Skip to content
Updated minutes ago
amd

Instinct MI300X

amd · cdna3 · 192 GB HBM3 · 750W TDP

VRAM

192 GB

BF16 TFLOPS

1307

Bandwidth

5300 GB/s

From

$1.79/hr

Calculate ROI with this GPU →

Spec Sheet

VRAM192 GB HBM3
Memory Bandwidth5300 GB/s
BF16 TFLOPS1307
FP16 TFLOPS1307
FP8 TFLOPS2614
INT8 TOPS2614
TDP750W
InterconnectINFINITY-FABRIC
Max per Node8
PCIe Gen5
Tensor CoresNo

Pricing by Provider

ProviderOn-DemandReservedSpotBadge
fluidstack$2.39/hr-$1.79/hrCheapest
vast_ai$2.79/hr-$1.99/hr
tensordock$2.69/hr-$1.99/hr
lambda$2.49/hr--
coreweave$3.39/hr$2.49/hr-
runpod$3.49/hr-$2.59/hr

Compatible Models (280)

Training Capabilities

Estimated GPU count for full fine-tuning (AdamW, BF16) and QLoRA

Model SizeFull Fine-TuneQLoRA
7B model1 GPU1 GPU
13B model2 GPUs1 GPU
70B model7 GPUs1 GPU

Energy Efficiency

Estimated tokens/second per Watt for popular models

Mistral 7B
0.97 t/s/WFP8
Qwen 2.5 7B
0.93 t/s/WFP8
Llama 3.1 8B
0.88 t/s/WFP8
DeepSeek V3
0.19 t/s/WFP8
Llama 3.1 70B
0.10 t/s/WFP8
Qwen 2.5 72B
0.10 t/s/WFP8

Similar GPUs

GPUVRAMBF16 TFLOPSBW (GB/s)From
Instinct MI325X256 GB13076000$2.49/hr
B100 SXM192 GB17508000$4.50/hr
GB200 NVL72 (per GPU)192 GB22508000$6.50/hr
GB300 NVL72 (per GPU)192 GB25008000$7.50/hr
H100 NVL 94GB (per GPU pair)188 GB16707876$5.49/hr

Methodology Note

Performance estimates for the Instinct MI300Xare based on InferenceBench's roofline performance model with CUDA kernel-level optimization including FlashAttention v2 and PagedAttention. Memory calculations account for model weights (192 GB HBM3 available), KV-cache allocation, and activation memory. Throughput predictions use the Instinct MI300X's rated 5300 GB/s memory bandwidth and 1307 BF16 TFLOPS compute capacity as roofline ceilings, with empirical correction factors per GPU architecture (cdna3). See our full methodology.

Frequently Asked Questions

How many AI models can run on Instinct MI300X?

The Instinct MI300X can run 280 AI models from our database within a single node. Compatible models range across various parameter sizes depending on the quantization precision (BF16, FP8, INT4). Smaller models fit on a single GPU while larger models may require multi-GPU setups up to 8x Instinct MI300X.

What is the Instinct MI300X inference throughput?

The Instinct MI300X delivers 1307 BF16 TFLOPS and 2614 FP8 TFLOPS with 5300 GB/s memory bandwidth. Actual inference throughput (tokens/sec) depends on the model size, precision, and batch size. Use our calculator for model-specific throughput estimates.

How much does Instinct MI300X cost per hour?

The Instinct MI300X is available starting from $1.79/hour via fluidstack. Prices vary by provider and pricing tier (on-demand, reserved, spot). Compare pricing across all providers in the table above.