Skip to content
Updated minutes ago
amd

Instinct MI100

amd · cdna · 32 GB HBM2 · 300W TDP

VRAM

32 GB

BF16 TFLOPS

184.6

Bandwidth

1229 GB/s

From

$0.40/hr

Calculate ROI with this GPU →

Spec Sheet

VRAM32 GB HBM2
Memory Bandwidth1229 GB/s
BF16 TFLOPS184.6
FP16 TFLOPS184.6
FP8 TFLOPS184.6
INT8 TOPS184.6
TDP300W
InterconnectINFINITY-FABRIC
Max per Node8
PCIe Gen4
Tensor CoresNo

Pricing by Provider

ProviderOn-DemandReservedSpotBadge
vast_ai$0.60/hr-$0.40/hrCheapest

Compatible Models (260)

Training Capabilities

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

Model SizeFull Fine-TuneQLoRA
7B model5 GPUs1 GPU
13B model8 GPUs1 GPU
70B model42 GPUs2 GPUs

Energy Efficiency

Estimated tokens/second per Watt for popular models

Mistral 7B
0.56 t/s/WFP8
Qwen 2.5 7B
0.54 t/s/WFP8
Llama 3.1 8B
0.51 t/s/WFP8
Llama 3.1 70B
0.06 t/s/WFP8
Qwen 2.5 72B
0.06 t/s/WFP8

Similar GPUs

GPUVRAMBF16 TFLOPSBW (GB/s)From
RTX 509032 GB2101792$0.89/hr
V100 32GB32 GB28.3900$0.19/hr
TPU v432 GB2751200$2.25/hr
TPU v6e (Trillium)32 GB4601640$1.75/hr
Cloud AI 10032 GB150134$0.00/hr

Methodology Note

Performance estimates for the Instinct MI100are based on InferenceBench's roofline performance model with CUDA kernel-level optimization including FlashAttention v2 and PagedAttention. Memory calculations account for model weights (32 GB HBM2 available), KV-cache allocation, and activation memory. Throughput predictions use the Instinct MI100's rated 1229 GB/s memory bandwidth and 184.6 BF16 TFLOPS compute capacity as roofline ceilings, with empirical correction factors per GPU architecture (cdna). See our full methodology.

Frequently Asked Questions

How many AI models can run on Instinct MI100?

The Instinct MI100 can run 260 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 MI100.

What is the Instinct MI100 inference throughput?

The Instinct MI100 delivers 184.6 BF16 TFLOPS and 184.6 FP8 TFLOPS with 1229 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 MI100 cost per hour?

The Instinct MI100 is available starting from $0.40/hour via vast_ai. Prices vary by provider and pricing tier (on-demand, reserved, spot). Compare pricing across all providers in the table above.