Skip to content
Updated minutes ago
nvidia

RTX 4090

nvidia · ada · 24 GB GDDR6X · 450W TDP

VRAM

24 GB

BF16 TFLOPS

165

Bandwidth

1008 GB/s

From

$0.39/hr

Calculate ROI with this GPU →

Spec Sheet

VRAM24 GB GDDR6X
Memory Bandwidth1008 GB/s
BF16 TFLOPS165
FP16 TFLOPS165
FP8 TFLOPS330
INT8 TOPS330
TDP450W
InterconnectPCIE
Max per Node4
PCIe Gen4
CUDA Compute Capability8.9
Tensor CoresYes

Pricing by Provider

ProviderOn-DemandReservedSpotBadge
fluidstack$0.59/hr-$0.39/hrCheapest
vast_ai$0.74/hr-$0.44/hr
tensordock$0.69/hr-$0.44/hr
runpod$1.10/hr-$0.79/hr
lambda$0.89/hr--

Pricing History

runpod
$1.10/hr 0.0%
2024-01-012025-03-01
Low: $1.10High: $1.44
vast
$0.54/hr 0.0%
2024-01-012025-03-01
Low: $0.54High: $0.89

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
0.31 t/s/WFP8
Qwen 2.5 7B
0.29 t/s/WFP8
Llama 3.1 8B
0.28 t/s/WFP8
Llama 3.1 70B
0.03 t/s/WFP8
Qwen 2.5 72B
0.03 t/s/WFP8

Similar GPUs

GPUVRAMBF16 TFLOPSBW (GB/s)From
L424 GB121300$0.29/hr
RTX 408016 GB97717$0.32/hr
RTX 4060 Ti 16GB16 GB44288$0.30/hr
RTX 4070 Ti12 GB93504$0.25/hr
RTX 4070 Super12 GB55504$0.22/hr

Methodology Note

Performance estimates for the RTX 4090are 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 4090's rated 1008 GB/s memory bandwidth and 165 BF16 TFLOPS compute capacity as roofline ceilings, with empirical correction factors per GPU architecture (ada). See our full methodology.

Frequently Asked Questions

How many AI models can run on RTX 4090?

The RTX 4090 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 4090.

What is the RTX 4090 inference throughput?

The RTX 4090 delivers 165 BF16 TFLOPS and 330 FP8 TFLOPS with 1008 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 4090 cost per hour?

The RTX 4090 is available starting from $0.39/hour via fluidstack. Prices vary by provider and pricing tier (on-demand, reserved, spot). Compare pricing across all providers in the table above.