You have no items in your shopping cart.
What is the NVIDIA Quadro GP100?
Until now, Pascal architecture combined with high-speed memory HBM of the second generation were used only in the rack based server environments, with NVIDIA’s latest Tesla P100. NVIDIA have now release the Quadro GP100 to fill this gap in the market for Desktop Workstation based users wanting to perform deep learning, FEA, Simulation or Analysis workflows.
The Quadro GP100 is based on the same architecture as the Tesla P100 which was initially released as the, compute card installed in NVIDIA’s supercomputers DGX-1. But after a few months, NVIDIA introduced the PCI Express version, which can work with 12 or 16 GB of HBM2 memory.
Now NVIDIA have broadened the availability of this supercomputing performance by offering actively cooled Quadro GP100 perfectly suited to office based Workstation solutions. In some areas, an even greater level of performance is required, so the demand for computing resources in Workstation solutions continues to grow. And there is no doubt that it will continue to grow in the future. Until now, the fastest professional graphics cards in the family were Quadro P5000 and P6000 based on the GPU GP102 and GP104 were perfectly suited for those working with visualisation and Media & Entertainment but the Quadro GP100 is more suited for those working with big data computation and scientific areas.
How Does it Perform?
The graphics processor has 3,584 stream processors in it. FP16 computing is performed with a capacity of 20.7 TFLOPS, FP32 computing can make with the performance of 10.3 TFLOPS and complex FP64 computing produces a massive 5.2 TFLOPS of performance. As a result, the performance of Quadro GP100 is even better than Tesla P100’s performance and currently the most powerful workstation based compute card in the world for FP64 double precision performance.
16 GB of HBM2 memory is connected by 4,096-bit interface with a bandwidth of 720 GB/s. Available video outputs of the video card: 4x DisplayPort 1.4 and 1x DVI which can support 4x4.096x2.160 pixels at 120 Hz or 4x 5.120x2.880 pixels at 60 Hz.
As a special function Quadro GP100 accelerators can be combined with help of NVLink now available via PCIe connection. Like SLI, for the connections are used four channels with a capacity of 160 GB/s. Physically, the connection looks like two bridges resembling SLI.
According to NVIDIA sources, in tasks on the basis of deep study GP100 model delivers 20.7 16-bit teraflops. It can be used as a development platform for the introduction of deep study (deep learning) environment in Windows and Linux.
NVIDIA Pascal is the most powerful architecture of NVIDIA GPUs, which has ever created to date. With a new video card Quadro GP100, you can easily work on any innovative products. With this video card, you will be able to complete even the most complex simulation in record times.
- 3584 CUDA Cores
- 16GB HBM2 Memory
- 5.2 TFLOPS FP64 Performance
- 10.2 TFLOPS FP32 Performance
- 20.7 TFLOPS FP16 Performance
- 4x DisplayPort 1.4 & 1x DVI-D Connectors