Knowledge Base

More Information on Graphics Cards (GPUs)

Aim

• To provide information about graphics cards and their software needed to perform video tracking with the Deep Learning technique.
• This topic does not apply if you track your subject with the Contour-Based method.

NOTE: Technically, the GPU indicates the main processor in the graphics card. The graphics card includes other components such as the video memory (VRAM), the printed circuit board (PCB), a power management unit, the PCIe connectors, and the display connectors. In this topic, GPU is assumed to coincide with the physical board that you install on your PC.

GPU Microarchitecture

Graphics cards are categorized based on their microarchitecture, also known as generations. To find which microarchitecture category your GPU belongs to, browse to https://en.wikipedia.org/wiki/CUDA.

• Under the table Compute Capability, GPU Semiconductors and Nvidia GPU Board Products, locate the name of the GPU in the column GeForce. The name under Micro-architecture tells which generation the GPU belongs to.
• In the table Compute Capability (CUDA SDK Support vs. Microarchitecture) you can see whether that GPU generation works with the CUDA SDK version used with EthoVision XT, currently 12.2.

GPU Speed

When comparing two GPUs for deep learning applications, one of the best indicators is the memory bandwidth, rather than single properties like the GPU memory speed. A GPU's memory bandwidth determines how fast it can move data from and to the memory (VRAM) to the computation cores.

Theoretically, the memory bandwidth depends on:

• The memory clock speed (in Hz).
• The width of the data bus between the card memory and the graphics processor, in bits (also known as memory interface). This is the physical count of bits that may fit along the bus every clock cycle.
• The memory type (the so-called memory clock type multiplier).

Memory bandwidth is measured in GB/s. For example, a graphics card with bandwidth of 1000 GB/s is expected to perform two times faster than one with 500 GB/s. This is a rough indication of the difference in performance, as real performance may depend on several other factors, including the neural network architecture. If there is no reading from memory at any particular clock cycle, that cycle's worth of bandwidth goes unused and cannot be stored for later use. Because GPUs tend to acquire data in spurts, the usable bandwidth is generally lower than the numerically available bandwidth.

Another useful measure is the number of Floating Point Operations per Second (FLOPS). For example:

• NVIDIA Quadro P2200: 3.8 TFLOPS (teraFLOPS, equal to 10¹² FLOPS).
• NVIDIA T1000: 2.5 TFLOPS.
• NVIDIA GeForce RTX 4060: 15.1 TFLOPS.

For the specifications of graphics cards, see: https://www.techpowerup.com/gpu-specs/

GPUs Tested and Supported

The following NVIDIA graphics cards have been successfully tested with Deep Learning-based tracking.

TIP: Install the most recent driver version available on the NVIDIA website for your GPU.

Desktop

NOTE: The GeForce RTX 4060 is a dual-slot GPU, occupying two PCIe slots. There are several versions of this board, with two or three fans. Make sure it can fit in your PC. Both the T1000 and the MSI GeForce 4060 Ventus 2X (dual fan) fit in the Dell Precision 3680 desktop computer.

Laptop

NOTE: If you install a new GPU on a computer that you have previously used to acquire data with Deep Learning, you must update the neural network model stored on your computer. See Update the Neural Network Model After Installing a New GPU.

Performance

A few NVIDIA GPUs were tested in an EthoVision XT experiment with four arenas, one subject per arena, using Deep Learning as the body point detection technique. A 5-minute video of resolution 1280 x 1024 at 30 fps was used. The average time taken to acquire one trial was as follows:

• Quadro P2200: 4 min 29 s.
• T1000: 4 min 24 s.
• GeForce RTX 4060 Ventus 2X: 1 min 58 s.

Importance of CUDA

The driver of the graphics card must support the CUDA component. CUDA is a parallel computing platform and programming model created by NVIDIA that helps speed up applications by using the power of graphics cards.

Each graphics card may have its own latest driver version in Windows Update. For the cards specified above, Windows Update installs the needed minimal version. However, other cards may need more recent versions than those available through Windows Update. When a driver version is insufficient for supporting CUDA, EthoVision XT displays the following message in the Experiment Settings: The graphics card driver is outdated or not installed, or the graphics card does not support CUDA.

It is recommended to download and install the latest driver from the card's manufacturer. This should make the card work properly with CUDA, provided that the hardware supports that version. To find out whether your NVIDIA graphics card is CUDA-enabled, refer to: https://developer.nvidia.com/cuda-gpus

CUDA Version

If your GPU is already installed on the PC, follow these steps to find out which version of CUDA is installed:

1. In Windows Explorer, locate the folder C:\ProgramData\Noldus\Components\Ethovision\TrackerInterfaceNN\[version number].
2. In that folder, find the file named CudaReport.txt.
3. Open this file. The runtime version should give an indication of the version of CUDA currently present on your PC.

If you do not see the folder specified above, create an experiment and in the Experiment Settings under Body Point Detection Technique select Deep Learning. Then check the folder again as described in step 1.

NOTE: The version of CUDA is not the same as the version of the GPU driver.

Which GPU Should I Choose?

• Choose a recent NVIDIA GPU of the 1000, 2000, 3000 or 4000 series. However, a very recent GPU or one of a new generation may not be compatible. Contact Noldus if in doubt.
• In principle, a graphics card of the Pascal microarchitecture category, with 4 GB memory, 1024 CUDA cores and memory bandwidth up to 140 GB/s represents the minimum specifications. Click one of the links below depending on the camera you use and locate the row with a specific graphics card and DL (Deep Learning) specified in the rightmost column:
  • GigE cameras
  • USB 3.0 cameras
  • Analog cameras
• The number of CUDA cores can be a useful indicator of performance when comparing GPUs within the same microarchitecture. However, when comparing cards between generations (e.g. Pascal vs. Turing), the difference in CUDA core count does not reliably predict actual performance. An older card with more CUDA cores may not perform as well as a more recent card with fewer CUDA cores.
• Overall, newer generations perform better. For example, an Ampere GPU should perform better than a Turing GPU.
• GPUs of the Maxwell generation / 900 series are not supported with EthoVision XT 19.
• A more powerful GPU requires more power from the PC. Make sure your PC has enough power to feed the GPU. See the prerequisites in Install a Graphics Card for Deep Learning.
• Performance of a graphics card depends not only on its own characteristics but also heavily on the sample rate and resolution chosen in EthoVision XT. See Video Source.

See Also

• Install a Graphics Card for Deep Learning
• Deep Learning: Requirements