NVIDIA H100 SXM vs NVIDIA L4 Full Specs

Summary

16,896 Shaders 1.98GHz	7,680 Shaders 2.04GHz
96GB HBM33.35TB/s	24GB GDDR6300.1GB/s
Peak·FP32·FP16 66.91 TFLOPS	Peak·FP32·FP16 31.33 TFLOPS
Form Factor SXM5	Form Factor PCIe Card
TDP 700W	TDP 72W
Power Connectors -	Power Connectors 1x 16-Pin 12VHPWR

Highlights

Peak OPS

Peak OPS FP32 FP16 Tensor FP8-16 FP8-16 Tensor Sparse Tensor FP8-32 FP8-32 Tensor Sparse Tensor FP16-16 FP16-16 Tensor Sparse Tensor FP16-32 FP16-32 Tensor Sparse FP64 Tensor FP64 BF16 Tensor BF16 BF16 Tensor Sparse Tensor TF32 Tensor INT8 Tensor INT4 Memory Bandwidth TDP

H100 SXM3.96 POPSINT8 Tensor Sparse

x3.95

L41 POPSINT4 Tensor Sparse

x1

Show Top Score

Theoretical Performance

Clock Speed1980MHz UC·Base·Boost·OC	Clock Speed2040MHz UC·Base·Boost·OC
Peak OPS 3.96 POPSINT8 Tensor Sparse	Peak OPS 1 POPSINT4 Tensor Sparse
Tensor FP8-16 1.98 PFLOPS FP8-16 Tensor Sparse 3.96 PFLOPS Tensor FP8-32 1.98 PFLOPS FP8-32 Tensor Sparse 3.96 PFLOPS	Tensor FP8-16 250.7 TFLOPS FP8-16 Tensor Sparse 501.4 TFLOPS Tensor FP8-32 125.3 TFLOPS FP8-32 Tensor Sparse 250.7 TFLOPS
Tensor FP16-16 989.4 TFLOPS FP16-16 Tensor Sparse 1.98 PFLOPS Tensor FP16-32 989.4 TFLOPS FP16-32 Tensor Sparse 1.98 PFLOPS	Tensor FP16-16 125.3 TFLOPS FP16-16 Tensor Sparse 250.7 TFLOPS Tensor FP16-32 62.67 TFLOPS FP16-32 Tensor Sparse 125.3 TFLOPS
BF16 133.8 TFLOPS Tensor BF16 989.4 TFLOPS BF16 Tensor Sparse 1.98 PFLOPS	BF16 31.33 TFLOPS Tensor BF16 62.67 TFLOPS BF16 Tensor Sparse 125.3 TFLOPS
Tensor TF32 494.7 TFLOPS	Tensor TF32 31.33 TFLOPS
FP32 66.91 TFLOPS	FP32 31.33 TFLOPS
FP64 33.45 TFLOPS Tensor FP64 61.84 TFLOPS	FP64 489.6 GFLOPS Tensor FP64 -
Tensor INT4 -	Tensor INT4 501.4 TOPS
Tensor INT8 1.98 POPS	Tensor INT8 250.7 TOPS
Ray -	Ray 72.43 TOPS
Pixel Rate 47.5 GPixel/s	Pixel Rate 163.2 GPixel/s
Texture Rate 1.04 TTexel/s	Texture Rate 489.6 GTexel/s

Core Config

Shaders 16,896 Shaders	Shaders 7,680 Shaders
TMUs 528 TMUs	TMUs 240 TMUs
ROPs 24 ROPs	ROPs 80 ROPs
Tensor Cores 528 T-Cores	Tensor Cores 240 T-Cores
RT Cores -	RT Cores 60 RT-Cores
SMs 132 SMs	SMs 60 SMs

Clocks

Base Clock 1.06GHz	Base Clock 795MHz
Boost Clock 1.98GHz	Boost Clock 2.04GHz
Tensor Clock 1.83GHz	Tensor Clock -

Cache

L2 Cache 51.2MB shared

L2 Cache 49.1MB shared

Memory

96GB HBM3	24GB GDDR6
Memory Bus 5120-bit	Memory Bus 192-bit
Memory Speed 5.2GT/s	Memory Speed 12.5GT/s
Memory Bandwidth 3.35TB/s	Memory Bandwidth 300.1GB/s

Power & Thermals

TDP 700W

TDP 72W

Encoder

-	Encoder Model 2x NVENC 8

Decoder

Decoder Model 7x NVDEC 5

Decoder Model NVDEC 5

Card

Form Factor SXM5	Form Factor PCIe Card
PCIe -	PCIe 1-Slots
-	Height 56 mm (2.21") Width 169 mm (6.65") Depth 20 mm (0.79")
Cooling Passive	Cooling Passive
Power Connectors -	Power Connectors 1x 16-Pin 12VHPWR

Chip

Manufacturer NVIDIA	Manufacturer NVIDIA
Chip Designer NVIDIA	Chip Designer NVIDIA
Architecture Hopper	Architecture Ada Lovelace
Family Server	Family Server
Branding	Branding
Codename NV180	Codename NV184
Chip Variant -	Chip Variant AD104-250
Market Segment Server	Market Segment Server
Release Date Mar 22, 2022	Release Date Mar 21, 2023

Fabrication

Foundry TSMC	Foundry TSMC
Fabrication Node 4N	Fabrication Node 4N
Die Size 814mm²	Die Size 295mm²
Transistor Count 80 Billion	Transistor Count 35.8 Billion
Transistor Density 98.28 MTr/mm²	Transistor Density 121.6 MTr/mm²

Images

No images available

No images available

Competitors

NVIDIAH100 SXM vs GPU NVIDIAB100 NVIDIAH100 SXM vs GPU NVIDIAB200 NVIDIAH100 SXM vs GPU NVIDIAGB200 Show 12 more

NVIDIAL4 vs GPU NVIDIARTX 4000 Ada NVIDIAL4 vs GPU NVIDIARTX 4500 Ada NVIDIAL4 vs GPU NVIDIARTX PRO 4000 Blackwell SFF Show 12 more