Nvidia Talks DirectX Compute in Windows 7
Use your GPU for more than just Crysis.
The latest entry on the Microsoft Partner blog is from Nvidia and pushes the idea of the GPGPU. While it is written from an Nvidia product-focused, the same principals and advantages exist with GPUs from other vendors (such as your Radeon).
"With the introduction of Windows 7, the GPU and CPU will exist in a co-processing environment where each can handle the computing task they are best suited for," wrote Chris Daniel, product manager for software at Nvidia. "The CPU is exceptionally good at performing sequential calculations, I/O, and program flow, whereas the GPU is perfectly suited for performing massive parallel calculations."
Microsoft is doing its part by putting DirectX Compute in Windows 7, so that developers can make better use of the GPU for tasks other than just graphics acceleration. Having the GPU pitch in where possible will help take the load off of the CPU so that it can focus on other tasks. The ideal end result of this is that the PC should be more responsive thanks to efficient use of processing power.
Daniel gives an example of how a GPGPU could speed up a task: "With new software designed to take advantage of this capability you would be able to copy and transcode (convert a video from one format to another – a very computationally intensive task) a movie to your MTP supported portable media device up to 5 times faster when using the GPU as a co-processor with DX Compute, as compared to only doing the processing on the CPU."
Microsoft also natively supports GPU acceleration with a new Windows Media Player and Windows Media Center for H.264 video content, most of which is encoded in high-definition formats and typically more taxing on the CPU.
"Parallel programming is the next big thing for the world of computing – it has started already," said Daniel. "DirectX Compute will accelerate this discontinuity by enabling massive parallelism to the masses. What we are talking about is co-processing— essentially using the right tool for the job."
- Foxconn: Employee Suicide Had It Coming
- Resident Evil 5 PC Specs
- QOTD: How Much Did Your Computer(s) Cost?
- Is Stargate Worlds Doomed? MGM Says Maybe
- Microsoft Sales Decline for First Time Ever
- Intel Talks About Windows 7 Optimizations
- Apple Tablet Rumored for Q1 of 2010
- Foxconn CEO Accused of Tax Evasion
- MSFT Edits Laptop Hunter Ad at Apple's Request
- Bill Gates Tries Facebook, Quits Facebook
- Microsoft to Give EU Users a Browser Choice
- Corrected: Intel Halts Shipments of 34nm SSDs
- AMD Introduces $1,800 ATI FirePro V8750
- China Bans Online Gang Games
- This is What Firefox 4.0 Could Look Like
- Crytek: No Crysis 2 This Year
- Asus, Acer Push Back Plans for New Netbooks
- QOTD: How Would You Change Your Browser?
...and they lived happily ever after.
Lets hope OpenCL will be "brand" agnostic. I got here 1600 Cores that love to game...and folding.
Lets hope sometime cooler comes along.
Still waiting for the apps that "most of us use" (key words here) to have this facility.
The most impressive feature of 7 I've seen yet. I hope developers take advantage of this as much as possible.
The most impressive feature of 7 I've seen yet. I hope developers take advantage of this as much as possible.
I doubt it. Most games (ie GTAIV,etc) are ported from consoles so I doubt they would take the time to optimize it to this level.
I built my rig when core2 quad's got to a reasonable price and set up sli as well to future proof, thinking everything was moving to multithread. Now if general purpose software gets offloaded unto the gpu's, it would be a boost to the useful lifespan of my computer.
If at the very least it would be good to know that when I'm not gaming the graphics cards are doing more then Just heading up my home
I doubt it. Most games (ie GTAIV,etc) are ported from consoles so I doubt they would take the time to optimize it to this level.
what i got from this article is that they're talking about applications in general, and not specifically games.
for example, if a program has fancy bells and whistles when you click a menu item, or press a button, it can draw power from the video card which would be better suited to deal with this type of processing.
of course, what i've said is very simplified but it's what i got from the article.
... where is the news?
whats the difference between this and Nvida CUDA technology?
whats the difference between this and Nvida CUDA technology?
One will be a standard for ALL cards(well ones that support it) not just Nvidia ones, kind of like ATI's own implementation, will be nice for everything to work on all hardware.
I'm hoping for DirectX compute support in Handbrake, but I'm sure we're more likely to get OpenCL support first due to it being cross platform. Either option should do well assuming nvidia and ati have support in their drivers for both options. Other applications that will probably adopt gpu computing would be photoshop(more general than the current 3d acceleration) and a boatload of professional level video and CAD applications. For average users, this all means nothing (average people don't do anything computing intensive, but here's hoping for accelerated adobe flash).
Transcoding and encoding is all i need now i can put to use my powerhouse gpu more and spend less on powerhouse cpu's. Doing Ultra low High quality encoding takes a long time even on a good set up.
At least it would if it wasn't illegal to transcode a movie, unless we're talking home movies here.
sounds kind of like CUDA, but yes Nukemaster, for all GPU's. and if thats the case, maybe programmers will get around to utilizing GPU's for 3D Animation, which is something ive been waiting for for a long time now. fingers crossed
g-thor, You can also transcode all that PVR stuff you recorded
twisted politiks, About time huh. If i recall, ATI cards did all the transformers trailers in real time...
hopefully it will be an open standard and that windows takes advantage with out having developers rewrite the wheel.
Both ATI and NVidia have some kind of acceleration already. I think problem is, that it needs to be transparent for the end user.
This is Nvidia's race against larrabee. Cuda and Stream for that matter are still not implemented well enough to supplant CPU based encoders. Even if it's implemented into DX you would still need applications to apporpriately code for the the acceleration. Look how long it took to get DXVA acceleration to work for h.264, and it still not 100%.
Since Larrabee is going to be x86, leveraging it for multithreaded encoding will be much easier. Intel doesn't like the idea of Nvidia and ATI/AMD GPUs doing CPU like work. I have a feeling larabee will crush the other GPUs at parallel computing.
And I'm still waiting for a proper, working, 100% FREE gpu based video transcoder.
* ATi's free video encoder's quality is just really not up-to-par.
Does this mean that if your going SLI or Crossfire then that extra GPU will also be able to take more load off the CPU? Cause that would make sli and crossfire even MORE useful then it already is.
Oh great so i can transcode my movies a bit faster for my iPod. What else? I'm still waiting for a real useful software that can use my GPU.
Microsoft also natively supports GPU acceleration with a new Windows Media Player and Windows Media Center for H.264 video content, most of which is encoded in high-definition formats and typically more taxing on the CPU.
Anyone know if this applies to the use of media player classic on win 7 as well? (part of kl mega codec pack ; codecguide.com)
The most impressive feature of 7 I've seen yet. I hope developers take advantage of this as much as possible.
lol wut, we makes the apps for you and your XP baseline OS. Maybe in 3 years we think about win7.
Oh great so i can transcode my movies a bit faster for my iPod. What else? I'm still waiting for a real useful software that can use my GPU.
you know a PC is not only used by ppl who put music to their ipods ... some ppl make video editing , photoediting , 3D animation and so on ... so you can stick your ipod somewhere for that matter is not about you here.
the vendor would be ATI, the product is Radeon.
Will it mean that in the future netbooks will come with eg: a single core CPU processor, and a (crappy) integrated graphics card,and that in windows we'll see over 16 cores (threads)?
I'd love to see if the OS can boot and do background tasks faster thanks to the smaller threads that the GPU can process.
I'd also love to see if programs like MS office, Adobe, etc.., could see a benefit in booting speed thanks to this technology, and that it won't be limited to video encoding, or x@home clients.