With 16GB of HBM2 vRAM, the PNY NVIDIA Quadro GP100 Graphics Card is the first workstation board from NVIDIA to utilize the HMB2 memory interface. HBM2 memory dies sit atop an interposer and connect directly with the GPU die, rather than separately from it. This design facilitates a memory bus that is an order of magnitude wider than traditional GDDR memory implementations, in this case, 4096 bits for a 717 GB/s total bandwidth. Otherwise, this card offers 3584 CUDA cores providing up to 10.3 TFLOPS of single-precision computing power. For applications that require double precision and half-precision computing, such as scientific multiphysics simulations and machine learning, the GP100 can crank out 5.2 and 20.7 TFLOPS in those operations, respectively.
The output panel on this card features four DisplayPort terminals and a dual-link DVI-D port. Up to four monitors can be simultaneously supported by a single card, without any extra hardware, to create high-resolution video walls with your workstation or for more desktop real-estate to increase productivity. DisplayPort 1.4 support allows this card to drive UHD, DCI 4K, 5K, and 8K displays. The DVI port and included DisplayPort to DVI adapters can be used for connecting to legacy hardware while other separately available adapters can be used with this card to drive other kinds of displays and projectors.
For cooling, the GP100 uses a blower-style cooler. This cooler uses a radial fan to force air through the card's heatsink and out through the rear of the case. This cooling configuration can be used in crowded workstation environments with multiple GPUs or where space between computer components can be a concern, as hot air is not recirculated back into the case.