GPU Comparison

NVIDIA

GEFORCE

NVIDIA GeForce RTX 3090

CORE STATE GA102

VRAM 24 GB

CLOCK SPEED 1695 MHz

TDP 350 W

BUS WIDTH 384 bit

ARCHITECTURE Ampere

nm

PROCESS 8 nm

LAUNCH DATE 2020

VS

NVIDIA

GEFORCE

H200 NVL

CORE STATE GH100

VRAM 141 GB

CLOCK SPEED 1785 MHz

TDP 600 W

BUS WIDTH 6144 bit

ARCHITECTURE Hopper

nm

PROCESS 5 nm

LAUNCH DATE 2024

PERFORMANCE BENCHMARKS

3dmark_3dmark_steel_nomad_dx12

5,118

N/A

geekbench_opencl

191,142

305,608

geekbench_vulkan

174,310

N/A

passmark_directx_10

182

N/A

passmark_directx_11

220

N/A

passmark_directx_12

110

N/A

passmark_directx_9

268

N/A

passmark_g2d

1,063

N/A

passmark_g3d

26,645

N/A

passmark_gpu_compute

15,356

N/A

DETAILED SPECIFICATIONS

SPECIFICATION

RTX 3090

H200 NVL

Core Specs

Shading Units

10,496

16,896 +61.0%

Shaders

10,496

16,896 +61.0%

TMUs

328

528 +61.0%

ROPs

112

24 -78.6%

SM Count

82

132 +61.0%

Clocks

Base Clock

1395 MHz

1365 MHz

Boost Clock

1695 MHz

1785 MHz

Memory Clock

1219 MHz 19.5 Gbps effective

1593 MHz 6.4 Gbps effective

Memory

Memory Size

24 GB

141 GB

VRAM (MB)

24,576

144,384 +487.5%

Memory Type

GDDR6X

HBM3e

Memory Bus

384 bit

6144 bit

Bandwidth

936.2 GB/s

4.89 TB/s

Cache

L1 Cache

128 KB (per SM)

256 KB (per SM)

L2 Cache

6 MB

50 MB

Performance

Pixel Rate

189.8 GPixel/s

42.84 GPixel/s

Texture Rate

556.0 GTexel/s

942.5 GTexel/s

FP32 (TFLOPS)

35.58 TFLOPS

60.32 TFLOPS

FP64 (TFLOPS)

556.0 GFLOPS (1:64)

30.16 TFLOPS (1:2)

FP16 (TFLOPS)

35.58 TFLOPS (1:1)

241.3 TFLOPS (4:1)

AI/RT

RT Cores

82

—

Tensor Cores

328

528 +61.0%

Power

TDP

350 W

600 W

TDP (W)

350

600 +71.4%

Suggested PSU

750 W

1000 W

Power Connectors

1x 12-pin

8-pin EPS

Architecture

Architecture

Ampere

Hopper

GPU Name

GA102

GH100

Generation

GeForce 30

Server Hopper (Hxx)

Process Size

8 nm

5 nm

Transistors

28,300 million

80,000 million

Die Size

628 mm²

814 mm²

Foundry

Samsung

TSMC

Density

45.1M / mm²

98.3M / mm²

API Support

DirectX

12 Ultimate (12_2)

—

OpenGL

4.6

—

Vulkan

1.4

—

OpenCL

3.0

3.0

CUDA

8.6

9.0

Shader Model

6.8

—

Physical

Slot Width

Triple-slot

Dual-slot

Length

336 mm 13.2 inches

267 mm 10.5 inches

Height

140 mm 5.5 inches

111 mm 4.4 inches

Outputs

1x HDMI 2.13x DisplayPort 1.4a

No outputs

Bus Interface

PCIe 4.0 x16

PCIe 5.0 x16

Other

Launch Price

1,499 USD

—

Production

End-of-life

Active

Predecessor

GeForce 20

Server Ada

Successor

GeForce 40

Server Blackwell

View GeForce RTX 3090 Details View H200 NVL Details