GPU comparison

MI300X vs MI325X

Specifications and current cloud pricing, side by side.

CDNA 3vsCDNA 3Updated 33 days ago

The MI325X provides the stronger choice for the most common use case of LLM inference because its 256 GB memory capacity exceeds the 192 GB on the MI300X while preserving the same 1307.4 TFLOPS dense FP16 rating.

MI300X listed from $0.95/GPU/hrMI325X listed from $4.62/GPU/hr

Specifications Compared

SpecMI300XMI325X
TDP750W1000W
VRAM192 GB256 GB
FP8 (dense)2,614.9 TFLOPS2,614.9 TFLOPS
Memory TypeHBM3HBM3e
ArchitectureCDNA 3CDNA 3
FP16 (dense)1,307.4 TFLOPS1,307.4 TFLOPS
Form FactorsOAMOAM
INT8 (dense)2,614.9 TOPS2,614.9 TOPS
InterconnectInfinity Fabric, PCIe 5.0Infinity Fabric, PCIe 5.0
FP32 Performance163.4 TFLOPS163.4 TFLOPS
FP64 Performance81.7 TFLOPS81.7 TFLOPS
Memory Bandwidth5,300 GB/s6,000 GB/s
FP8 (with sparsity)5,229.8 TFLOPS5,229.8 TFLOPS
FP16 (with sparsity)2,614.9 TFLOPS2,614.9 TFLOPS
INT8 (with sparsity)5,229.8 TOPS5,229.8 TOPS

Performance Analysis

Dense FP16 performance equals 1307.4 TFLOPS on the MI300X and the MI325X alike so training throughput per GPU stays constant when sparsity remains unused. The FP16 rating with sparsity reaches 2614.9 TFLOPS on each accelerator and the FP8 dense rating also equals 2614.9 TFLOPS. FP32 performance remains fixed at 163.4 TFLOPS across both devices.

Current On-Demand Offers

Cheapest secure on-demand offer per provider that is reported in stock right now, per GPU per hour; multi-GPU instances show the whole-instance price alongside. Deploy opens the DeployGPU console for providers on the platform, otherwise the provider's own site. Stock is rechecked every minute.

MI300X

MI300X is not offered on-demand by any provider we track right now. See the MI300X rental page for last-seen listed prices and a price alert.

MI325X

MI325X is not offered on-demand by any provider we track right now. See the MI325X rental page for last-seen listed prices and a price alert.

Which GPU to watchWatch the price of

Notify me when MI300X drops below a price

One email when the cheapest in-stock on-demand price per GPU-hour falls below your threshold.

QuantaCloud

Comparing providers? We broker across all of them.

Stop tab-switching between pricing pages. Tell us what you need, 16+ GPUs reserved or cluster capacity, and we return one quote at partner rates within 24 hours.

No waitlist24hr quote turnaroundInfiniBand fabric

Compare real-time pricing across 25+ providers

When to Choose the MI300X

The MI300X fits deployments constrained to a 750W TDP ceiling where the 192 GB HBM3 capacity meets model size needs. Its 5300 GB/s bandwidth supports workloads that do not require the additional 700 GB/s offered by the alternative.

When to Choose the MI325X

The MI325X fits deployments that require 256 GB HBM3e capacity to hold larger models or to increase batch sizes at 6000 GB/s bandwidth. Its higher TDP of 1000W accommodates sustained operation under those expanded memory conditions.

Use Cases

LLM Training
MI325X

The MI325X supplies 256 GB HBM3e memory which exceeds the 192 GB on the MI300X and supports larger training batches at 6000 GB/s bandwidth.

LLM Inference
MI325X

The MI325X delivers 256 GB capacity against 192 GB on the MI300X allowing longer context lengths while maintaining 1307.4 TFLOPS dense FP16.

Fine-tuning
Either

Both GPUs share identical 1307.4 TFLOPS dense FP16 and 163.4 TFLOPS FP32 ratings so selection depends on whether 192 GB or 256 GB memory meets the workload.

Stable Diffusion
MI300X

The MI300X operates at a 750W TDP which reduces power draw compared with the 1000W TDP on the MI325X for diffusion workloads that fit in 192 GB.

Scientific Computing
Either

FP32 performance stays fixed at 163.4 TFLOPS on both the MI300X and MI325X so either device meets requirements when memory fits within 192 GB or 256 GB.

Frequently Asked Questions

How much more memory does the MI325X provide than the MI300X?▾

The MI325X provides 256 GB HBM3e memory while the MI300X provides 192 GB HBM3 memory resulting in a 64 GB increase.

Do the MI300X and MI325X share the same FP16 performance?▾

Both GPUs deliver 1307.4 TFLOPS dense FP16 and 2614.9 TFLOPS FP16 with sparsity so peak rates match exactly.

What is the TDP difference between the two GPUs?▾

The MI300X lists a 750W TDP while the MI325X lists a 1000W TDP creating a 250W increase on the later model.

How does memory bandwidth compare on the MI300X versus the MI325X?▾

Memory bandwidth equals 5300 GB/s on the MI300X and rises to 6000 GB/s on the MI325X for a 700 GB/s gain.

Are the interconnect options identical on both accelerators?▾

Both the MI300X and MI325X use Infinity Fabric together with PCIe 5.0 in the OAM form factor.

Which GPU offers higher INT8 throughput with sparsity?▾

Both GPUs reach 5229.8 TOPS INT8 with sparsity so the rating remains the same across the MI300X and MI325X.

Which is cheaper to rent, the MI300X or the MI325X?▾

Cloud rental prices for both the MI300X and MI325X vary by provider, configuration, and availability. This page shows live pricing from 25+ providers updated every 60 seconds. Scroll to the Live Cloud Pricing section to compare current rates.

How much VRAM does the MI300X have compared to the MI325X?▾

The MI300X has 192 GB of HBM3 memory. The MI325X has 256 GB of HBM3e memory.

Can I find MI300X and MI325X GPUs available to rent right now?▾

Yes. This page shows real-time availability across 25+ cloud GPU providers. The Live Cloud Pricing section displays only in-stock offers with current pricing.

What is the main difference between the MI300X and the MI325X?▾

The MI300X uses the CDNA 3 architecture (2023) while the MI325X uses CDNA 3 (2024). Both deliver the same dense FP16 throughput (1,307.4 TFLOPS without sparsity), and the MI325X has 1.1x the memory bandwidth of the MI300X.

Rent these GPUs

Each GPU page lists every current on-demand offer by provider, per GPU-hour, updated every minute.

Related comparisons

How this page is made

  • Specifications come from the NVIDIA, AMD and Intel datasheets for the MI300X and the MI325X. Dense and sparse throughput are listed separately.
  • Prices and availability are live: every offer shown is an in-stock, on-demand listing from a provider we track, rechecked every minute.
  • The written comparison (overview, performance notes, when to choose each card, verdict, use cases and FAQ) was drafted with an AI model from the specification table above and passed an automated check that rejects any figure not in that table. It was last generated on . It contains no prices; those are always read live.
  • Read how we collect the data, or report an error on this page.

Next steps