<img alt="" src="https://secure.insightful-enterprise-intelligence.com/783141.png" style="display:none;">

Access NVIDIA H100s from just $2.06/hour. Reserve here

Deploy 8 to 16,384 NVIDIA H100 SXM GPUs on the AI Supercloud. Learn More

NVIDIA L40
nvidia-l40
NVIDIA L40

Introducing the NVIDIA L40 GPU for Powerhouse Performance

Leveraging Ada Lovelace architecture, the NVIDIA L40 GPU integrates RT, Tensor, and CUDA cores for groundbreaking visual computing. Its 48GB GDDR6 memory handles intensive applications, adaptable via vGPU software. Designed for 24x7 enterprise operations, NEBS Level 3 compliant, and available in passively cooled, power-efficient designs, the NVIDIA L40 GPU Benchmark shows high-performance scalability for diverse data centre needs. Get yours at Hyperstack now!

product-banner-curve

Unrivalled performance in…

tick_check

Real Time Ray Tracing

3rd generation RT Cores deliver up to 2x faster real-time ray tracing compared to previous generations

tick_check

AI Acceleration

4th generation Tensor Cores with support for structural sparsity and optimised TF32 format deliver 5x higher inference performance than third-generation tensor cores

tick_check

Ultra-Fast Memory

Equipped with 48GB of ultra-fast GDDR6 memory to tackle memory-intensive applications

tick_check

Networking and Storage

Swift data transfer and efficient handling of high-volume workloads, essential for demanding AI and large-scale computational tasks

NVIDIA L40 GPU

Pricing starts from $0.70/hour

nvidia-l40-reserve-now

Benefits of NVIDIA L40 GPU

get-hyper-realistic-visuals

Get Hyper-Realistic Visuals

Witness your designs come to life with 2x faster ray tracing thanks to third-generation RT Cores. Experience stunning product prototypes, architectural models, and engineering simulations in real time with hardware-accelerated motion blur.

level-up-ai-training

Level up AI Training

Accelerate your AI and data science workflows by 5x with fourth-generation Tensor Cores, optimised for structural sparsity and TF32 format. Train complex models faster, gain deeper insights, and get AI-powered solutions in record time.

largest-memory

Largest Memory

Carry on demanding tasks like data science, simulations, and high-resolution rendering with a massive 48GB of ultra-fast GDDR6 memory. Allocate memory efficiently with vGPU software, allowing creative, data science, and design teams to collaborate seamlessly on large projects.

operations

24/7 Operations

Built for 24/7 operation, the NVIDIA L40 thrives in demanding data centre environments. Its power-efficient design and NEBS Level 3 compliance ensure reliable performance and scalability for diverse workloads.

be-secure

PCI Express Gen 4.0

Protect your valuable data with Secure Boot and Root of Trust technology. The NVIDIA L40 prioritises security, offering peace of mind for mission-critical applications.

scalable-and-adaptable

Scalable and Adaptable

Easily deploy at scale and integrate into a wide range of NVIDIA-certified systems

Technical Specifications

GPU: NVIDIA L40

CUDA Cores: 18,176

Graphics Memory: 48GB GDDR6 with ECC

GPU Architecture
NVIDIA Ada Lovelace architecture
GPU Memory
48GB GDDR6 with ECC
Memory Bandwidth
864GB/s
CUDA Cores
18,176
RT Cores
142
Tensor Cores
568
RT Core performance TFLOPS
209
FP32 TFLOPS
90.5
TF32 Tensor Core TFLOPS
90.5 181
BFLOAT16 Tensor Core TFLOPS
181.05 362.1
FP16 Tensor Core
181.05 362.1
FP8 Tensor Core
362 724
Peak INT8 Tensor TOPS
362
Peak INT4 Tensor TOPS
724
Max Power Consumption
300W
Thermal
Passive
Virtual GPU (vGPU) software support
Yes
NVENCI NVDEC
3x 1 3x (Includes AVI Encode & Decode)
Secure Boot with Root of Trust
Yes
NEBS Ready
Level 3
GPU Architecture
NVIDIA Ada Lovelace architecture
GPU Memory
48GB GDDR6 with ECC
Memory Bandwidth
864GB/s
CUDA Cores
18,176
RT Cores
142
Tensor Cores
568
RT Core performance TFLOPS
209
FP32 TFLOPS
90.5
TF32 Tensor Core TFLOPS
90.5 181
BFLOAT16 Tensor Core TFLOPS
181.05 362.1
FP16 Tensor Core
181.05 362.1
FP8 Tensor Core
362 724
Peak INT8 Tensor TOPS
362
Peak INT4 Tensor TOPS
724
Max Power Consumption
300W
Thermal
Passive
Virtual GPU (vGPU) software support
Yes
NVENCI NVDEC
3x 1 3x (Includes AVI Encode & Decode)
Secure Boot with Root of Trust
Yes
NEBS Ready
Level 3

Available on
Hyperstack

nvidia-l40-available-on-hyperstack

Frequently Asked Questions

Hyperstack's support team for NVIDIA L40 ensures smooth deployment, offering expert guidance and rapid resolution for any hurdle.

How fast is the NVIDIA L40 GPU card?

The NVIDIA L40 GPU card offers a peak FP32 performance of 130 Tensor PetaFLOPS and is one of the fastest GPUs in the market. This means significantly faster processing for AI and graphics workloads.

How many CUDA cores does a NVIDIA L40 have?

The NVIDIA L40 has 8192 CUDA cores, offering immense parallel processing power for demanding tasks like deep learning training and high-fidelity rendering.

What is the difference between NVIDIA A40 and NVIDIA L40?

The NVIDIA L40 offers significantly higher FP32 performance (130 vs. 427 TOPs) and nearly double the CUDA cores compared to the NVIDIA A40/A6000.

What applications or workloads is the NVIDIA L40 GPU optimised for?

The NVIDIA L40 GPU is optimised for neural graphics, virtualisation, computing, and AI capabilities for GPU-accelerated data centre workloads.

Can the NVIDIA L40 GPU card be used for deep learning and AI applications?

Yes, the latest fourth-generation Tensor Cores of NVIDIA L40 GPU deliver groundbreaking performance for deep learning and AI inference applications.

What is the NVIDIA L40 GPU price?

The NVIDIA L40 GPU price is $1.00 per hour on-demand, and starts at $0.70 for reservation.   

A condimentum vitae sapien

Est velit egestas dui id ornare arcu odio ut. Varius quam quisque id diamEst velit egestas dui id ornare arcu odio ut. Varius quam quisque id diam

prd-01

A condimentum vitae sapien

Est velit egestas dui id ornare arcu odio ut. Varius quam quisque id diamEst velit egestas dui id ornare arcu odio ut. Varius quam quisque id diam

prd-01

A condimentum vitae sapien

Est velit egestas dui id ornare arcu odio ut. Varius quam quisque id diamEst velit egestas dui id ornare arcu odio ut. Varius quam quisque id diam

prd-01