百度智能云
All Product Document

          GPU Compute Service

          • GPU Compute Service
          • >
          • Product Description
          • >
          • Cluster of GPU Instance Specifications
          • >
          • Compute Optimized Specification Cluster

          Compute Optimized Specification Cluster

          Last Updated：2021-11-20

          Compute Optimized GPU

          Cluster of Specifications GPU series Availability zone
          Compute optimized GN3 Ampere A100 East China-Beijing
          Volta V100 North China - Beijing, North China - Baoding
          Turing T4 North China - Beijing, North China-Baoding, South China-Guangzhou and East China-Suzhou
          Compute optimized GN2 Volta V100 South China-Guangzhou, East China-Suzhou
          Compute optimized GN1 Pascal P4 North China - Beijing, North China-Baoding, South China-Guangzhou and East China-Suzhou
          Compute optimized vGN3-C Volta V100 East China-Beijing
          Turing T4 East China-Beijing

          ## Compute Optimized GN3

          With the NVIDIA mainstream data center GPU, you can have access to strong computing power for high-density computing scenarios. According to the actual scenario needs, the GPU accelerator card can be NVIDIA A100, NVIDIA V100 SXM, or NVIDIA T4.

          Ampere A100 series

              This series is under test and is not yet available for online sale.

          • Applicable Scenarios:

            • Large-scale AI training: such as image recognition, target detection, unmanned driving, semantic analysis, and ads recommendation scenarios.
            • High-performance computing (HPC): life science, CAE, meteorological prediction, and collision simulation scenarios.
            • High-performance data analysis: data set visualization and big data analysis scenarios.

          • Features of of Specifications：

            • Processor: Intel Xeon Gold 6271C, high clock speed 2.6GHz
            • GPU: NVIDIA A100, FP16 computing power up to 312TFLOPS, single GPU video memory 40GB HBM2.
            • Memory: DDR4, memory bandwidth 2933MT/s, CPU memory ratio: 1:8.

            • Storage: Support CDS cloud disk

              • Network: The network performance of instances corresponds to the specification, and the upper limit of internal bandwidth is higher.

          Instance Specifications

          Instances Number of A100 GPU card cpu Memory size（ GB） Private Bandwidth (Gbps) Network packet receiving and sending (10K pps) Number of columns
          bcc.lgn3.c10m80.1A100 1 pc 10 core 80 4 150 8
          bcc.lgn3.c20m160.2A100 2 pc 20 core 160 7 280 8
          bcc.lgn3.c40m320.4A100 4 pc 40 core 320 13 500 16
          bcc.lgn3.c80m640.8A100 8 pc 80 core 640 25 700 16

          ### Volta V100 Series

          • Applicable Scenarios:

            • AI training and large-scale reasoning: such as image recognition, target detection, unmanned driving, semantic analysis, and ads recommendation scenarios.
            • High-performance computing (HPC): life science, CAE, meteorological prediction, and collision simulation scenarios.
            • High-performance data analysis: data set visualization and big data analysis scenarios.
            • Codec: video streaming and picture stream codec scenarios.

          • Features of of Specifications：

            • Processor: Intel Xeon Gold 6271C, high clock speed 2.6GHz
            • GPU: NVIDIA V100 SXM2, FP16 computing power up to 125TFLOPS, single GPU video memory 32GB HBM2, GPU interconnection supporting bandwidth 300GB/s NVLink.
            • Memory: DDR4, memory bandwidth 2933MT/s, CPU memory ratio: 1:8.
            • Storage: Support CDS cloud disk
            • Network: The network performance of instances corresponds to the specification, and the upper limit of internal bandwidth is higher.

          Instance Specifications

          Instances Number of V100-32GB GPU card Number of vcpus Memory size（ GB） Private Bandwidth (Gbps) Network packet receiving and sending (10K pps) Number of columns
          bcc.lgn3.c10m80.1v100-32g 1 pc 10 core 80 4 150 8
          bcc.lgn3.c20m160.2v100-32g 2 pc 20 core 160 7 280 8
          bcc.lgn3.c40m320.4v100-32g 4 pc 40 core 320 13 500 16
          bcc.lgn3.c80m640.8v100-32g 8 pc 80 core 640 25 700 16

          ### Turing T4 Series

          • Applicable Scenarios:

            • AI reasoning: such as image recognition, target detection, unmanned driving, semantic analysis, and ads recommendation scenarios.
            • High-performance data analysis: data set visualization and big data analysis scenarios.
            • Codec: video streaming and picture stream codec scenarios.

          • Features of of Specifications：

            • Processor: Intel Xeon Gold 6271C, high clock speed 2.6GHz
            • GPU: NVIDIA T4, INT8 computing power up to 130TFLOPS, single GPU video memory 16GB
            • Memory: DDR4, memory bandwidth 2666MT/s, CPU memory ratio: 1:4.
            • Storage: Support CDS cloud disk
            • Network: The network performance of instances corresponds to the specification, and the upper limit of internal bandwidth is higher.

          Instance Specifications

          Instances Number of T4 GPU card Number of vcpus Memory size（ GB） Private Bandwidth (Gbps) Network packet receiving and sending (10K pps) Number of columns
          bcc.gn3.c16m64.1t4 1 pc 16 core 80 6 280 8
          bcc.gn3.c32m128.2t4 2 pc 32 core 160 11 500 8
          bcc.gn3.c64m256.4t4 4 pc 64 core 320 22 700 16
          bcc.gn3.c20m80.1t4 1 pc 20 core 80 6 280 16
          bcc.gn3.c40m160.2t4 2 pc 40 core 160 11 700 16
          bcc.gn3.c40m160.2t4 4 pc 80 core 320 22 700 16

          ## Compute Optimized GN2

          With the NVIDIA mainstream data center GPU V100 SXM2 16GB, you can have access to strong computing power for high-density computing scenarios.

          • Applicable Scenarios:

            • AI training and reasoning: such as image recognition, target detection, unmanned driving, semantic analysis, and ads recommendation scenarios.
            • High-performance computing (HPC): life science, financial service, meteorological prediction, and collision simulation scenarios.
            • High-performance data analysis: data set visualization and big data analysis scenarios.
            • Codec: video streaming and picture stream codec scenarios.

          • Instance Specifications:

            • Processor: Processor: Intel® Xeon® Skylake 6148, high clock speed 2.4GHz
            • GPU: NVIDIA V100 SXM2, FP16 computing power up to 125TFLOPS, single GPU video memory 16GB HBM2, GPU interconnection supporting bandwidth 300GB/s NVLink.
            • Memory: DDR4, memory bandwidth 2,666MT/s. The CPU memory ratio can be 1:4 or 1:8.
            • Storage: Support CDS cloud disk

          Instance Specifications

          Instances Number of V100-16G GPU card Number of vcpus Memory size（ GB）
          bcc.lgn2.c8m32.1v100-16g 1 pc 8 core 32
          bcc.lgn2.c16m64.2v100-16g 2 pc 16 core 64
          bcc.lgn2.c32m128.41v100-16g 4 pc 32 core 128
          bcc.lgn2.c64m256.8v100-16g 8 pc 64 core 256
          bcc.lgn2.c8m64.1v100-16g 1 pc 8 core 64
          bcc.lgn2.c16m128.2v100-16g 2 pc 16 core 128
          bcc.lgn2.c32m256.4v100-16g 4 pc 32 core 256

          ## Compute optimized GN1

          With the NVIDIA mainstream data center GPU P4, it is suitable for AI reasoning in the data center.

          Pascal P4 Series

          • Applicable Scenarios:

            • AI reasoning: such as image recognition, target detection, and semantic segment scenarios.
            • Data analysis: data set visualization and big data analysis scenarios.
            • Codec: video streaming and picture stream codec scenarios.

          • Features of of Specifications：

            • Processor: Intel Xeon Gold 6271C, high clock speed 2.6GHz
            • GPU: NVIDIA P4, integer computing power up to 22TFLOPS, single GPU video memory 8GB
            • Memory: DDR4, memory bandwidth 2666MT/ s.
            • Storage: Support CDS cloud disk

          Instance Specifications

          Instances Number of P4 GPU card Number of vcpus Memory size（ GB）
          bcc.lgn1.c6m24.1p4 1 pc 6 core 24
          bcc.lgn1.c12m40.1p4 2 pc 12 core 40
          bcc.lgn1.c16m64.1p4 4 pc 16 core 64
          bcc.lgn1.c24m80.2p4 2 pc 24 core 80
          bcc.lgn1.c48m160.4p4 4 pc 48 core 160

          ## Compute Optimized vGN3- C 
          The specification family is currently opened in the form of the white list. You can submit the ticket if necessary.

          With the NVIDIA mainstream data center, multipart virtualized vGPU, it improves GPU's cost performance and resources utilization.

          Volta V100 Series

          This instance is suitable for computing scenarios, such as AI computing and development machines. If you demand that the instance supports the overall graphics and image functions, such as OpenGL or Windows DirectX, please use the Baidu AI Cloud rendering GPU instance vGN3-Q.

          • Applicable Scenarios:

            • AI reasoning: such as image recognition, target detection, unmanned driving, semantic analysis, and ads recommendation scenarios.
            • Small-scale AI training: such as image recognition and target detection scenarios.
            • AI model development and teaching environment

          • Instance Specifications:

            • Processor: Intel Xeon Gold 6271C, high clock speed 2.6GHz
            • GPU: NVIDIA® Tesla® V100 vGPU, GPU memory supports 8G or 16G
            • Memory: DDR4, memory bandwidth 2666MT/s, CPU memory ratio： 1:6.
            • Storage: Support CDS cloud disk
            • Network: The network performance of instances corresponds to the specification, and the upper limit of internal bandwidth is higher.

          Instance Specifications

          Instances Number of V100 GPU card Video memory size (GB) Number of vcpus Memory size（ GB） Private bandwidth (Gb) Network packet receiving and sending (10K pps)
          bcc.vgn3.c2m12.1v100-32g-8C 1/4 pcs 8 2 12 1.5 40
          bcc.vgn3.c4m24.1v100-32g-16C 1/2 pcs 16 4 24 1.5 80

          ### Turing T4 Series

          This instance is suitable for computing scenarios, such as AI reasoning. If you demand that the instance supports the overall graphics and image functions, such as OpenGL or Windows DirectX, please use the Baidu AI Cloud rendering GPU instance vGN3-Q.

          • Applicable Scenarios:

            • AI reasoning: such as image recognition, target detection, unmanned driving, semantic analysis, and ads recommendation scenarios.
            • AI model development and teaching environment

          • Instance Specifications:

            • Processor: Intel Xeon Gold 6271C, high clock speed 2.6GHz
            • GPU: NVIDIA T4 vGPU, single instance GPU video memory supporting 4 GB or 8 GB.
            • Memory: DDR4, memory bandwidth 2666MT/s, CPU memory ratio: 1:4.
            • Storage: Support CDS cloud disk
            • Network: The network performance of instances corresponds to the specification, and the upper limit of internal bandwidth is higher.

          Instance Specifications

          Instances Number of T4 GPU card Video memory size (GB) Number of vcpus Memory size（ GB） Private bandwidth (Gb) Network packet receiving and sending (10K pps)
          bcc.vgn3.c4m16.1t4-4C 1/4 pcs 4 4 16 1.5 80
          bcc.vgn3.c10m40.1t4-8C 1/2 pcs 8 10 40 4 150
          Previous
          GPU Card Details and Application Scenarios
          Next
          Render Type Instance Specification