藏在您繪圖工作站中看不見的GPU運算 (英文)
The value in a professional graphics workstation is higher than you might realize.
When most of us look for the “right” workstation for our CAD and design needs, we shop for excellent performance, the right level of storage capacity, stability and reliability, and software certifications and support. Being smart shoppers, we judge the value of each system and the value of the key technology inside.
I suspect that one place many of us economize to our own detriment is by selecting a less expensive consumer graphics solution over a comparable professional solution. Being realistic, there are times when you can do that, but besides the issue of reliability, testing, and support for your applications – and the disastrous impact to your projects, profits, and productivity when something goes wrong with a workstation – many professionals will be bypassing significant performance gains as well.
In todays professional applications, the computing environment is moving to a heterogenous computing system. Application developers are creating products that use both GPUs and CPUs for computing. This is a recognition that a GPU is a super-computer-on-a-chip inside a professional workstation. A couple of examples in the CAD and design domain include CATIA V6 from Dassault Systèmes and THEIA RT from optical simulation specialists, Optis.
In the first case, CATIA Shape has integrated iRay from Mental Images – the benefit for industrial designers around the world is simple – interactive ray-traced images of your design. Critical for design productivity, the designer makes a change and interactively visualizes the design with ray-traced quality. How this is possible is due simply to the intelligence of the Mental Images developers. Ray-tracing, although a graphics-specific algorithm for rendering physically accurate images, is a highly intensive computational problem. It has nothing at all to do with the more common, albeit amazing, images generated in real-time using the hardware-implemented graphics pipeline of the GPU. Ray-tracing needs a computing processor, not a GPU, and until recently, ray-tracing was anything but interactive.
So inside CATIA Shape, the Mental Images developers leverage the extreme computing horse-power of the Quadro GPU (or GPUs) in the workstation – a massively parallel-processing GPU – to do the heavy-lifting for the ray-tracing algorithm.
A French specialist in optical simulation, Optis, has been going far beyond the realism of ray-tracing for more than 20 years. Optis is not specialized in creating amazingly realistic looking images, they are specialized in creating amazingly accurate images. The difference is that Optis solutions are optical simulations of the light and its interaction with the entire environment – materials, lighting, natural lighting, reflections, etc. In brief, Optis software is highly regarded by its customers, for example in the automotive, aerospace, and luxury goods industries, where the simulation of light and lighting is critical to the product.
Like so many hard problems in simulation, real-time and interactive were adjectives rarely found – even in the most simple scenarios. When Optis partnered with the GPU experts at AMD, the collaboration resulted in a product Optis calls THEIA RT, permitting exactly that which had not been possible before – interactive visual simulation of complex industrial designs. As in the previous solution, Optis developers leveraged the massively parallel-processing capabilities of the GPU. In this case, by teaming with AMD and using the OpenCL environment to access the GPU's power.
And AMD themselves are uniquely positioned and, I suspect, highly motivated to push this platform to new capabilities. Just a short time ago, I spoke with AMD about the development of their Fusion heterogenous computing environment. As CAD professionals, we don't need to care about this, but the people developing our applications do care a great deal about this platform. We all know that so many problems today take huge computational resources – FEA, CFD, almost any area of engineering, design, or architectural simulation – and when the developers of your applications have an easy way to access the super-computer hidden inside your workstation, they are going to make it possible for you to do thing tomorrow which you don't even consider today. All of which will help you and your company to be more productive, more creative, and more competitive.
So remember, when you are looking at your new workstation requirements, do some checking to see if you can unlock a graphics super-computer. Because in the examples above, and more and more new examples, the results are stunning. And if you are not looking, then the fact that your workstation leverages professional GPU-computing is invisible to you.
註釋
GPU computing blog comment
2011年 十一月 23日(星期三) 3:27:53 pm
TL