Strong FP32/FP64; good for training and batch jobs.
VRAM 16 GB
FP16 18.7 TF
Bandwidth 0.73 TB/s Pascal architecture FP32 9.3 TF
AP South AP Southeast EU West +1 more
High priorityAvailability: Available
Up to 4 reservable · SLA
− 0 +
Low priorityAvailability: Available
Up to 5 · shared queue
− 0 +
Efficient inference for small and mid-size models.
VRAM 16 GB
FP16 65 TF
Bandwidth 0.32 TB/s Turing architecture FP32 8.1 TF
AP Northeast AP South EU West +2 more
High priorityAvailability: Available
Up to 4 reservable · SLA
− 0 +
Low priorityAvailability: Available
Up to 5 · shared queue
− 0 +
Large models, fine-tuning and high-throughput inference.
VRAM 80 GB
FP16 312 TF
Bandwidth 2.0 TB/s Ampere architecture FP32 19.5 TF
High priorityAvailability: Available
Up to 2 reservable · SLA
− 0 +
Low priorityAvailability: Available
Up to 4 · shared queue
− 0 +