The quadro gp100 arrives with 16 gb of hbm2 memory for workstations

Nvidia has gone through the SOLIDWORKS World event to announce a new professional card based on the powerful Pascal GP100 silicon, the most powerful and that we will not see on any card intended for video games. It is the new Quadro GP100 that seeks to offer all the best of Pascal in a lower range than the Tesla.

Nvidia Quadro GP100: features, availability and price

The new Quadro GP100 becomes the new top of the range series above the Quadro P6000 that was announced last summer with the GP102 chip. The new card offers great performance with a computing power of 10.3 TFLOPs in single precision, 20.7 TFLOPs in medium precision operations, 5.7 TFLOPs in double precision operations and a total of 16 GB of HBM2 memory with a bandwidth of 716 GB / s. Its most important features continue with error correction technology (ECC) that makes it valid for high performance tasks (HPC) and CAD / CAE that require high reliability and precision.

The Nvidia Quadro GP100 needs an 8-pin PCI-Express connector to cover its 235W TDP and includes four video outputs in the form of DisplayPort 1.4 and a DVI-D. It supports the NVLink interface for SLI configurations with up to two cards. Nothing has been said about its price but it will not be cheap considering that the Quadro P6000 is above 6000 euros, it will arrive in March.

NVIDIA Quadro Specifications
	GP100	P6000	M6000	K6000
CUDA Cores	3584	3840	3072	2880
TMUs	224	240	192	240
ROPs	128?	96	96	48
Turbo frequency	~ 1430MHz	~ 1560MHz	~ 1140MHz	N / A
Memory frequency	1.4 Gbps HBM2	9Gbps ​​GDDR5X	6.6Gbps GDDR5	6Gbps GDDR5
Bus	4096-bit	384-bit	384-bit	384-bit
VRAM	16 GB	24GB	24GB	12GB
ECC	Yes	Do not	Do not	Yes
FP64	1/2 FP32	1/32 FP32	1/32 FP32	1/3 FP32
TDP	235W	250W	250W	225W
GPU	GP100	GP102	GM200	GK110
Architecture	Pascal	Pascal	Maxwell 2	Kepler
Node	TSMC 16nm	TSMC 16nm	TSMC 28nm	TSMC 28nm
Launching	March 2017	October 2016	22/03/2016	23/07/2013

Source: anandtech