<< Click to Display Table of Contents >> Navigation:  XStream® HDVR® Background > CPU vs. GPU

C++ C++ Java Java .NET .NET

Summary

 

Fovia’s XStream® HDVR® software runs on off-the-shelf CPUs, circumventing the constraints of restrictive GPU volume rendering solutions, and allowing OEMs to offer cost-effective, flexible advanced visualization to their customers. Fovia’s CPU-based solution surpasses the performance and quality of current hardware-based solutions.

 

In the imaging world, some people may have the misconception that it is difficult to have top-of-the-line quality and performance using a CPU. Additionally, developers may assume that a GPU solution is “easier to implement.” We do our best to educate people about the pitfalls, so they can make an informed decision. This page will help you understand the benefits of Fovia’s CPU-based high performance solution over a GPU-based solution.

 

General Information

 

GPU-based volume rendering engines scale poorly, and their interactivity degrades rapidly once sampling rate, image resolution and/or dataset-size increases. The result is low quality, heavily pixelated rendering of data during user interactivity. Conversely, the scalability of Fovia’s CPU-based XStream HDVR volume rendering engine is dramatically better than available GPU solutions. There is virtually no difference between interactive and final image quality, resulting in a high fidelity user experience.

 

Summary of Fovia’s CPU-based solution advantages:

 

 

In general, CPU-based engines excel at supporting today’s in-demand architecture—the thin-client and zero-download model. Fovia’s engine delivers without comprising performance or quality. The thin-client architecture of XStream HDVR utilizes server-side rendering, and the software can run anywhere, on iOS, Android, or HTML5, with virtually no dependency on the end-user display device. Conversely, server-side rendering using a GPU solution, although technically feasible, is difficult to implement efficiently. Specifically, GPU-based solutions fail in cloud-based deployments that require the system to scale effectively for users, dataset sizes and output resolutions. In order to scale, GPU systems need more real estate to accommodate additional expensive hardware, they consume more power and generate increased heat in the storage area—all contributing to higher costs. Fovia eliminates all of these issues and restrictions with its XStream HDVR software-based solution that can run anytime, anywhere.

 

Unlike GPU-based solutions, Fovia’s CPU-based solution lowers development costs and reduces implementation complexity. XStream HDVR requires no drivers or special hardware, and is easier to program and deploy. Extremely flexible, Fovia's solution runs on any system regardless of the video card. In addition, adoption is much easier for a developer because there are no OpenGL driver headaches.

 

Summary GPU-based solution drawbacks:

 

 

Some of the vital issues listed above are elaborated in the information provided below:

 

1. Inevitable driver or OS updates can cause compatibility issues with various applications, and can be a huge detriment to the day-to-day operation of a facility.

 

2. During development, testing must be done across a wide matrix of frequently changing configurations of graphics cards, video drivers, and operating systems. Fovia’s CPU-based solution presents a much quicker and easier development path, because testing is solely dependent upon the Intel or AMD chip.

 

3. High equipment costs are driven by the need for graphics cards and specialized hardware.

 

4. Memory issues: Most GPU-based solutions require 3-5 times more memory than the size of the dataset being rendered. CPU-based XStream HDVR needs only 40-45% memory overhead above the original dataset size. More important, XStream HDVR can render the original 2D slices of data or share memory with your application, resulting in an even bigger savings. Memory overhead reduction becomes more significant as dataset size increases, and/or when a single system loads multiple large studies.

 

More information on High Definition Volume Rendering® (XStream® HDVR®)

 

Fovia’s innovative, CPU-based, software-only High Definition Volume Rendering® solution overcomes the many limitations of currently available GPU-based imaging technologies, and enables local, enterprise-wide and cloud-based volumetric rendering available anytime, anywhere on any AMD or Intel chip.

 

XStream® HDVR® is an advanced visualization technology for analyzing extremely large three-dimensional data sets in real time, without any down-sampling. As data demands increase, Fovia supplies the software and processor markets with a tool that has no expensive proprietary hardware requirements.

 

Fovia’s solution successfully leverages the scalability and flexibility of Intel’s multi-core CPU Xeon processors, creating a combination that far surpasses the abilities of GPU-based imaging systems. XStream HDVR software uses volumetric ray casting, enabled by a rendering engine based on algorithms that minimize computational costs and maximize quality and performance. As a result, XStream HDVR allows for complete control over performance and quality by super-sampling to a sub-voxel level resulting in superior, high definition pixel output, surpassing any GPU-based solution on the market.

 

Prior to the introduction of Fovia’s XStream HDVR software, existing solutions were unable to handle the large volumes of data being generated, and exhibited many limitations, including: requiring expensive graphics cards or specialized hardware (ASICs); forcing significant trade-offs between image quality, speed and cost; inability to easily distribute images remotely (therefore requiring a dedicated, advanced visualization workstation); inflexibility and non-scalability; and a high obsolescence factor. Fovia’s XStream HDVR technology has successfully overcome these volumetric data challenges. Fovia’s software creates a paradigm shift in volumetric imaging by enabling the visualization and analysis of large volumes of data in high definition—both locally on workstations and remotely through server-based, thin-client technology—using only CPUs.

 

See what Intel and AMD have said about XStream HDVR:

 

“Truly amazing scalability – with every core you throw at the problem you are receiving almost pure linear scalability, truly a feat both for the platform and for the software.”—AMD.

 

“Helps scientists, doctors and geophysicists visualize and comprehend their complex datasets with greater interactivity than ever before.”—Intel

 

With respect to workflows and CPUs vs. GPUs, even “pro-GPU” publications such as VizWorld, proclaimed that High Definition Volume Rendering has advantages over GPU solutions:

 

“Once liberated from the restrictive instruction sets of most current GPU designs, they were able to create vastly more complex visualizations using adaptive ray-sampling, adaptive step sizes, and many other optimizations not easily implemented in GPU algorithms”

 

“Easily render semitransparent geometry or mix polygonal geometry with volumetric geometry with no performance penalty, unlike most GPU solutions, which require either depth-sorting or depth-peeling algorithms.”

 

“Fovia is very proud (and rightfully so) of the speed and detail of their solution. While GPU solutions can probably match them on speed, those solutions typically lack in detail. It’s easy to load up a dataset as a 3D Texture in video memory and render it in hardware. However, the result is not going to look as good as the Fovia solution. The argument is a combination of Speed and Detail. A GPU could probably win in raw speed, but lose in the resulting detail.”