AWS G6 instance types

G6 is a gpu & accelerated Amazon EC2 family with 8 instance types, from 4 – 192 vCPUs and 16 GiB – 768 GiB of memory with NVIDIA L4 accelerators, running on AMD EPYC 7R13 Processor. Linux on demand starts at $0.8048 per hour.

gpu & acceleratedx86-64
sizes
8
vCPUs
4 – 192
memory
16 GiB – 768 GiB
accelerator
L4
on-demand from
$0.8048

per hour · 20 regions

G6 sizes and prices

Linux, hourly, US East (N. Virginia) (other regions on each size's page)

SizevCPUMemoryAcceleratorLocal storageOn-demandMonthlySpot1-yr
g6.xlarge416 GiB1 × NVIDIA L4250 GB NVME$0.8048$588$0.105$0.5239
g6.2xlarge832 GiB1 × NVIDIA L4450 GB NVME$0.9776$714$0.1259$0.6364
g6.4xlarge1664 GiB1 × NVIDIA L4600 GB NVME$1.323$966$0.1723$0.8614
g6.8xlarge32128 GiB1 × NVIDIA L4900 GB NVME (2 disks)$2.014$1,471$0.2623$1.311
g6.12xlarge48192 GiB4 × NVIDIA L43.8 TB NVME (4 disks)$4.602$3,359$0.6045$2.996
g6.16xlarge64256 GiB1 × NVIDIA L41.9 TB NVME (2 disks)$3.397$2,480$0.4404$2.211
g6.24xlarge96384 GiB4 × NVIDIA L43.8 TB NVME (4 disks)$6.675$4,873$0.8597$4.346
g6.48xlarge192768 GiB8 × NVIDIA L47.5 TB NVME (8 disks)$13.35$9,746$1.76$8.691

G6 equivalents on other clouds

closest match to g6.8xlarge (32 vCPU, 128 GiB)

AzureNV24ads V710 v5NVadsV710v5 family · 24 vCPU · 128 GiB · 1 × AMD Radeon Pro V710$2.00/hr · East USNVadsV710v5 family →
Google Cloudg2-standard-32G2 family · 32 vCPU · 128 GiB · 1 × NVIDIA L4$1.734/hr · IowaG2 family →
AkamaiRTX4000 Ada x1 X-LargeGPU (RTX 4000 Ada) family · 32 vCPU · 128 GiB · 1 × NVIDIA RTX 4000 Ada$1.53/hr · Newark, NJGPU (RTX 4000 Ada) family →

Other gpu & accelerated families on AWS

G6 FAQ

What sizes does the AWS G6 family have?

G6 has 8 instance types, from g6.xlarge (4 vCPU, 16 GiB) to g6.48xlarge (192 vCPU, 768 GiB). They run on AMD EPYC 7R13 Processor.

What is the cheapest G6 instance type?

g6.xlarge at $0.8048 per hour (Linux, on demand) in US East (N. Virginia), about $588 per month.

What is G6 best for?

G6 is a gpu & accelerated family. GPUs and AI accelerators for training, inference, rendering and video.