Tesla S870

Specifications of the NVIDIA Tesla S870 GPU

Name / Brand / Architecture
Manufacturer:	NVIDIA
Model:	Tesla S870
Reference card?:	Yes
Target market segment:	Server / High Performance Computing
Die name:	4 x G80
Fabrication process:	90 nm
Transistors:	681 million
Bus interface:	PCI-E 1.0 x 16
Launch date:	May 2007
Frequency
Graphics clock:	600 MHz
Processor clock:	1350 MHz
Memory specifications
Memory size:	4 x 1536 MB
Memory type:	GDDR3
Memory clock:	800 MHz
Memory clock (effective):	1600 MHz
Memory interface width:	4 x 384-bit
Memory bandwidth:	4 x 76.8 GB/s
Cores / Texture
CUDA:	1.0
CUDA cores:	4 x 128
ROPs:	4 x 24
Texture units:	4 x 32
RAMDACs:	400 MHz
Electric characteristics
Maximum power draw:	800 W
Performance
Pixel fill rate:	4 x 14.4 Gigapixels/s
Texture fill rate:	4 x 19.2 Gigatexels/s
Single presision compute power:	4 x 345.6 GFLOPS
Dimensions
Width:	Dual-Slot
Other features / Support
OpenGL support:	3.3
DirectX support:	10.0
Shader model:	4.0
Notes
Tesla S870 is a rackmount GPGPU solution that connects to the system via a PCI-E 1.0 card.