Skip to content
Updated minutes ago
nvidia

RTX 3090

nvidia · ampere · 24 GB GDDR6X · 350W TDP

VRAM

24 GB

BF16 TFLOPS

35.6

Bandwidth

936 GB/s

From

$0.19/hr

Calculate ROI with this GPU →

Spec Sheet

VRAM24 GB GDDR6X
Memory Bandwidth936 GB/s
BF16 TFLOPS35.6
FP16 TFLOPS35.6
FP8 TFLOPS35.6
INT8 TOPS71.2
TDP350W
InterconnectPCIE
Max per Node4
PCIe Gen4
CUDA Compute Capability8.6
Tensor CoresYes

Pricing by Provider

ProviderOn-DemandReservedSpotBadge
tensordock$0.29/hr-$0.19/hrCheapest
vast_ai$0.36/hr-$0.20/hr
runpod$0.54/hr-$0.34/hr

Pricing History

vast
$0.29/hr 0.0%
2024-01-012025-03-01
Low: $0.29High: $0.49

Compatible Models (240)

Training Capabilities

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

Model SizeFull Fine-TuneQLoRA
7B model6 GPUs1 GPU
13B model11 GPUs1 GPU
70B model55 GPUs2 GPUs

Energy Efficiency

Estimated tokens/second per Watt for popular models

Mistral 7B
0.37 t/s/WFP8
Qwen 2.5 7B
0.35 t/s/WFP8
Llama 3.1 8B
0.33 t/s/WFP8
Llama 3.1 70B
0.04 t/s/WFP8
Qwen 2.5 72B
0.04 t/s/WFP8

Similar GPUs

GPUVRAMBF16 TFLOPSBW (GB/s)From
A10G24 GB35600$0.30/hr
A3024 GB165933$0.35/hr
RTX A500024 GB27.8768$0.25/hr
A400016 GB76448$0.17/hr
A216 GB18200$0.15/hr

Methodology Note

Performance estimates for the RTX 3090are based on InferenceBench's roofline performance model with CUDA kernel-level optimization including FlashAttention v2 and PagedAttention. Memory calculations account for model weights (24 GB GDDR6X available), KV-cache allocation, and activation memory. Throughput predictions use the RTX 3090's rated 936 GB/s memory bandwidth and 35.6 BF16 TFLOPS compute capacity as roofline ceilings, with empirical correction factors per GPU architecture (ampere). See our full methodology.

Frequently Asked Questions

How many AI models can run on RTX 3090?

The RTX 3090 can run 240 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 4x RTX 3090.

What is the RTX 3090 inference throughput?

The RTX 3090 delivers 35.6 BF16 TFLOPS and 35.6 FP8 TFLOPS with 936 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 RTX 3090 cost per hour?

The RTX 3090 is available starting from $0.19/hour via tensordock. Prices vary by provider and pricing tier (on-demand, reserved, spot). Compare pricing across all providers in the table above.