Nexus/Vulcan/GPUs

There are several different types of NVIDIA GPUs in the Vulcan partitions in the Nexus cluster, listed in order of newest to oldest architecture.

Quadro RTX A6000

There are 36 RTX A6000 cards available to be scheduled in Vulcan. These cards can be requested with the --gres=gpu:rtxa6000 submission argument(s).

• Ampere architecture
• 10752 Cuda cores
• 48 GB GDDR6 Memory
• 40 TFLOPS (FP32)

Quadro RTX A5000

There are 40 RTX A5000 cards available to be scheduled in Vulcan. These cards can be requested with the --gres=gpu:rtxa5000 submission argument(s).

• Ampere architecture
• 8192 Cuda cores
• 24 GB GDDR6 Memory
• 27.8 TFLOPS (FP32)

Quadro RTX A4000

There are 56 RTX A4000 cards available to be scheduled in Vulcan. These cards can be requested with the --gres=gpu:rtxa4000 submission argument(s).

• Ampere architecture
• 6144 Cuda cores
• 16GB GDDR6 Memory
• 19.2 TFLOPS (FP32)

GeForce RTX 2080 Ti

There are 28 RTX 2080 Ti cards available to be scheduled in Vulcan. These cards can be requested with the --gres=gpu:rtx2080ti submission argument(s).

• Turing architecture
• 4352 Cuda cores
• 11 GB GDDR5X Memory
• 13.45 TFLOPS (FP32)

GeForce GTX 1080 Ti

There are 128 GTX 1080 Ti cards available to be scheduled in Vulcan. These cards can be requested with the --gres=gpu:gtx1080ti submission argument(s).

• Pascal architecture
• 3584 Cuda cores
• 11 GB GDDR5X Memory
• 11.3 TFLOPS (FP32)

Quadro P6000

There are 64 P6000 cards available to be scheduled in Vulcan. These cards can be requested with the --gres=gpu:p6000 submission argument(s).

• Pascal architecture
• 3840 Cuda cores
• 24 GB GDDR5X Memory
• 12 TFLOPS (FP32)

Tesla P100

There is one P100 card available to be scheduled in Vulcan. This card can be requested with the --gres=gpu:p100 submission argument(s).

• Pascal architecture
• 3584 Cuda cores
• 16GB CoWoS HBM2 Memory
• 9.3 TFLOPS (FP32)