Print this article Edit this article
GPU Recommendations for 3D rendering and computation
GPU recommendations for Engineering Software
Many pieces of engineering software can leverage a system’s GPU for higher performance not just in 3d applications, but for general computing processes as well. This document is meant to provide resources from some software developers as well as reference data for GPUs.
MatLab:
To speed up computation, Parallel Computing Toolbox leverages NVIDIA GPUs with compute capability 2.0 or higher. See the compute capabilities of all NVIDIA GPUs. MATLAB does not support computation acceleration using AMD or Intel GPUs at this time.
Dassault Systems (Catia)
Solidworks 2018:
Ansys:
The powerful GPU computing capabilities in ANSYS was developed on NVIDIA Professional GPUs. A CUDA-capable NVIDIA GPU (CUDA v4.2 or above) should be used. Suggested GPUs from the NVIDIA® Tesla® Kepler™ product family or NVIDIA® Quadro® Kepler product family are recommended.
NVIDIA GPU computing products are designed to deliver the highest computational performance to accelerate computations in ANSYS software.
Generally, more CUDA cores equal more compute power but as newer architectures are released newer cards with similar core counts will likely be faster and more energy efficient overall.
Another note to consider is that Quadros are specifically designed for double-precision computation, making them 2x-4x faster in those scenarios.
Another consideration to keep in mind is that 3d over Remote Desktop Protocol in Windows is limited/slow/buggy for GeForce cards due to drivers, where it is not for Quadros. If you plan on regularly doing work via Remote Desktop on a Windows system we would recommend purchasing a Quadro.
Model | Cuda Cores | Memory | Minimum Power Supply* |
Titan V | 5120 | 12GB | 700W |
Titan Xp | 3840 | 12GB | 600W |
GTX 1080Ti | 3584 | 11GB | 600W |
GTX 1080 | 2560 | 8GB | 500W |
GTX 1070Ti | 2432 | 8GB | 500W |
GTX 1070 | 1920 | 8GB | 500W |
GTX 1660Ti | 1535 | 6GB | 450W |
GTX 1660 | 1408 | 6GB | 450W |
GTX 1060 | 1280 | 6GB | 400W |
GTX 1650** | 896 | 4GB | 300W |
GTX 1050Ti** | 768 | 4GB | 300W |
GTX 1050 | 640 | 2GB | 300W |
GTX 1030 | 384 | 2GB | 300W |
CUDA-Capable Quadro GPUs
Model | Cuda Cores | Memory | Minimum Power Supply* |
P6000 | 3840 | 24GB | 850 |
M6000 | 3072 | 12GB | 600 |
K6000 | 2880 | 12GB | 500 |
P5000 | 2560 | 16GB | 750 |
K5200 | 2304 | 8GB | 500 |
M5000 | 2048 | 4GB | 500 |
P4000 | 1792 | 8GB | 750 |
M4000 | 1664 | 8GB | 500 |
K5000 | 1536 | 4GB | 500 |
K4200 | 1344 | 4GB | 500 |
P2000 | 1024 | 5GB | 600 |
K4000 | 768 | 3GB | 450 |
K2200 | 640 | 2GB | 350 |
P1000 | 640 | 4GB | 500 |
K1200 | 512 | 4GB | 350 |
K620 | 384 | 2GB | 300 |
P600 | 384 | 2GB | 400 |
P400 | 256 | 2GB | 300 |
*Power supply recommendations are relative- higher wattage supplies may be required depending on the CPU or any additional peripherals/expansion cards. Depending on the model/make of the PC or Power Supply, sufficient power connectors may or may not be available even if wattages are sufficient. Feel free to ask us for assistance if you are unsure if your PC meets the requirements for a GPU.
**Do not require PCIe power rails
Reference Link for 3D Performance Benchmarks: https://www.videocardbenchmark.net/high_end_gpus.html
Last Modified:
May 13, 2020 9:46 am America/New_York
Created:
Mar 22, 2018 10:44 am GMT-4
by
admin
JumpURL: