A wide range of IaaS compute options with the best price-performance available

Compute options range from VMs to GPUs to bare metal servers, and includes options for dense I/O workloads, high performance computing (HPC), and AMD EPYC processors.

Oracle Cloud Infrastructure Compute offers significant price-performance and control improvements compared to on-premises data centers, while providing the elasticity and cost savings of the public cloud. Oracle has a longstanding reputation for providing technologies that empower enterprises to solve demanding business problems––and Oracle Cloud Infrastructure is the first cloud that's purpose-built to enable enterprises to run business-critical production workloads.



Avoid the performance overhead associated with virtualization that impacts other public clouds. Oracle’s approach to cloud networking ensures servers running virtual machines (VMs) aren't taxed by network virtualization operations. Oracle does not oversubscribe core counts or networks so users do not have to worry about noisy neighbors. Oracle also provides high-core, specialty compute instances and 100 Gbps of networking for remote direct memory access (RDMA) with high performance computing (HPC) instances.


Oracle offers low, predictable pricing and improved price-performance compared to other public clouds. Oracle counts processors by Oracle compute units (OCPUs)––and each OCPU include the complete processor core, including the hyper-threading unit. Oracle does not count processors by virtual central processing units (vCPUs), fractional central processing units (CPUs), oversubscribed cores, or similar constructs used by competing public clouds. The price per OCPU is the same across VMs and bare metal compute instances. AMD EPYC processor-based compute instances are about 65% less than comparable instances on other clouds.


Compute instance configurations and pools make it easy to standardize configurations and manage groups of identical compute instances. Autoscaling provides the ability to scale a pool of resources up and down automatically in response to changing business demands and, as a result, pay only for the resources they need. AMD EPYC processor-based compute instances provide the highest memory bandwidth, and lowest cost per core-hour. From single-processor VMs to 64-core bare metal servers, Oracle offers cloud compute options that meet every need.


Deploy applications across multiple regions, multiple availability domains (ADs), and multiple fault domain configurations. Each AD is housed in a highly secure building and provides fully independent resources like power generators, cooling equipment, networking, and servers, while fault domains partition compute resources within each AD for additional protection.


Oracle supports an expansive list of Microsoft Windows Server and enterprise Linux operating systems, including Oracle Linux, CentOS, Ubuntu, and many others. Customers can run their own operating systems and hypervisors on Oracle compute instances––and they can use the same image across configuration sizes. Users can get business applications images from Oracle or its robust ecosystem of third-party partners.


Cloud compute instances are available in two main categories: VMs and bare metal machines. Both groups offer a dense I/O option for high performance local NVMe SSD storage. Both groups are also available with NVIDIA Tesla GPU's (Pascal or Volta) for HPC workloads. All instances run on the same virtual cloud networks with up to 1 PB per instance, and networks are accessible through the same console and API set.
Virtual Machine Standard
1 - 24 full cores on Intel Xeon or AMD EPYC processors
15 - 320 GB RAM
Up to 1 PB Block Storage
Up to 24.6 Gbps Network Bandwidth
Virtual Machine Dense I/O
8 - 24 full cores on Intel Xeon processors
120 - 320 GB RAM
6.4 - 25.6 TB of local NVMe SSD
Up to 1 PB Block Storage
Up to 24.6 Gbps Network Bandwidth
虛擬機器 GPU
6 - 24 個核心、90 - 360 GB RAM
1 - 4 個 Tesla GPU (Pascal 或 Volta)
最高可達 1 PB 區塊儲存空間
最高可達 25 Gbps 網路介面
Bare Metal Standard
52 or 64 full cores on Intel Xeon or AMD EPYC processors
512-768 GB RAM
Up to 1 PB Block Storage
2x 25 Gbps Network Interfaces
Bare Metal Dense I/O
52 Cores, 768 GB RAM
51.2 TB of local NVMe SSD
Up to 1 PB Block Storage
2x 25 Gbps Network Interfaces
裸機 GPU
28 - 52 個核心、192 - 768 GB RAM
2 - 8 個 Tesla GPU (Pascal 或 Volta)
最高可達 1 PB 區塊儲存
2x 25 Gbps 網路介面

Oracle Cloud Infrastructure Compute - Use Cases

Oracle Cloud Infrastructure Compute instances serve as the building blocks for business applications––from the smallest websites to the biggest enterprise workloads. Take a look at the main use cases for each of Oracle Cloud Infrastructure Compute shape.

一般工作負載 - VM

虛擬機器 (VM) 執行處理提供許多資源配置的運算資源,從單一 OCPU 到 24 個 OCPU 都有,能夠滿足各種工作負載和軟體架構需求。所有 Oracle Cloud Infrastructure VM 資源配置都支援遠端區塊儲存,但密集 I/O 資源配置特別針對需要低延遲、數百萬次 IOPS 及高本機儲存容量的應用系統,提供最高可達 25.6 TB 的本機 NVMe SSD 儲存空間。請參閱虛擬機器執行處理功能 >>

需要大量運算和儲存的工作負載 - 裸機

裸機執行處理支援需要大量核心、大量記憶體及高記憶體頻寬的應用系統。您可以建置效能等同或優於其他雲端或企業內部部署基礎架構的雲端環境。裸機運算執行處理可為客戶提供優異的隔離環境、資訊能見度及控制。請參閱裸機執行處理功能 >>

客製化應用系統工作負載 - 裸機

裸機運算執行處理具備機外網路,可在應用系統支援方面發揮極致的多功能性。「自備映像檔」可讓您在裸機執行處理上執行任何作業系統或虛擬機器管理程式。由於與完全由軟體定義的 Oracle Cloud Infrastructure 交換式網路架構相互結合,因此能提供無與倫比的處理能力,從傳統到最尖端的應用系統堆疊支援一應俱全。請參閱自備映像檔文件 >>

人工智慧和 HPC 工作負載 - GPU

若要加速運算,每項服務必須具備一貫快速的基礎架構。藉由 GPU 執行處理,您將能以更有效率的方式處理及分析大量資料集,讓這些資料集適用於複雜的機器學習 (ML)、人工智慧 (AI) 演算法及許多工業 HPC 應用系統。請參閱裸機 GPU 執行處理功能 >>

Oracle Cloud Infrastructure Compute Customer Success Stories

Frozen Mountain
「我們需要性價比最佳的可用服務。與先前的解決方案相比,效能提昇了一倍,也就意謂著成本的降低。使用 Oracle Cloud Infrastructure 就是能夠獲得更佳的運算能力。」
Anton Venema | CTO
"For our customer workloads, scalable performance in the cloud is driven by the network and storage architecture. In both those areas, Oracle Cloud Infrastructure is delivering a significant step change in the price-performance available to our HPC Cloud customers."
Jamil Appa | 共同創辦人暨協理
Oracle Data Cloud
"Oracle Cloud Infrastructure hardware is simply on another level, from the bare metal servers with NVMe SSD drives to the network backbone. We obtain fabulous performance with very low latency for our end customers. Dense I/O compute instances with 25 Gbps of bandwidth and millions of read and write IOPS is a game changer for big data workloads like ours."
Matt Abrams | VP of Engineering


選擇適合您需求的資源配置,包括具有 GPU 和 X7 型執行處理的高效能運算
觀看這個 10 分鐘 Oracle Cloud Infrastructure Compute 基本構成要素簡介的影片
1-800-633-0738 (美國)

1-800-633-0738 (美國)

Oracle Cloud 討論論壇