NVIDIA HGX GPU 8x A100 640GB 935-23587-0000-204

0 out of 5 (0)

✓ Architecture: Ampere

✓ Theoretical Performance: 156 TFLOPS

✓ Graphics Memory Type: HBM2

✓ Number of Stream Processors: 55,296

✓ Output to the Monitor: None

✓ Profile: HGX

✓ Bus: HGX

✓ Consumption (W): 1500 W

✓ Part Number: 935-23587-0000-204

✓ Alternative Part No.: GPU-NVHGX-A100-88D-EX2Y

✓ Supermicro Part No.: GPU-NVTHGX-A100-SXM4-88D

✓ Manufacturer: NVIDIA

✓ Mass: 2 kg

✓ PCI Express: Generation 4.0, 2 GB/s per line, 16 lines

✓ CUDA Technologies: CUDA stream processors

2 weeks lead time. All sales final. No returns or cancellations. For bulk inquiries, consult a live chat agent or call our toll-free number.

CYCLESAFE™ Plus

Get this product for

$169,000.00

Get it in 10 days

Estimate for 682345

Will be delivered to your location via DHL

Inquiry to Buy

Description Specifications Download Reviews (0)

Description

NVIDIA HGX GPU 8x A100 640GB 935-23587-0000-204

High-Performance NVIDIA Server Graphics Cards
NVIDIA products represent a range of professional graphics cards intended for both rackmount servers and powerful workstations. These are high-quality graphics products
with large memory and a special set of drivers, which are adapted for work in CAD and other professional applications, remote collaboration, virtual desktop,
3D visualization and rendering, and machine learning. Compared to desktop graphics cards, they are often optimized for lower power consumption.

Cutting-Edge Technology and Performance
NVIDIA server graphics cards feature the Ampere graphics chip with a theoretical performance of 156 TFLOPS. They utilize VRAM, primarily GDDR, which offers faster buffering of data
through memory chips with higher throughput and multiple data transfer rates. The CUDA architecture enables the use of graphic CUDA stream processors, significantly speeding up specific
calculations compared to classic processors with fewer cores. Passively cooled, these cards leverage optimized airflow in server environments to dissipate heat efficiently. The latest PCI
Express 4.0 interface ensures high performance and throughput with a speed of 2 GB/s per line, using 16 lines to connect to the processor.

Specifications

Specification	Details
Architecture	Ampere
Theoretical Performance	156 TFLOPS
Graphics Memory Type	HBM2
Number of Stream Processors	55,296
Output to the Monitor	None
Profile	HGX
Bus	HGX
Consumption (W)	1500 W
Part Number	935-23587-0000-204
Alternative Part No.	GPU-NVHGX-A100-88D-EX2Y
Supermicro Part No.	GPU-NVTHGX-A100-SXM4-88D
Manufacturer	NVIDIA
Mass	2 kg
PCI Express	Generation 4.0, 2 GB/s per line, 16 lines
CUDA Technologies	CUDA stream processors