NVIDIA H200 SXM 141 GB vs NVIDIA GeForce RTX 5090 D
We compared two Desktop platform GPUs: 141GB VRAM H200 SXM 141 GB and 32GB VRAM GeForce RTX 5090 D to see which GPU has better performance in key specifications, benchmark tests, power consumption, etc.
Main Differences
NVIDIA H200 SXM 141 GB 's Advantages
More VRAM (141GB vs 32GB)
Larger VRAM bandwidth (4.89TB/s vs 1.79TB/s)
NVIDIA GeForce RTX 5090 D 's Advantages
Boost Clock has increased by 22% (2407MHz vs 1980MHz)
4864 additional rendering cores
Lower TDP (575W vs 700W)
Score
Benchmark
FP32 (float)
H200 SXM 141 GB
66.91 TFLOPS
GeForce RTX 5090 D
+56%
104.8 TFLOPS
Graphics Card
Nov 2024
Release Date
Jan 2025
Server Hopper
Generation
GeForce 50
Desktop
Type
Desktop
PCIe 5.0 x16
Bus Interface
PCIe 5.0 x16
Clock Speeds
1500 MHz
Base Clock
2017 MHz
1980 MHz
Boost Clock
2407 MHz
1593 MHz
Memory Clock
1750 MHz
Memory
141GB
Memory Size
32GB
HBM3e
Memory Type
GDDR7
6144bit
Memory Bus
512bit
4.89TB/s
Bandwidth
1.79TB/s
Render Config
-
-
-
132
SM Count
170
16896
Shading Units
21760
528
TMUs
680
24
ROPs
176
528
Tensor Cores
680
-
RT Cores
170
256 KB (per SM)
L1 Cache
128 KB (per SM)
50 MB
L2 Cache
96 MB
-
-
-
Theoretical Performance
47.52 GPixel/s
Pixel Rate
423.6 GPixel/s
1045 GTexel/s
Texture Rate
1637 GTexel/s
267.6 TFLOPS
FP16 (half)
104.8 TFLOPS
66.91 TFLOPS
FP32 (float)
104.8 TFLOPS
33.45 TFLOPS
FP64 (double)
1.637 TFLOPS
Board Design
700W
TDP
575W
1100 W
Suggested PSU
950 W
No outputs
Outputs
1x HDMI 2.1b
3x DisplayPort 2.1b
8-pin EPS
Power Connectors
1x 16-pin
Graphics Processor
GH100
GPU Name
GB202
-
-
-
Hopper
Architecture
Blackwell 2.0
TSMC
Foundry
TSMC
5 nm
Process Size
5 nm
80 billion
Transistors
92.2 billion
814 mm²
Die Size
750 mm²
Graphics Features
N/A
DirectX
12 Ultimate (12_2)
N/A
OpenGL
4.6
3.0
OpenCL
3.0
N/A
Vulkan
1.4
9.0
CUDA
10.1
N/A
Shader Model
6.8
