Ebmgn7

vgn5i

vgn6i

gn5i

gn5

gn6i

gn6v

gn6e

sccgn6

 타입/목적

X-Dragon-based 

GPU-accelerated

instances

 Lightweight

GPU-accelerated 

instances

Lightweight

GPU instances

General-purpose

GPU-accelerated

instances

General-purpose

GPU-accelerated

instances 

 General-purpose

GPU-accelerated

instances 

 General-purpose

GPU-accelerated

instances 

 General-purpose

GPU-accelerated

instances 

 supercomputing

GPU-accelerated

instances 

 가상화 레벨

 Bare metal

Slice virtualization

Slice virtualization

Pass-through

virtualization  

Pass-through

virtualization

Pass-through

virtualization

Pass-through

virtualization

Pass-through

virtualization

Pass-through

virtualization

 GPU모델

8 x A100 

(1/8-1) x P4

(1/4-1/2) x T4

1/2 x P4

1/2/4/8 x P100

1/2/4 x T4

1/4/8 x V100

1/4/8 x V100

8 x V100

 GPU vRAM

(GB)

40

1,2,4,8

4,8

1/2 x 8

1/2/4/8 x 16

1/2/4 x 16

1/4/8 x 16

1/4 x 32

8 x 16

Computing Power

AAAAA

A

A

AA

AAA

AAA

AAAA

AAAA

AAAA

네트워크 타입

VPC

VPC

VPC

VPC

VPC

VPC

VPC

VPC

VPC+RDMA Latency

aslow as 2μs

 대역폭

30G

10G

25G

10G

25G

25G

25G

25G

25G + 50G

내부 통신모델

NVLink

PCle

PCle

PCle

PCle

PCle

NVLink

NVLink

NVLink

Video 인코딩

및 디코딩

supported

supported

supported

supported

supported

supported

supported

supported

supported