Windows 7 compatibility test with the CUDA architecture

Graphics giant NVIDIA's GPU-generic computing technology Cuda believes everyone knows, but as Windows 7 becomes a popular concern, some people question whether Windows 7 supports Cuda and can support Cuda well. The reason is straightforward--because Windows 7 supports Microsoft's own direct compute!

But is that really the case? Today, the author once again for the concept of cuda, architecture and so on, and respectively in Windows Vista and Windows 7 operating systems on the performance of the Cuda test, to achieve the effect of order.

Before conducting a formal comparison, we must first understand what Cuda is, and the author also finds that it is precisely because many people do not have a truly accurate and objective understanding of cuda that leads to misunderstanding of CUDA and direct compute. Let's take a look at how Nvidia officially explains the concept of Cuda:

Cuda is a general-purpose parallel computing architecture introduced by NVIDIA, which enables the GPU to solve complex computational problems. It contains the CUDA instruction set architecture (ISA) and the parallel computing engine within the GPU. Developers can now write programs for the CUDA architecture using the C language, which is the most widely used high-level programming language. The program is written so that it can run at ultra-high performance on processors that support Cuda. Other languages will also be supported in the future, including Fortran and C + +.

The C language compiler version is updated quickly in the Cuda development package provided by NVIDIA

As you can see, Cuda is a universal parallel computing architecture in real sense, but it also contains the CUDA instruction set architecture and the parallel computing engine inside the GPU. As in the case of CPUs, the X86 architecture also contains the hardware architecture of ISA and execution directives. Various applications are developed based on this architecture and run on this. It can be said that the CUDA architecture of the GPU is a graphics rendering architecture and parallel computing architecture!

Cuda is meant to be a framework, but many people think that Cuda is a language, or a development platform, which may be related to Nvidia's initial propaganda and promotion strategy. At the beginning of the Cuda architecture (GeForce 8), there is no development platform and programming language specifically for GPU parallel computing, so NVIDIA must provide a simple, easy-to-use and complete software development solution for programmers, Thus Nvidia's release of the Cuda package includes the C language compiler for the GPU, the error-correcting device/tooling, the dedicated driver, and the standard function library.

In addition to supporting your own C-language compiler, more languages and APIs can be run in the CUDA architecture

The diagram above reflects the relationship between Cuda and the application Interface (API) and various language compilers, in which the DX11 computation, direct Compute, shows that there is no conflict between the Cuda schema and direct Compute. Including Cuda's own compiler used in the C language extension, OpenCL application interface, Fortran and even C + + can run on the Cuda architecture, the future CUDA will also support more languages. In the whole industry under the common impetus, GPU computing is a promising future!

In fact, in a Microsoft Direct compute PowerPoint, we can also see that direct compute is completely based on a GPU that conforms to the unified rendering architecture. In fact, since DirectX 10, Nvidia and AMD have watered down the concept of pipelines and vertices to allow the flow processor to perform more non-3D applications, and now Microsoft's direct Compute is undoubtedly a big boost for GPU general-purpose computing applications.

To clarify the relationship between CUDA and direct compute, let's go into the beta phase to see if direct compute in Windows 7 has any effect on Nvidia Cuda, and once again verify direct Whether there is conflict and competition between compute and CUDA.

This time we use the encapsulation interface for LGA1156 i7 with just released P55 motherboard for testing, graphics card using Inno3D launched ichill-gtx260+ Taurus Edition, and in the not supported direct compute windows The Vista operating system is contrasted with the Windows 7 operating system that supports direct compute. Where Windows Vista is 64bit,windows 7 is 32Bit.

Test procedures We selected three most representative Cuda software, respectively, for domestic transcoding software MediaCoder, CyberLink launched the video editing software Power Director (director) and everyone very familiar with the transcoding software Badaboom, and test them in the same way under Windows Vista and Windows 7, respectively.

Badaboom is the GeForce GPU video transcoding tool produced by NVIDIA United Elemetal, which is the only video transcoding software that supports cuda acceleration in the early stages of the CUDA standard, through which the software can be used directly to accelerate the use of Cuda to the iphone, ITouch, PSP, Apple TV and many other devices to convert video. Currently, the latest version of the software is 1.2.1.

This test was used in the film "X-Men 3" of a 11-minute trailer, the original resolution of 1920x1080, in this test we will convert it to 720P MP4 format, the following are in Windows Vista and Windows 7 performance screenshots:

Windows Vista: Conversion time to 5 minutes 23 seconds