NVIDIA H200 SXM 141 GB vs NVIDIA Tesla S2050

Comparison of NVIDIA H200 SXM 141 GB with 141 GB HBM3e and 16,896 cores vs NVIDIA Tesla S2050 with 3 GB GDDR5 and 448 cores.

Loading...

Performance Rating

A100 A100
H200 H200
MI325X MI325X

NVIDIA H200 SXM 141 GB

67.4

NVIDIA H200 SXM 141 GB

67.4
RX 7900 XTX RX 7900 XTX
MI250 MI250
Instinct MI300X Instinct MI300X

NVIDIA Tesla S2050

0.0

NVIDIA Tesla S2050

0.0

Contents:

Memory ML Performance Compute Power Architecture & Compatibility ML Software Support Clocks & Performance Power Consumption Rendering Benchmarks Additional

Memory

Memory Size

🔥 +4,600% 141 ГБ
3 ГБ ×4 (12 ГБ)

Memory Type

HBM3e GDDR5

Memory Bandwidth

🔥 4.89 TB/s
148.4 GB/s ×4 (593.6 GB/s)

Memory Bus Width

6,144 бит 384 бит ×4 (1536 бит)

ML Performance

FP16 (Half Precision)

🔥 267.6 TFLOPS
No

BF16 (Brain Float)

No No

TF32 (TensorFloat)

No No

Compute Power

FP32 (Single Precision)

🔥 66.91 TFLOPS
No

FP64 (Double Precision)

🔥 +6,409% 33.45 TFLOPS
0.5139 TFLOPS

CUDA Cores

🔥 +3,671% 16,896
448 ×4 (1792)

RT Cores

No No

Architecture & Compatibility

GPU Architecture

Hopper Fermi

SM (Streaming Multiprocessor)

🔥 +843% 132
14

PCIe Version

PCIe 5.0 x16 PCIe 2.0 x16

ML Software Support

CUDA Version

🔥 9.0
2.0

Clocks & Performance

Base Clock

🔥 1,500
No

Boost Clock

🔥 1,980
No

Memory Clock

🔥 +106% 1,593
773

Power Consumption

TDP/TGP

🔥 -22% 700 W
900 W

Recommended PSU

🔥 -15% 1100 W
1300 W

Power Connector

8-pin EPS No

Rendering

Texture Units (TMU)

🔥 +843% 528
56 ×4 (224)

ROP

No No

L2 Cache

🔥 50 MB
768 KB

Benchmarks

MLPerf, llama2-70b-99.9 (UNSET)

3 534 tokens/s

MLPerf, llama2-70b-99.9 (fp16)

3 553 tokens/s

MLPerf, llama2-70b-99.9 (fp8)

2 444 tokens/s

MLPerf, llama3.1-405b (fp16)

40.8 tokens/s

MLPerf, llama3.1-405b (fp8)

25.3 tokens/s

MLPerf, llama3.1-8b (fp8)

5 161 tokens/s

MLPerf, deepseek-r1 (fp8)

1 113 tokens/s

MLPerf, mixtral-8x7b (fp8)

7 132 tokens/s

Additional

Slots

🔥 SXM Module
Dual-slot

Release Date

Nov. 18, 2024 July 25, 2011

Display Outputs

No outputs
No outputs

Renting is cheaper than buying