CPU and GPU Architecture Differences for Evaluation Boards

December 10, 2020 , in Blog

CPU with external pin connections

CPU external connection architecture

One of the sometimes underappreciated effects of the continuing expansion of computers into all aspects of our lives is how they have transformed learning. Prior to computers, it was a struggle for teachers to develop a pedagogy or teaching method that was truly aligned with the way many students actually learn the best. That is visual. For many, the ability to see something being done more rapidly translates into an activity they are able to understand and perform through mimicking than simply hearing it described. Proof positive of the adage “Give a man a fish, and you feed him for a day. Teach a man to fish, and you feed him for a lifetime. The origin of this thought is highly contested.”

Whether having a laptop at every desk, video lessons, virtual reality, or some other visual pedagogy tool is used, the result is that graphics-based learning is only possible due to the advancement of computing. Computers, of course, are driven by central processing units (CPUs) that perform or initiate all of the tasks requested by the operating program instructions. This includes communicating with the graphics processing unit (GPU), which is responsible for creating the imagery for display devices, such as monitors and projector screens. Let’s take a look at these two often complimentary electronic devices by comparing CPU vs GPU architecture and circuit board layout.

What is the Difference Between CPUs and GPUs?

Prior to discussing architecture and routing differences between CPUs and GPUs, it is probably informative to definitely draw a line between them in terms of attributes, as is done below.

Attributes of CPU and GPU

Attributes CPU GPU
Control logic More complex Simpler
Computational density Low High
Clock speed Faster Slower
Pipelines Low (< 30) High (hundreds)
Latency tolerance Low High
Processing structure Serial Parallel
Trends More parallel processing More complex logic

As shown above, there are distinct differences between CPUs and GPUs, which is not surprising as their objectives are different. The CPU is meant to control program execution. Specifically, this means ensuring that instructions are handled sequentially and associated operations, such as subroutines, are completed prior to invoking the next instruction. The GPU, on the other hand, is one of several peripheral devices that interact with the computer’s MPU, which contains the CPU. The function of the GPU is to access memory locations quickly and assemble or create imagery structures or frames that can be sent to a display device for view or interaction.

The parallel and serial operational design for the GPU and CPU, respectively, can be better understood by examining the processing structure shown in the table below, where the data handling characteristics of a general-purpose GPU (GPGPU) are contrasted with those of a CPU.

CPU vs GPU Processing Structure Comparison

Process structure comparison of CPU and GPU

CPU vs GPU data handling comparison 1

Now, that the operation and design intent of CPUs and GPUs is better understood, the discussion of CPU vs GPU architecture can be had.

1. Table from https://bensontao.wordpress.com/2013/06/11/gpu_in_automotive/. Accessed on 4/26/20.

Considerations for CPU versus GPU Architecture Layout for Evaluation Boards

Once, graphics boards or video cards - as they are sometimes called - were relegated to displaying (what we would call today low quality) imagery on computer screens. Today, however, these boards are found in all types of electronic devices that have graphics display capability. In most cases, they are still accompanied by a separate processor; and their primary function of rapidly retrieving, structuring (creating frames), and outputting imagery remains, albeit at a much better quality or resolution. In fact, the complexity of GPUs has reached the point where they are sometimes the primary semiconductor device under review for evaluation boards; instead of simply a peripheral. There are, of course, different considerations for laying out evaluation boards for CPU vs GPU architectures as shown below.

CPU vs GPU Architecture Layout

  • CPU Concerns
  • Instruction execution speed
        • As fast as possible
  • Minimizing latency 
        • Executing fetch through retire stages as quickly as possible
  • Multiple input/output data types
        • SPI
        • Power
        • GPIO
        • Multiplexer
        • Control
        • Image data
  • GPU Concerns
  • Memory access speed
        • As fast as possible
  • Memory buffer size
        • As large as possible
  • Frame size or image resolution 
        • Highest number of pixels
  • Image transfer bandwidth
      • As wide as possible

Advances in CPU technology have propelled computers into every industry and this growth coupled with improved GPU boards have opened the door to graphical applications that could not have been imagined a couple of decades ago.

Tempo‘s Custom Demo and Evaluation Board PCBA Manufacturing Service
  • ISO-9001, IPC-600, and IPC-610 commitment to quality certifications.
  • Accurate quote in less than a day.
  • DFX support, including DFM, DFA, and DFT from Day 1 of design.
  • Entire turnkey PCB manufacturing in as fast as 3 days.
  • Agile manufacturing process to quickly adapt to design changes.
  • Specialization in high-speed and HDMI trace optimization.
  • Reliable supplier component sourcing for quality, reliability, and traceability.
  • Performs multiple automated inspections during PCB assembly to ensure quality for prototyping and low volume production.
  • Comprehensive process documentation, including testing regimens and verifications.

However, building the evaluation boards that will lead to future capabilities requires an understanding of the important considerations for routing CPUs and GPUs. At Tempo Automation, we have the experience, expertise, and advanced manufacturing process to help you build the next generation of CPU and/or GPU evaluation board.

And to help you get started on the best path, we furnish information for your DFM checks and enable you to easily view and download DRC files. If you’re an Altium Designer or Cadence Allegro user, you can simply add these files to your PCB design software. For Mentor Pads or other design packages, we furnish DRC information in other CAD formats and Excel.

If you are ready to have your design manufactured, try our quote tool to upload your CAD and BOM files. If you want more information on CPU vs GPU architecture for evaluation boards, contact us.

The latest PCB news delivered to your inbox.

Search Sign In