Product Review

Everything you need to know about the Quadro GV100

April 26, 2018 • 3 min read


With greater research and exponential advancement in the world of deep learning, the demand for better and more effective GPUs is constantly increasing. The Quadro GV100, powered by NVIDIA’s Volta architecture, is transforming the whole game and is hence driving another wave of research in the deep learning industry.

The Quadro GV100 brings astonishing potential in simulation, rendering, and complex scientific calculations that can help engineers and scientists alike. Moreover, the generative design along with the capability of conducting complex simulation tasks can help businesses explore a greater range of design options. This would optimize the design for both their cost and performance, resulting in highly innovative products.

Along with the Quadro vDWS, this high-grade GPU is capable of addressing the constantly increasing requirements of the largest business industries in the world by accelerating deep learning infused applications and enabling surreal VR, providing greater security. The industries that the Quadro GV100 specifically impact include the automotive, architecture, healthcare, entertainment, and engineering industry.


This impressive piece of hardware is packed with 7.9 TFLOPS peak double-precision performance, 14.8 TFLOPS peak single-precision performance, and 59.3 TFLOPS peak integer operations performance. Along with that, it has a 118.5 TFLOPS deep learning performance.

With a PCI Express 3.0 x16 system interface, the Quadro GV100 contains an impressive number of 5120 CUDA Cores along with 640 Tensor cores.

It carries a capacity of 32 GB HBM2 (high bandwidth memory) with a bandwidth of 870 GB/s and a 4096-bit memory interface.

The Volta architecture

Volta architecture is the latest development by NVIDIA. This architecture is highly different from Pascal architecture. Differences include core layouts, thread execution and scheduling, memory controller, and more. Nevertheless, the new tensor cores employed in the Volta architecture can be the primary game-changers.

Tensor cores are a newly developed type of cores that are primarily less flexible (although they can still be programmed) and optimized for tensor math calculations. These cores are capable of performing an immense number of matrix multiplications within a single unit. In general, the performance of a single tensor core is equivalent to the operations of 64 FMA per clock. Since the Quadro GV100 contains 8 tensor cores in each SM, there are about 1024 FLOPS in every clock per SM.

Quadro GV100 vs. Quadro GP100
  • Architecture

The Quadro GV100 and Quadro GP100 are inherently different from each other. While the GV100 is built on the latest Volta architecture (specifically designed for workstations), the GP100 uses the Pascal architecture (which performs decently well for desktop-based tasks).

  • Memory

Both of these GPUs make use of a PCIe 3.0 x16 interface. The Quadro GV100 has a memory capacity of 32 GB HBM2 while the Quadro GP100 carries i.e. 16 GB HBM2.

  • Floating-point performance

While the Quadro GV100 carries a floating-point performance of 14,817 gFLOPS, the Quadro GP100 offers 10,329 gFlOPS.

  • Power consumption

Due to heavier usage and a greater number of calculations, the Quadro GV100 uses 250 Watts of power which is slightly greater than the 235 Watts of power used by the Quadro GP100.

  • Pipelines

The Quadro GV100 contains a surprisingly high number of 5120 CUDA cores along with 640 tensor cores, whereas the Quadro GP100 contains 3584 CUDA cores.


The Quadro GV100 drives high-end research for deep learning. The latest Volta architecture is capable of carrying out a high number of complex matrix calculations in a single unit. This is what makes it an effective tool to be used in deep learning applications.

From a gaming perspective or desktop usage, however, the Quadro GP100 would be a better choice. In this case, the Quadro GV100 is not for gaming purposes because it is optimized for professional workstations.







Related Content