AMD Instinct MI300P vs NVIDIA H100 Full Specs

Summary

9,728 Shaders 2.1GHz	14,592 Shaders 1.75GHz
64GB HBM32.66TB/s	80GB HBM2e2.04TB/s
Peak·FP32·FP16 81.72 TFLOPS	Peak·FP32·FP16 51.22 TFLOPS
Form Factor OAM Module	Form Factor PCIe Card
TDP 300W	TDP 350W
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 Tensor FP32 FP64 Tensor FP64 BF16 Tensor BF16 BF16 Tensor Sparse Tensor TF32 Tensor INT8 Tensor INT4 Memory Bandwidth TDP

Instinct MI300P2.62 POPSINT4 Tensor Sparse

x1

H1003.03 POPSINT8 Tensor Sparse

x1.16

Show Top Score

Theoretical Performance

Clock Speed2100MHz UC·Base·Boost·OC	Clock Speed1755MHz UC·Base·Boost·OC
Peak OPS 2.62 POPSINT4 Tensor Sparse	Peak OPS 3.03 POPSINT8 Tensor Sparse
Tensor FP8-16 1.31 PFLOPS FP8-16 Tensor Sparse 2.62 PFLOPS Tensor FP8-32 1.31 PFLOPS FP8-32 Tensor Sparse 2.62 PFLOPS	Tensor FP8-16 1.51 PFLOPS FP8-16 Tensor Sparse 3.03 PFLOPS Tensor FP8-32 1.51 PFLOPS FP8-32 Tensor Sparse 3.03 PFLOPS
Tensor FP16-16 653.7 TFLOPS FP16-16 Tensor Sparse 1.31 PFLOPS Tensor FP16-32 653.7 TFLOPS FP16-32 Tensor Sparse 1.31 PFLOPS	Tensor FP16-16 756.4 TFLOPS FP16-16 Tensor Sparse 1.51 PFLOPS Tensor FP16-32 756.4 TFLOPS FP16-32 Tensor Sparse 1.51 PFLOPS
BF16 - Tensor BF16 653.7 TFLOPS BF16 Tensor Sparse 1.31 PFLOPS	BF16 102.4 TFLOPS Tensor BF16 756.4 TFLOPS BF16 Tensor Sparse 1.51 PFLOPS
Tensor TF32 326.9 TFLOPS	Tensor TF32 378.2 TFLOPS
FP32 81.72 TFLOPS Tensor FP32 81.72 TFLOPS	FP32 51.22 TFLOPS Tensor FP32 -
FP64 40.86 TFLOPS Tensor FP64 81.72 TFLOPS	FP64 25.61 TFLOPS Tensor FP64 47.28 TFLOPS
Tensor INT4 1.31 POPS	Tensor INT4 -
Tensor INT8 1.31 POPS	Tensor INT8 1.51 POPS
Pixel Rate -	Pixel Rate 42.1 GPixel/s
Texture Rate 1.28 TTexel/s	Texture Rate 800.3 GTexel/s

Core Config

Shaders 9,728 Shaders	Shaders 14,592 Shaders
TMUs 608 TMUs	TMUs 456 TMUs
ROPs -	ROPs 24 ROPs
Tensor Cores 608 T-Cores	Tensor Cores 456 T-Cores
CUs 152 CUs	SMs 114 SMs

Clocks

Base Clock 1GHz	Base Clock 1.06GHz
Boost Clock 2.1GHz	Boost Clock 1.75GHz
Tensor Clock -	Tensor Clock 1.62GHz

Cache

L2 Cache 16.4MB shared	L2 Cache 51.2MB shared
L3 Cache 128MB shared	L3 Cache -

Memory

64GB HBM3	80GB HBM2e
Memory Bus 4096-bit	Memory Bus 5120-bit
Memory Speed 5.2GT/s	Memory Speed 3.2GT/s
Memory Bandwidth 2.66TB/s	Memory Bandwidth 2.04TB/s
ECC Yes	ECC -

Power & Thermals

TDP 300W

TDP 350W

Encoder

Encoder Model 2x VCN 2.6

-

Decoder

Decoder Model 2x VCN 2.6

Decoder Model 7x NVDEC 5

Card

Form Factor OAM Module	Form Factor PCIe Card
PCIe -	PCIe 2-Slots
-	Height 111 mm (4.37") Width 268 mm (10.55") Depth 40 mm (1.57")
Cooling Passive	Cooling Passive
Power Connectors -	Power Connectors 1x 16-Pin 12VHPWR

Chip

Manufacturer AMD	Manufacturer NVIDIA
Chip Designer AMD	Chip Designer NVIDIA
Architecture CDNA 3	Architecture Hopper
Family Instinct	Family Server
Branding	Branding
Codename Aqua Vanjaram	Codename NV180
Chip Variant Aqua Vanjaram XL	Chip Variant -
Market Segment Server	Market Segment Server
Release Date Dec 6, 2023	Release Date Mar 22, 2022

Fabrication

Foundry TSMC	Foundry TSMC
Other Foundries TSMCActive Interposer	Other Foundries -
Fabrication Node N5 N6Active Interposer	Fabrication Node 4N -
Die Size 740mm²	Die Size 814mm²
Transistor Count 73 Billion	Transistor Count 80 Billion
Transistor Density 98.65 MTr/mm²	Transistor Density 98.28 MTr/mm²

Images

No images available

No images available

Competitors

AMDInstinct MI300P vs GPU AMDInstinct MI300A AMDInstinct MI300P vs GPU NVIDIARTX PRO 5000 Blackwell AMDInstinct MI300P vs GPU NVIDIARTX 5880 Ada Show 12 more

NVIDIAH100 vs GPU NVIDIARTX PRO 5000 Blackwell NVIDIAH100 vs GPU NVIDIAB100 NVIDIAH100 vs GPU NVIDIARTX PRO 6000 Blackwell Max-Q Show 12 more