Webinar: Learn about the price-to-performance advantage of AMD MI355X GPUs on Vultr
with 48-month contract, 730 hours / month
Exceptional acceleration for the world’s most demanding AI and HPC workloads
Key Features
-
Next-generation architecture
Powered by the cutting-edge AMD CDNA 4 architecture
-
Expanded datatype support
Supports FP4 and FP6 datatypes and enhanced FP16 and FP8 processing for maximizing computational throughput
-
Increased efficiency
288 GB HBM3E memory and 8 TB/s memory bandwidth per GPU ensures larger models can be processed in-memory on each GPU, while direct liquid cooling decreases energy requirements
-
Built-in security
Device Secure Boot helps ensure only trusted firmware runs, while AMD Infinity Fabric™ Link security protects high-speed GPU to GPU networking
-
Open software stack
Harnesses the latest AMD ROCm™ open software stack, the premier alternative for AI and HPC
Process more data in-memory for greater efficiency and high performance. Download the datasheet to learn how.
No information is required for download
Enterprise-ready at
any scale and any location
Clusters at any size
Vultr's enterprise-ready infrastructure seamlessly supports any cluster size of AMD GPUs. Whether you require a small cluster or a massive deployment, Vultr ensures reliable, high-performance computing to meet your specific needs.
Globally available, locally accessible
Large clusters of AMD GPUs are available where you need them, thanks to Vultr's extensive infrastructure. With 32 cloud data center regions across six continents, we guarantee low latency and high availability, enabling your enterprise to achieve optimal performance worldwide.
Enterprise-grade compliance and security
Vultr ensures our platform, products, and services meet diverse global compliance, privacy, and security needs, covering areas such as server availability, data protection, and privacy. Our commitment to industry-wide privacy and security frameworks demonstrates our dedication to protecting our customers' data.
Powered by AMD ROCm™
open software ecosystem
AMD ROCm™ is the only open software platform for GPU computing, ensuring freedom from vendor lock-in and from limited architectural options.
AMD ROCm™ works out of the box and is supported on hundreds of open-source platforms, letting developers design the solutions they need to advance their models without the concerns of being locked into a proprietary solution.
AMD MI355X Platform
Specifications
- AMD MI355X GPUs
- 8
- Form factor
- UBB 2.0 module
- GPU compute units
- 2048
- Matrix cores
- 8192
- Peak engine clock
- 2200 MHz
- Memory capacity
- Up to 288 GB HBM3E per GPU
- Memory bandwidth
- 8.0 TB/s max. peak theoretical per GPU
- Memory interface
- 8192 bits
- AMD Infinity Cache™ (last level)
- 256 MB per GPU
AI peak theoretical performance
with sparsity
- FP16 (PFLOPs)
- 20.1
- 40.3
- BFLOAT16 (PFLOPs)
- 20.1
- 40.3
- INT8 (PFLOPs)
- 40.3
- 80.5
- INT4 (POPs)
- 40.3
- 80.5
- FP8 (PFLOPs)
- 40.3
- 80.5
- FP6 (PFLOPs)
- 80.5
- 161.1
- FP4 (PFLOPs)
- 80.5
- 161.1
HPC peak theoretical performance
(TFLOPS)
- FP64 vector
- 628.8
- FP32 vector
- 1.3
- FP64 matrix
- 628.8
- FP32 matrix
- 1.3
Customer success stories and use cases
Explore how leading organizations in the manufacturing and energy industry, equipped with the right tools, achieve security, connectivity, and efficiency using Vultr’s cloud solutions.
Access advanced hardware without the typical complexities
Discover how Vultr's integration of AMD MI355X GPUs is reshaping what's possible for AI and HPC in the cloud. This IDC report explores the strategic impact of Vultr’s partnership with AMD, the advantages of cutting-edge GPU architecture, and what it means for developers, enterprises, and innovators alike.
No information is required for download
Additional resources
FAQ
AMD GPU accelerators are AMD’s trademark data center GPUs, architectured for artificial intelligence (AI), high-performance computing (HPC), and machine learning (ML) workloads. Integration with AMD’s CDNA™ architecture and ROCm™ software enables high-density, high-throughput deployments with low latency.
The AMD MI355X is built on the AMD CDNA™ 4 architecture, which supports advanced datatypes like FP6 and FP4, whereas the MI300X and MI325X are built on CDNA™ 3 architecture. Additionally, 288GB of HBM3E memory and 8TB/s memory bandwidth is up from 256GB and 6TB/s, respectively, in the MI325X.
The AMD MI355X is optimized for AI inference, training, and scientific simulation, benefitting any enterprise, innovator, or developer looking to scale and deploy agentic AI applications.
AMD GPUs offer the following for AI training and inference:
- Strong memory bandwidth for AI training
- Accelerated matrix computations for deep learning models
- Price-to-performance advantages in scaling cloud AI infrastructure
- Efficient performance per watt for HPC workloads
AMD MI355X GPUs can be deployed as:
- Vultr Bare Metal servers: For full control of your environment, deploy AMD MI355X as single-tenant dedicated servers, affording direct access to the full capacity of a GPU server with no virtualization overhead.
- Vultr Cloud GPU virtual machines: Offered as 8-GPU plans with the lightest of virtualization, providing near bare metal performance with simple integration of virtual machines. These plans also offer seamless integration with Vultr Cloud Storage.
AMD’s GPUs and ROCm™ ecosystem on Vultr enable rapid scaling and deployment of AI workloads, as well as seamless integration. Read how Vultr and AMD’s collaboration accelerates AI efficiency.
AMD solutions available on Vultr include the AMD MI300X, MI325X, and MI355X GPUs, as well as AMD EPYC™ 9004 Series, AMD EPYC 7003 Series, and AMD EPYC 4005 Series processors. As a Vultr Cloud Alliance member, AMD helps power a composable, open ecosystem of best-in-class partners delivering integrated cloud computing solutions and services.
What are AMD GPU accelerators?
AMD GPU accelerators are AMD’s trademark data center GPUs, architectured for artificial intelligence (AI), high-performance computing (HPC), and machine learning (ML) workloads. Integration with AMD’s CDNA™ architecture and ROCm™ software enables high-density, high-throughput deployments with low latency.
Reserve AMD
MI355X GPUs now
Get ready to build, test, and deploy on The Everywhere Cloud.