AMD

Instinct MI355X

Instinct MI355X — illustration of the card's form factor
VRAM
288 GB
FP32 TFLOPS
157.3 TFLOPs
CUDA Cores
16,384
TDP
1400 W

Provider Marketplace

Cheapest
$4.50/hour
Starting from
DigitalOcean logo
DigitalOceanVisit
Best Value
—
Awaiting listings
No additional provider yet
Enterprise Choice
—
Awaiting listings
No additional provider yet

All Cloud Providers

1 Options available
DigitalOcean logo
DigitalOceanCheapest
Spot · preemptible
$4.50/ hour
Estimated Cost
Provision

Estimates only — rates are collected automatically from public provider pages and may be out of date. Prices vary by region, commitment term, and availability, and typically exclude storage, egress, and tax. Confirm current pricing with the provider before purchasing. Last collected .

Compute Performance

FP6478.6 TFLOPs
FP32157.3 TFLOPs
FP16157.3 TFLOPs
BF162.5 PFLOPs TFLOPS
FP85 PFLOPs TFLOPS
INT85 POPs TOPS
INT410.1 PFLOPs TOPS

Architecture

MicroarchitectureCDNA4
Process NodeTSMC 3nm | 6nm FinFET
Transistors185 Billion
Compute Units256 CUs
CUDA Cores16384
Tensor CoresMatrix Cores 1024
Matrix EngineMXFP4/MXFP6/MXFP8/OCP-FP8
Boost Clock2400 MHz
Sparse AccelerationSupported (structured sparsity)
Dynamic PrecisionSupported (MXFP4/MXFP6/MXFP8/OCP-FP8/FP16/FP32/BF16/INT8/FP64)

Memory & VRAM

Memory TypeHBM3E
Total Capacity288 GB
Bandwidth8 TB/s
Bus Width8192 bits
ECC SupportYes (Full-Chip)
Unified MemoryYes (Coherent shared memory between all eight accelerators)
Memory PoolingCoherent shared memory between accelerators

Connectivity & Scaling

InterconnectAMD Infinity Fabric
IB Bandwidth153 GB/s
PCIe InterfacePCIe 5.0 xx16
TopologyCoherent shared memory between all eight accelerators on a universal baseboard
Max GPUs/Node8
Scale-OutEthernet
P2P MemoryHybrid hardware/software memory coherency between all eight accelerators

Virtualization

SR-IOVSupported
K8s ReadinessSupported (AMD GPU Operator)
GPU SharingSR-IOV (up to 8 partitions)
Virt EfficiencyNear bare-metal (vendor claim)

Power & Efficiency

TDP1400 W
Thermal LimitsCooling Passive & Active

Physical Design

Form FactorOAM Module
CoolingPassive & Active
Rack DensityHigh-density optimized

Thermals & Cooling

ThrottlingSustain higher performance over time; minimizes throttling and maximizes throughput during prolonged or intensive workloads.

Software Ecosystem

PyTorchSupported
TensorFlowSupported
JAXSupported
HuggingFaceSupported (collaboration mentioned)
Triton ServerSupported

Server & Deployment

OEM AvailabilityCollaborations between AMD and leading cloud service providers (CSPs), original equipment manufacturers (OEMs), and platform designers drive a robust ecosystem of AMD Instinct MI350 Series powered servers, delivering a comprehensive and diverse portfolio of AI and HPC solutions to the market.
PreconfiguredWhen deployed in an 8-GPU AMD Instinct MI355X platform using the AMD Universal Base Board (UBB 2.0) that can be integrated into server designs as compact as 2U.
Rack-ScalePurpose built for high density computing environments.

System Compatibility

Required PCIePCIe® 5.0 x16
MotherboardOAM Module

Benchmarks & Throughput

Structured Sparsity

Supported

Scaling Efficiency

Hybrid hardware/software memory coherency between all eight accelerators with 160 GB/s bidirectional bandwidth between each GPU.

Multi-GPU Scalability

Scaling Efficiency

8-GPU160 GB/s bidirectional bandwidth between each GPU

Scaling Characteristics

ParallelismSR-IOV for up to 8 partitions

Workload Readiness

LLM Training

supported

LLM Inference

supported

Vision Training

supported

HPC / Simulation

supported

Scientific Computing

supported

Real-Time Serving

supported

Market Authority

Cloud Adoption

Broad adoption by cloud service providers and OEMs; Microsoft and Meta cited as users.

Enterprise Cases

AMD cites Microsoft and Meta as users of Instinct GPUs to power large-scale AI models.

Key Strengths

Limitations

Also in the Lineup

AMD
Instinct MI100
AMD
Instinct MI200
AMD
Instinct MI210 PCIe Gen4 Passive Accelerator
AMD
Instinct MI250
AMD
Instinct MI250X
AMD
Instinct MI300A APU
AMD
Instinct MI300X
AMD
Instinct MI325X

Expert Insight

The Instinct MI355X represents a powerful alternative for diversified workloads. When comparing cloud providers, consider not just the hourly rate, but also the interconnect bandwidth (InfiniBand/NVLink) and regional availability which can significantly impact total cost of ownership for large-scale training.

Glossary Terms

FP32 TFLOPS
VRAM
TDP
Cores
Information updated daily. Cloud pricing subject to vendor availability.