Nvidia CUDA and PhysX

[Taarank]

Let me tell you that I am not having enough knowledge about the graphic card. What is a graphics card can actually do anything like that? Over the years, the GPU (Graphics Processing Unit), but made many steps forward. While all units at the beginning of the video card still hardwired and so the tasks were already set (called a fixed-function pipeline), there was the introduction of the Direct3D 8 API and the associated hardware for the first time a programmable unit. Now I am here to know more about the Nvidia CUDA and PhysX. I think that you guys will definitely help me (its my assignment). I am expecting some information sooner from your side.

[Aamin]

With the Direct3D 9 hardware and Shader Model 2 and its successor SM3 opportunities were increasing and the success of the Direct3D 10 graphics cards are no longer a pure GPU more. For with structured parallel programs for themselves a high-end multi-core CPU has as its difficulties, while a GPU can do the tasks in your sleep (nevertheless can not replace a GPU, a CPU, and will probably never do). Both ATI and Nvidia have recognized the signs of the times and offer the software developer recently of techniques with which one programs can leave off the 3D graphics run on the graphics card.

[Deklan]

The corresponding software developers in their work, however, was always aware that they are programmed on a 3D accelerator, because you had to use a language understood by the hardware: Direct3D or OpenGL as an alternative. It was therefore more or less programmed "graphic" can and have, accordingly, the necessary knowledge, even though the programs have nothing to look at the hat.

[Imdra]

Meanwhile, both ATI and Nvidia one step further and offer especially for GPU computing (computing on the GPU without graphics) is a high-level language, with which it is easier to develop GPU programs without having to rely on Direct3D or OpenGL need. This is the first important step was done to make GPU computing more attractive, even if the implementations of ATI and Nvidia (different again). Further steps seem so inevitable. With the Direct3D 11 API seems to already have a solution in sight. The Microsoft API should support a so-called "Compute Shader", that is a new and particularly uniform standard for general calculations, making the GPU computing should celebrate the breakthrough - as well as enforce the 3D graphics with only a single API could. Until then, you have to take but still with the (certainly not bad) competing solutions, the manufacturer preferences.

[Wajeeh]

Nvidia's self-developed high-level answers to the name CUDA (Compute Unified Device Architecture) and is the traditional C programming would be very similar. With CUDA programmed programs can benefit from an important advantage: Not only do they work on all CUDA-enabled GPUs (from the GeForce 8000 series are generally all NVIDIA GPUs, CUDA-capable, though some still the driver is missing), it addition, the programs require no adjustments for future graphics cards. The software will run properly on, for example, as a GT300.

[Rixwel]

In addition, there will be many other mainstream tools a CUDA version, with respect are not too many details are known. It can, however, is an open secret that Adobe is working on a Photoshop version of the GPU to display the massive use of filters and effects on the graphics card charge. Also there will be an audio tool with which you can convert audio files from one codec to another convert. Further details are still open.

[Zeverto]

Technologically after the graphic quality of the physics performance is one of the most interesting factors in play. In order to implement physics in games, there are two possibilities. Either you write your own physics library, which takes time and money in claims, but also brings benefits, or it relies on a commercially available physics environment that has been tested many times, and above all improved again and again. To select the moment there are two direct competitors: Havok from Intel and Nvidia PhysX.

[Broot]

To select the moment there are two direct competitors: Havok from Intel and Nvidia PhysX.

At this point we want to address more specifically PhysX, as the PhysX libraries can be calculated on a GPU. We will not rule out that Havok will be accelerated in future ATI cards, but you can currently only speculate in this regard. PhysX was originally developed by Ageia, who designed not only the pure CPU-physics environment, but also developed a hardware accelerator, on the parallel processors could accelerate the presentation of physics in games. And this in a manner which enabled the developer physics effects on a CPU that could not be played back smoothly.

[StephanieRice]

While the native PhysX hardware was sown in the PCs of the customer only rare, so the developers can now draw on a much larger user base, since PhysX works on all CUDA-capable graphics cards (the GeForce 8000 series). The software started with next to the second CPU test in 3DMark Vantage once the first-person shooter Unreal Tournament 3 Condition was and is the PhysX mod pack , based on the Ageia hardware additional effects in three selected levels already facilitated, based on a CPU that could not be displayed smoothly. The new maps HeatRay DM, CTF and CTF-Tornado-Lighthouse offers a variety of elements that are based strongly on physics effects. These include collapsing walls, a Tornado responded to the local area, breaking glass and destructible environments.

[Pandya]

Compatible including the free game, which was developed specifically for PhysX hardware and produced without the only unplayable frame rates. Flying around elements that interact with the environment, as well as destructible towels are only two reasons why a main processor is overwhelmed. In addition, there is the game "Ghost Recon Advanced Warfighter 2", which uses the PhysX hardware, but only conditionally. In particular, with possible new particle effects can benefit the gaming experience yet. As early alpha versions controls an Nvidia tech demo of the game "Nurien" when, in which there is to inspect a small fashion show. The movements of the models clothes flapping all about calculated physically correct to the actress. The second alpha version is from a multiplayer first-person shooter called Metal Knight Zero. Due to the unfinished status of the program prepared us, however, some difficulties. Finally there will be self-programmed by Nvidia tech demo Nvidia Particle fluid that simulates a water course.

[Gadin]

And what happened with PhysX when two Nvidia graphics cards are installed in an SLI system? Nvidia PhysX implementation offers two 3D accelerators in two different modes: the "SLI" - and the "multi GPU" mode. In SLI mode, both cards render as usual at two different pictures in AFR mode and calculate also the physics. Currently this works only in a few applications, so Nvidia recommends a general principle to enable PhysX in games the multi-GPU mode. This takes care of a graphics card to render the 3D graphics, while the other is solely responsible for PhysX. In the second graphics card may well belong to another generation (instead of only GeForce 8 GeForce GTX 200), the CUDA capability must be given. One can, for example, a GeForce GTX 280 (graphics) with a GeForce 9500 GT (PhysX) combine, with an SLI motherboard is not a requirement in this case. The mixing operation of a Radeon with a GeForce card is not possible, because Windows Vista does not accept two different video card drivers.

[Zeppelin]

The PhysX control panel you can now choose to be calculated on what PhysX card. Currently there are on Windows Vista but still a disturbing limitation that Nvidia would diminish in the future with a workaround (Windows XP is due to the different driver models not affected). So it is necessary under Windows Vista, that either a second monitor or alternatively be the same monitor with a different cable to the GeForce PhysX card must be connected, even though they charged no picture. In addition, we must extend the Windows desktop on the second and the individual monitor. Both are currently still at a very early stage and should be improved with the next driver versions.

[Langward]

Without doubt, for that matter GPU Computing NVIDIA currently has the nose on the home user sector forward and can offer a greater variety of programs than ATi. But the Canadians are not idle and support as their counterparts in the professional segment, various programs that run faster with GPU. As for the mainstream market, but it does not look so rosy. Currently there are up to Folding @ Home is not a program that can benefit from the blatantly GPU computing capabilities of ATI graphics cards. Although they even wrote a video converter to convert video codecs, but this is only compatible with the Radeon X1000-series.

[Daniela007]

How will it just go with ATI for the traditional customers further, is currently still open. One can assume that there will be for ATi products customized programs and has managed to date only in the Folding @ Home ATi mainstream list (Adobe will follow soon). In addition, we seek to, to run physics and AI calculations on the graphics card. So it is possible that ATi GPUs accelerate in the future using the Havok Physics Interface effects on the GPU. In addition, we have programmed for the RV770 chip own tech demo that implements the complete artificial intelligence of some 10,000 units on the 3D accelerator. Just like Nvidia and ATI has developed its own "GPU computing language" with which one can carry out without the knowledge of 3D programming calculations on the GPU. While the Canadians originally with CTM (Close to Metal), only a hardware-programming environment (low-level programming offered), CTM now has two "layers" split: HAL (Hardware Abstract Layer) and CAL (Compute Abstract Layer).