H200 SXM 141 GB
Graphics Card H200 SXM 141 GB — Specifications, Tests & Benchmarks
H200 SXM 141 GB
GeForce Hxx
•
GH100
•
18.11.2024
•
141 GB
•
HBM3e
•
6144 bit
•
700W
🏆 Top #14129
Performance
Synthetic Score
100
Efficiency
low
The H200 SXM 141 GB graphics adapter is based on the advanced Hopper architecture. NVIDIA announced this model on 18.11.2024. The graphics processor is manufactured using a modern process. The 814 mm² die contains approximately 80.0 billion transistorsistors transistors. The core configuration includes 16896 shader units, 0 RT Cores for ray tracing, and 528 Tensor Cores for AI tasks. The HBM3e video memory capacity is 141 GB. Thanks to the 6144 bit-bit bus, bandwidth reaches 5007.36 GB/s, eliminating delays when working with demanding textures. The graphics chip operates at a base clock of 1500 MHz with dynamic acceleration capability up to 1980 MHz. The nominal thermal design power is 700 W. For stable system operation, a PSU of 1100 W or more will be required.
General Information
| Release Date | 18.11.2024 |
| Production Status | Active |
| Usage Class | Office Use |
| Market Segment | Desktop |
| Series | GeForce Hxx |
GPU
| Architecture | Hopper |
| Codename | GH100 |
| Process Node | 5 nm (TSMC) |
| Transistor Count | 80,000,000.0 million |
| Die Size | 814 mm² |
| Transistor Density | 98,300,000 million / mm² |
Rendering Configuration
| Shading Units (ALU) | 16896 |
| Texture Mapping Units (TMU) | 528 |
| Render Output Units (ROP) | 24 |
| Tensor Cores | 528 |
| Compute Units (SM/CU) | 132 |
Clock Speeds
| Base Clock | 1500 MHz |
| Boost Clock | 1980 MHz |
| Memory Clock (Actual) | 1593 MHz |
| Memory Clock (Effective) | 6400 MHz |
Memory
| Memory Capacity | 141 GB |
| Memory Type | HBM3e |
| Memory Bus | 6144 bit bit |
| Memory Bandwidth | 5007.36 GB/s |
Cache Memory
| L1 Cache | 33792 KB |
| L2 Cache | 50 MB |
Board Design & Power
| Interface | PCIe 5.0 x16 |
| Slot Width | 0 |
| Power Consumption (TDP) | 700 W |
| Recommended PSU | 1100 W |
| Power Connectors | 8-pin EPS |
API Support
| OpenCL | |
| CUDA |
Theoretical Performance
| Pixel Fill Rate | 47520 GPixel/s |
| Texture Fill Rate | 1045400 GTexel/s |
| FP16 Performance (Half) | 267.60 TFLOPS (4:1) |
| FP32 Performance (Float) | 66.91 TFLOPS |
| FP64 Performance (Double) | 33.45 TFLOPS (1:2) |
Synthetic Tests & Benchmarks
| Performance Tier | low |
Всі тести
Тестів поки немає.
Comments
0Please log in to your account to leave a comment.
Log In or RegisterNo comments yet. Be the first to comment!
Are you sure you want to delete this comment? This action cannot be undone.