Compute Comparison
AMDCDNA 32023

MI300X 192GB

See rental prices ↓

Current median: $3.56/hr across 5 providers (observed), 0.00 (0.0%) vs yesterday. Typical range: $2.15–$20.32/hr (P10–P90). 30-day trend unlocks after 30 days of captures

TDP 750W · 1307 TFLOPS · 192GB VRAM · 5300 GB/s bandwidth

The MI300X 192GB has a 750W TDP, 1307 TFLOPS of FP16 compute, 192GB of HBM3 VRAM, 5300 GB/s memory bandwidth.

AMD's flagship AI GPU. 192GB HBM3 — most VRAM of any single GPU at launch. Strong ROCm software stack. Widely available across major cloud providers.

VRAM
192GB
HBM3
FP16
1.3k
TFLOPS
Bandwidth
5.3k
GB/s
TDP
750W
power
Best for:Memory-bound LLM inferenceLarge context windowsOpen-source model serving

MI300X 192GB currently rents for a median of $3.56/hr across 5 providers (observed on-demand prices). MI300X 192GB has 192GB of memory. Compare its specs and listed cloud rental prices below. The workload must fit in VRAM with room for runtime memory; a listed price does not confirm stock.

MI300X 192GB Overview

AMD’s MI300X is a CDNA 3 accelerator built around 192GB of HBM3, giving it one of the largest single-accelerator memory pools available. It delivers 1,307 TFLOPS of FP16/BF16 performance and targets large language-model training and inference through the ROCm ecosystem.

Its 5,300 GB/s memory bandwidth is a defining advantage for memory-bound LLM serving, and 192GB can hold very large quantized models or extensive KV cache without immediately requiring tensor parallelism. The platform uses Infinity Fabric and PCIe rather than NVIDIA NVLink, so distributed scaling and software tuning follow a different stack.

MI300X is especially compelling for large-context, open-source model serving and memory-heavy training where its capacity can reduce GPU count. The trade-off is ROCm compatibility: teams must verify kernels, inference engines, and operational tooling rather than assuming CUDA-oriented software will transfer unchanged.

Memory

VRAM192 GB
Memory TypeHBM3
Bandwidth5300 GB/s

Compute Performance

FP6481.7 TFLOPS
FP32163.4 TFLOPS
FP161307 TFLOPS
BF161307 TFLOPS
FP82614 TFLOPS
INT82614 TOPS

Hardware Specifications

Chip

ArchitectureAqua Vanjaram
GenerationCDNA 3
Process NodeTSMC 5nm / 6nm
Transistors153B
Release DateDecember 6, 2023

Clocks

Boost Clock1,100 MHz

Memory

VRAM192 GB
Memory TypeHBM3
Memory Bus8192-bit
Bandwidth5300 GB/s

Power

TDP750 W
InterconnectInfinity Fabric / PCIe 5.0

Relative Performance

FP16 Compute17%
VRAM Capacity67%
Mem Bandwidth33%

Relative to highest-spec GPU in database

Limitations

ROCm software ecosystem lags CUDA in library coverage
Fewer cloud providers offer MI300X vs H100/A100
Higher power draw (750W) than comparable NVIDIA options

Live Cloud PricingOn-demand hourly rates

Live7 providers · -89% price spread
ProviderOn-DemandSpotRegionAvailability
$31.50/hr—EU/USunknown
$3.56/hr—US-Westunknown
$1.99/hr—USunknown
$3.56/hr—Variousunknown
$2.39/hr—USunknown
$1.85/hr—US/EU/APACunknown
$3.45/hr—us-east1unknown
Cheapest / month
$22995
$31.50/hr × 730h
Median / month
$2599
at median price
Spot savings
N/A
vs on-demand (best)

Job Cost Estimator

At median $3.56/hr across 5 providers (observed on-demand). Spot pricing typically 40–70% lower.

7B LLM fine-tune
8× GPU, 24 h · Full fine-tune, single node
$684
spot ~$308
70B LLM fine-tune
64× GPU, 72 h · Multi-node, 3-day run
$16k
spot ~$7.4k
Inference server
1× GPU, 730 h · 1 month continuous
$2.6k
spot ~$1.2k
Image gen batch
4× GPU, 8 h · Batch generation job
$114
spot ~$51

Estimates only. Actual cost depends on provider, region, availability, and utilization. Use the GPU cost calculator for custom inputs.

Compare MI300X 192GB vs…

Use Case Guidance

Memory-bound LLM inference
Large context windows
Open-source model serving

LLM Model Size Guidance

Max model (FP16)~96Bparameters at FP16 precision
Max model (INT8)~192Bparameters at INT8 precision
Max model (INT4)~384Bparameters at INT4/GGUF

Estimates only. Actual capacity depends on context length, KV cache, and framework overhead.

Related Guides

LLM APIs Running on This GPU Class

Providers that serve frontier LLM inference on CDNA 3-class hardware.

Browse all 42 LLM models

MI300X 192GB vs Alternatives — Spec Comparison

SpecMI300X 192GB thisMI300X 192GBMI355XMI355X 288GB
VRAM192GB HBM3192GB HBM3288GB HBM3e▲288GB HBM3e▲
Memory Bandwidth5300 GB/s5300 GB/s8000 GB/s▲8000 GB/s▲
FP16 TFLOPS1307▲1307▲13001300
BF16 TFLOPS1307▲1307▲13001300
FP8 TFLOPS2614▲2614▲26002600
INT8 TOPS2614▲2614▲26002600
TDP750W▲750W▲750W▲750W▲
Process NodeTSMC 5nm / 6nm▲TSMC 5nm / 6nm▲TSMC 3nmTSMC 3nm
ArchitectureAqua VanjaramAqua VanjaramAntaresAntares
Release Year202320232025▲2025▲
Max model (FP16)~96B params~96B params~144B params▲~144B params▲
Max model (INT4)~384B params~384B params~576B params▲~576B params▲
▲ indicates best value in row · FP16/BF16 TFLOPS at full precision · Max model estimates at 2 bytes/param (FP16) and 0.5 bytes/param (INT4)Full side-by-side comparison

Related GPUs

MI300X 192GB Frequently Asked Questions

What is the MI300X 192GB TDP?

The MI300X 192GB has a TDP (Thermal Design Power) of 750W.

How many TFLOPS does the MI300X 192GB have?

The MI300X 192GB delivers 1307 TFLOPS of FP16 compute and 1307 TFLOPS BF16, and 2614 TFLOPS FP8.

How much VRAM does the MI300X 192GB have?

The MI300X 192GB has 192GB of HBM3 VRAM.

What is the MI300X 192GB's memory bandwidth?

The MI300X 192GB has a memory bandwidth of 5300 GB/s.

MI300X 192GB User Reviews

MI300X 192GB User Reviews

Loading reviews...

Ready to rent?

Compare MI300X 192GB prices across 130+ providers

Live on-demand & spot rates · monthly cost estimates · availability status

Compare rental prices