Slide 10
Slide 10 text
GPUラインアップ 一覧
10 Copyright © 2025, Oracle and/or its affiliates
Bare Metal
192 GB
GPU Memory
4 x L40S
GPU.L40S.4
VM & Bare Metal
320 GB
GPU Memory
8 x A100
40GB
GPU.4.8
100 Gb/sec
Front-end network
VM & Bare Metal
24 GB
48 GB
GPU Memory
GPU.A10.1
GPU.A10.2
GPU.A10.4
1 x A10
2 x A10
Bare Metal
768 GB
GPU Memory
72 x B200
192GB
GPU.GB200.4
NVL72
(18 servers/rack)
7,200 Gb/sec
Front-end network
36 x Grace
CPU
Bare Metal
1,128 GB
GPU Memory
30.7 TB
Local NVMe
200 Gb/sec
Front-end network
8 x H200
141GB
GPU.H200.8
Bare Metal
1,536 GB
GPU Memory
30.7 TB
Local NVMe
400 Gb/sec
Front-end network
8 x B200
192GB
GPU.B200.8
Bare Metal
640 GB
GPU Memory
8 X A100
80GB
GPU.A100-v2.8
Bare Metal
8 x H100 80GB
GPU.H100.8
640 GB
GPU Memory
61.4 TB
Local NVMe
100 Gb/sec
Front-end network
4月GA
1.7 TB
CPU+GPU Memory
15.3 TB
Local NVMe
4月GA
Bare Metal
8 x MI300X
192GB
GPU.MI300X.8
1,536 GB
GPU Memory
30.7 TB
Local NVMe
100 Gb/sec
Front-end network
Bare Metal
2,304 GB
GPU Memory
8 x 2 B300
288GB
GPU.B300
NVL16
予約受付中
Bare Metal
20.7 TB
GPU Memory
72 x B300
GPU.GB300.4
NVL72
(18 servers/rack)
7,200 Gb/sec
Front-end Network
36 x Grace
CPU
38.7 TB
CPU+GPU Memory
553 TB
Local NVMe
予約受付中
小 → 大規模
3,840GPU
800 Gb/sec
RDMA
1,600 Gb/sec
RDMA
200 Gb/sec
Front-end network
100 Gb/sec
Front-end network
32,768GPU 16,384GPU
3,200 Gb/sec RDMA
16,384GPU 65,536GPU 131,072GPU 131,072GPU
28,800 Gb/sec
RDMA
6,400 Gb/sec
RDMA
57,600 Gb/sec
RDMA
OCI Object & File Storage FSS HPMT—20/40/80 Gb/sec per MT | Managed Lustre—1 Gb/sec per TB
AI向けストレージ
AI向けネットワーク: OCI Supercluster