAM
- 96 GB
- CUDA
- 24,064
- Bandwidth
- 1,790 GB/s
WorkstationRTX Ada Generation
RTX 2000 Ada Generation
Ada Lovelace
- VRAM
- 16 GB
- CUDA
- 2,816
- Bandwidth
- 256 GB/s
WorkstationRTX Ada Generation
RTX 4000 Ada Generation
Ada Lovelace
- VRAM
- 20 GB
- CUDA
- 6,144
- Bandwidth
- 360 GB/s
WorkstationRTX Ada Generation
RTX 4000 SFF Ada Generation
Ada Lovelace
- VRAM
- 20 GB
- CUDA
- 6,144
- Bandwidth
- 280 GB/s
WorkstationRTX Ada Generation
RTX 4500 Ada Generation
Ada Lovelace
- VRAM
- 24 GB
- CUDA
- 7,680
- Bandwidth
- 432 GB/s
WorkstationRTX Ada Generation
RTX 5000 Ada Generation
Ada Lovelace
- VRAM
- 32 GB
- CUDA
- 12,800
- Bandwidth
- 576 GB/s
WorkstationRTX Ada Generation
RTX 5880 Ada Generation
Ada Lovelace
- VRAM
- 48 GB
- CUDA
- 14,080
- Bandwidth
- 864 GB/s
WorkstationRTX Ada Generation
RTX 6000 Ada Generation
Ada Lovelace
- VRAM
- 48 GB
- CUDA
- 18,176
- Bandwidth
- 960 GB/s
WorkstationRTX A Series
RTX A1000
Ampere
- VRAM
- 8 GB
- CUDA
- 2,304
- Bandwidth
- 192 GB/s
WorkstationRTX A Series
RTX A2000
Ampere
- VRAM
- 6–12 GB
- CUDA
- 3,328
- Bandwidth
- 288 GB/s
WorkstationRTX A Series
RTX A400
Ampere
- VRAM
- 4 GB
- CUDA
- 768
- Bandwidth
- 96 GB/s
WorkstationRTX A Series
RTX A4000
Ampere
- VRAM
- 16 GB
- CUDA
- 6,144
- Bandwidth
- 448 GB/s
WorkstationRTX A Series
RTX A4500
Ampere
- VRAM
- 20 GB
- CUDA
- 7,168
- Bandwidth
- 640 GB/s
WorkstationRTX A Series
RTX A5000
Ampere
- VRAM
- 24 GB
- CUDA
- 8,192
- Bandwidth
- 768 GB/s
WorkstationRTX A Series
RTX A5500
Ampere
- VRAM
- 24 GB
- CUDA
- 10,240
- Bandwidth
- 768 GB/s
WorkstationRTX A Series
RTX A6000
Ampere
- VRAM
- 48 GB
- CUDA
- 10,752
- Bandwidth
- 768 GB/s
WorkstationQuadro RTX
Quadro RTX 4000
Turing
- VRAM
- 8 GB
- CUDA
- 2,304
- Bandwidth
- 416 GB/s
WorkstationQuadro RTX
Quadro RTX 5000
Turing
- VRAM
- 16 GB
- CUDA
- 3,072
- Bandwidth
- 448 GB/s
WorkstationQuadro RTX
Quadro RTX 6000
Turing
- VRAM
- 24 GB
- CUDA
- 4,608
- Bandwidth
- 672 GB/s
WorkstationQuadro RTX
Quadro RTX 8000
Turing
- VRAM
- 48 GB
- CUDA
- 4,608
- Bandwidth
- 672 GB/s
Data center / AITesla Turing
Tesla T4
Turing
- VRAM
- 16 GB
- CUDA
- 2,560
- Bandwidth
- 320 GB/s
Data center / AIRTX PRO Blackwell Server
NVIDIA RTX PRO 4500 Blackwell Server Edition
Blackwell
- VRAM
- 32 GB
- CUDA
- 10,496
- Bandwidth
- 800 GB/s
Data center / AIRTX PRO Blackwell Server
RTX PRO 6000 Blackwell Server
Blackwell 2.0
- VRAM
- 96 GB
- CUDA
- 24,064
- Bandwidth
- 1,790 GB/s
Data center / AIHopper Data Center
H100 NVL
Hopper
- VRAM
- 94 GB
- CUDA
- 16,896
- Bandwidth
- 3,940 GB/s
Data center / AIHopper Data Center
H100 PCIe
Hopper
- VRAM
- 80 GB
- CUDA
- 14,592
- Bandwidth
- 2,040 GB/s
Data center / AIHopper Data Center
H100 SXM5
Hopper
- VRAM
- 80 GB
- CUDA
- 16,896
- Bandwidth
- 3,360 GB/s
Data center / AIHopper Data Center
H200 NVL
Hopper
- VRAM
- 141 GB
- CUDA
- 16,896
- Bandwidth
- 4,890 GB/s
Data center / AIHopper Data Center
H200 SXM
Hopper
- VRAM
- 141 GB
- CUDA
- 16,896
- Bandwidth
- 4,890 GB/s
Data center / AIHopper Data Center
NVIDIA H20
Hopper
- VRAM
- 96 GB
- CUDA
- 9,984
- Bandwidth
- 4,000 GB/s
Data center / AIHopper Data Center
NVIDIA H20 NVL16
Hopper
- VRAM
- 96 GB
- CUDA
- 9,984
- Bandwidth
- 4,000 GB/s
Data center / AIBlackwell Data Center
B200
Blackwell
- VRAM
- 180 GB
- CUDA
- 18,944
- Bandwidth
- 8,000 GB/s
Data center / AIBlackwell Data Center
B300
Blackwell Ultra
- VRAM
- 288 GB
- CUDA
- 18,944
- Bandwidth
- 8,000 GB/s
Data center / AIAmpere Data Center
A10 PCIe
Ampere
- VRAM
- 24 GB
- CUDA
- 9,216
- Bandwidth
- 600 GB/s
Data center / AIAmpere Data Center
A100 PCIe
Ampere
- VRAM
- 40–80 GB
- CUDA
- 6,912
- Bandwidth
- 1,940 GB/s
Data center / AIAmpere Data Center
A100 SXM4
Ampere
- VRAM
- 40–80 GB
- CUDA
- 6,912
- Bandwidth
- 2,040 GB/s
Data center / AIAmpere Data Center
A16 PCIe
Ampere
- VRAM
- 16 GB
- CUDA
- 1,280
- Bandwidth
- 200 GB/s
Data center / AIAmpere Data Center
A2 PCIe
Ampere
- VRAM
- 16 GB
- CUDA
- 1,280
- Bandwidth
- 200 GB/s
Data center / AIAmpere Data Center
A30 PCIe
Ampere
- VRAM
- 24 GB
- CUDA
- 3,584
- Bandwidth
- 933 GB/s
Data center / AIAmpere Data Center
A40 PCIe
Ampere
- VRAM
- 48 GB
- CUDA
- 10,752
- Bandwidth
- 696 GB/s
Data center / AIAda Data Center
L4
Ada Lovelace
- VRAM
- 24 GB
- CUDA
- 7,424
- Bandwidth
- 300 GB/s
Data center / AIAda Data Center
L40
Ada Lovelace
- VRAM
- 48 GB
- CUDA
- 18,176
- Bandwidth
- 864 GB/s
Data center / AIAda Data Center
L40S
Ada Lovelace
- VRAM
- 48 GB
- CUDA
- 18,176
- Bandwidth
- 864 GB/s
Understand the condition and history of a GPU lot
Refurbished GPUs
Ask what was inspected, serviced or repaired and request test results. Refurbished does not automatically mean manufacturer-certified or covered by a warranty.
Modified / high-VRAM GPUs
Aftermarket memory upgrades and other hardware modifications are separate from refurbishment. Confirm the actual VRAM, firmware and compatibility.
Ex-corporate hardware, off-lease & liquidation lots
Possible sources include IT asset disposal (ITAD), data-center decommissioning, lease returns and business liquidation sales. Government or police-seized asset auction lots need documents proving lawful ownership and resale rights. These are source categories, not claims of available stock.
Before buying in bulk
Ask about mining history, repairs, quantities, serial numbers, tests and return terms. Ex-mining GPUs and faulty cards sold for parts or repair must be identified clearly; provenance is not a condition grade.