GPGPU, DX11 Direct Compute Benchmarking

Now here are benchmarks for DX11 Direct Compute, or Microsoft's DX version of OpenCL, GPGPU.

Download: http://www.friendsea.com/ComputeMark/



G73 Stock Score: 888
G73 800/1100 Score: 933, 5763 180s Burn Test.

 

pretty good.

when is the mobility 5970 coming out for laptops?

also, still thinking of getting the g73 and was wondering how dx11 plays vs dx9 with the tesslation feature. im a geek when it comes to new graphics or technology.

heck i bought the g71 just to see crysis on veryhigh..pretty sweet if you ask me.

 

The 5970 is a dual-gpu 5870, so probably never. The only thing that might happen is that the mobility 5870's (really a 5770 desktop chip) could be configured in crossfire.

 

Maybe Asus will make another W90?

 

but what does this mean for the computer illiterate wanting to learn? (me) lol

 

The GTX480 on par with 5770?

 

No they won't. Asus have said they will not be going into ultra high end laptops anymore, G73 is as good as it gets.

Kal, GPU have numerous shader cores. Each of these cores are running at 700 mhz for my HD5870M for example. Because there are 800 shader cores, this mean for certain applications, my HD5870 can process some information much faster and better than the i7 Intel.

So not just for gaming, there are many applications where you can now offload the processing onto the GPU. This is highlighted by CUDA from Nvidia and Stream from ATi, but both are really just shader cores.

CUDA, Stream are proprietary to each company, the open source is called OpenCL. From Microsoft, DirectX 11 allows some processes that were traditionally only on CPU to be offloaded onto GPU. This is called Direct Compute, this is what excites a lot of developers. This means we can have for example Physics in games now processed on these shader cores.

GPGPU is what they are calling it. The computational power is called a flop. The HD5870M can has 1 teraflops of computational power, desktop HD5870 has 2 teraflops. Floating Point Operations Per Second.

 

The months of delays and preposterous size and power requirements on Fermi just keep paying off for <s>nVidia</s> AMD.

 

Woopie.. Ferni had it's but kicked by 5870 and 5970 again! What does this mean? It mean's " its is not only the begining of the end but its the end of the begining". Woopie, another nail in NVDIA's coffin..

 

"Woopie.. Ferni had it's but kicked by 5870 and 5970 again! What does this mean? It mean's " its is not only the begining of the end but its the end of the begining". Woopie, another nail in NVDIA's coffin.."

einsteins exact words

 

Indeed.

It is odd that the advertised feature of Fermi's (extreme compute performance) is not as good as the competitor. ATI has a nice series of cards this time. But I think that Nvidia has not optimized for Direct Compute as they promote CUDA (or even native C) for computing tasks. Considering many games are written for Windows and use DirectX, you think it would be advantageous to optimize Direct Compute performance (even though it isn't common yet). $0.02

 

This only means that ATI's implementation of DirectCompute/OpenCL is more powerful than nVidia's implementation of DirectCompute/OpenCL. That graph says nothing about the graphics performance of GPUs.

When nVidia is holding a majority of the GPGPU computing market with its CUDA platform.