Compare GCP Instances

Instance Type	a4-highgpu-8g	a2-highgpu-8g
Details	Optimized for accelerated high performance computing workloads. a – Accelerator optimized 4 – Generation highgpu – High amound of GPU memory - 40x RAM/GPU ratio 8g – Number of GPUs	Optimized for accelerated high performance computing workloads. a – Accelerator optimized 2 – Generation highgpu – High amound of GPU memory - 40x RAM/GPU ratio 8g – Number of GPUs
Instance Family	Accelerator optimized	Accelerator optimized
Description	Accelerator Optimized: 8 NVIDIA B200 GPU, 224 vCPUs, 3968GB RAM	Accelerator Optimized: 8 NVIDIA Tesla A100 GPUs, 96 vCPUs, 680GB RAM
vCPUs	224	96
Memory	3968 GiB	680 GiB
CPU Architecture	x86_64	x86_64
Processor	Intel Emerald Rapids	Intel Cascade Lake
Coremark Score		1269327
Shared CPU	no	no
GPU	yes	yes
GPUs	8	8
GPU Name	nvidia-b200	nvidia-tesla-a100
GPU Total RAM (GiB)		680
Sustained Discounts (SUDs)	no	no
Sole Tenancy	no	yes
Nested Virtualization	no	yes
Custom Shapes	no	no
Confidential Compute	no	no
Maximum Persistent Disks	128	128
Maximum Persistent Disks Size	524288 GB	524288 GB
Disk Interface Type	NVMe	SCSI (PD, Local SSD) NVMe (Local SSD)
Local SSD	yes	yes
Max Local SSD	12000 GB	3000 GB
Extreme Persistent Disks	no	no
Hyperdisk Extreme	yes	no
Hyperdisk Throughput	no	no
Network Interfaces	gVNIC, MRDMA	gVNIC, VirtIO-Net
Avg Price Diff		Infinityx

Regional Prices

Geography	Region	a4-highgpu-8g	a2-highgpu-8g
Hidden	Hidden (hidden-1)	n/a	29.9838
Hidden	Hidden (hidden-2)	n/a	31.4748
Hidden	Hidden (hidden-3)	n/a	32.1101
Hidden	Hidden (hidden-4)	n/a	32.3258
Hidden	Hidden (hidden-5)	n/a	32.3967
Hidden	Hidden (hidden-6)	n/a	32.3967
Hidden	Hidden (hidden-7)	n/a	34.6005
North America	Iowa (us-central1)	n/a	29.3871
North America	Oregon (us-west1)	n/a	29.3871
North America	South Carolina (us-east1)	n/a	29.3871