Intel Gaudi 3 vs NVIDIA H200 SXM Full Specs

Summary

16,384 Shaders 438MHz	16,896 Shaders 1.98GHz
128GB HBM2e3.7TB/s	141GB HBM3e4.81TB/s
Peak·FP32·FP16 -	Peak·FP32·FP16 66.91 TFLOPS
Form Factor OAM Module	Form Factor SXM5
TDP 900W	TDP 700W

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 Tensor FP32 FP64 Tensor FP64 BF16 Tensor BF16 BF16 Tensor Sparse Tensor TF32 Tensor INT8 Memory Bandwidth TDP

Gaudi 31.84 PFLOPSFP8-32 Tensor

x1

H200 SXM3.96 POPSINT8 Tensor Sparse

x2.15

Show Top Score

Theoretical Performance

Clock Speed438MHz UC·Boost·OC	Clock Speed1980MHz UC·Base·Boost·OC
Peak OPS 1.84 PFLOPSFP8-32 Tensor	Peak OPS 3.96 POPSINT8 Tensor Sparse
Tensor FP8-16 - FP8-16 Tensor Sparse - Tensor FP8-32 1.84 PFLOPS FP8-32 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 FP16-16 - FP16-16 Tensor Sparse - Tensor FP16-32 459.3 TFLOPS FP16-32 Tensor Sparse -	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
BF16 - Tensor BF16 1.84 PFLOPS BF16 Tensor Sparse -	BF16 133.8 TFLOPS Tensor BF16 989.4 TFLOPS BF16 Tensor Sparse 1.98 PFLOPS
Tensor TF32 459.3 TFLOPS	Tensor TF32 494.7 TFLOPS
FP32 - Tensor FP32 229.6 TFLOPS	FP32 66.91 TFLOPS Tensor FP32 -
-	FP64 33.45 TFLOPS Tensor FP64 61.84 TFLOPS
-	Tensor INT8 1.98 POPS
Pixel Rate -	Pixel Rate 47.5 GPixel/s
Texture Rate -	Texture Rate 1.04 TTexel/s

Core Config

Shaders 16,384 Shaders	Shaders 16,896 Shaders
TMUs -	TMUs 528 TMUs
ROPs -	ROPs 24 ROPs
Tensor Cores 64 T-Cores	Tensor Cores 528 T-Cores
CUs 8 CUs	SMs 132 SMs

Clocks

Base Clock -	Base Clock 1.06GHz
Boost Clock 438MHz	Boost Clock 1.98GHz
Tensor Clock -	Tensor Clock 1.83GHz

Cache

L2 Cache -

L2 Cache 51.2MB shared

Memory

128GB HBM2e	141GB HBM3e
Memory Bus 8192-bit	Memory Bus 6144-bit
Memory Speed 3.6GT/s	Memory Speed 6.3GT/s
Memory Bandwidth 3.7TB/s	Memory Bandwidth 4.81TB/s
ECC Yes	ECC -
ESRAM 96MBSize 12800GB/sBandwidth	ESRAM -

Power & Thermals

TDP 900W

TDP 700W

Decoder

Decoder Model 14x Gaudi

Decoder Model 7x NVDEC 5

Card

Form Factor OAM Module	Form Factor SXM5
Cooling Passive	Cooling Passive

Chip

Manufacturer Intel	Manufacturer NVIDIA
Chip Designer Intel	Chip Designer NVIDIA
Architecture Gaudi 3	Architecture Hopper
Family Gaudi	Family Server
Branding	Branding
Codename Gaudi 3	Codename NV180
Market Segment Server	Market Segment Server
Release Date Apr 9, 2024	Release Date Nov 13, 2023

Fabrication

Foundry TSMC	Foundry TSMC
Fabrication Node N5	Fabrication Node 4N
Die Size -	Die Size 814mm²
Transistor Count -	Transistor Count 80 Billion
Transistor Density -	Transistor Density 98.28 MTr/mm²

Images

No images available

No images available

Competitors

IntelGaudi 3 vs GPU AMDInstinct MI250X IntelGaudi 3 vs GPU AMDInstinct MI250 IntelGaudi 3 vs GPU AMDInstinct MI300X Show 12 more

NVIDIAH200 SXM vs GPU NVIDIAB100 NVIDIAH200 SXM vs GPU NVIDIAB200 NVIDIAH200 SXM vs GPU NVIDIAGB200 Show 12 more