Whitehat Virtual Technologies Blog

CPU vs. GPU

Written by Brad Truman | May 18, 2014 9:14:00 AM

We are hearing a lot about graphics processing units or “GPUs” these days, so what are they and what makes them valuable? We’ll shed some light on this in this article.

Most people are aware of “CPUs”, or central processing units, since that term has been around for several decades now. Intel invented the microprocessor in the late 1970s, and the most famous one was in the original IBM PC from 1982, and it was called the 8088. It had a 16-bit instruction set and was very powerful for its time. Boy, have things come a long way since then! Now there are multiple CPU “cores” on each physical CPU chip that are all loosely combined together to do simultaneous multi-processing (see the diagram below). Each core is a CPU, and there can be as many as 12 cores on a physical CPU chip. Each core can process independently, so more work can get done.

If you remember back from the early days of the PC, you could buy a “Math Co-Processor” to aid in numerical/scientific calculations. This concept has long since been incorporated into the primary CPU chips, but the concept is the same for the GPU. So what makes the GPU different, and necessary at all?

With a CPU core processes data in a serial format… one task at a time, so you are limited to simultaneous processing only up to the limit of the number of cores you have. Graphics manipulation requires that thousands, even millions of points be manipulated simultaneously, so the GPU was invented to do that. It has dozens/hundreds of cores that can process in a massively parallel fashion on one task… manipulating a single graphics image. The GPU has a way of dividing up that image that needs to be processed and acting on each of those image portions simultaneously, providing the tremendous processing power needed to display 3D images, and motion on those images. Think of it as today’s math co-processor.

Ok, so how does this fit in with virtual desktop infrastructure VDI hosted desktops? As you can imagine, manipulating and displaying a 3D image, especially where motion is involved, is difficult enough when you are within the same workstation computer. Trying to do that with a hosted VDI session, and then displaying that to a remote workstation over LAN, or even worse…WAN connectivity, has been an area that people didn’t even attempt… until now.

Citrix XenDesktop combines its low bandwidth and graphics compression capability with the new shared vGPU capabilities developed jointly with NVIDIA, to open up a whole new world. This solution provides:

  • Low-bandwidth access to state-of-the-art VDI sessions.
  • Cost-effective sharing of expensive GPU resources among many users.
  • The ability to use low-powered workstations, thin clients, and tablets.
  • Effective, full-function access to full versions of graphical programs remotely.

This is the next frontier in virtualization, and it’s about to explode.