GPU Elastic Physical Compute provides services including hardware (GPU, CPU, memory, and hard disk) and network. The price of the instance hardware is in the table below.
|GPU Standard Compute||P3E (Series 3)||
It provides a total of 48TFlops single-precision floating-point computing power and 96GB of memory, which can be used for large-scale neural network training and online reasoning.
Currently, two types are supported: GPU physical instance standard type and GPU physical instance standard cluster type.
|GPU Inference Compute||P3IE (Series 3)||
Based on NVIDIA Tesla P4 GPU, it satisfies the computational power requirements in deep learning (especially inference) scenarios with single-precision floating-point computing power of up to 44 TFLOPS and INT8 processing capability of 176 TOPS. At the same time, the single-card power consumption is only 75W, which has a very high performance-to-power ratio.
|GPU Standard Compute||P4E (Series 4)||Based on NVIDIA's Tesla V100 GPU, P4E has superior computing power, supports single fine floating point computing capability of up to 14*8TFLOPS and mixed precision (FP16/FP32) matrix computing capability of 112 * 8TFLOPS, providing solutions for deep learning and reasoning.|
|P3E||E5-2690 V4*2||128||800G SSD*8||Nvidia_P40*4||RAID 5/50/Single Disk RAID 0||4,179.11
||E5-2690 V4*2||512||800G SSD*8||Nvidia_P4*4||RAID 10/5/50||2,611.94|
|P4E.56F8||E5-2690 V4*2||512||960G SSD*6||Nvidia_V100*8||RAID 5/50||8,358.21|