ATI has recently released its fastest graphics card, which is RADEON HD 5970.
It seems that with its latest released RADEON HD 5970; ATI is putting an extreme pressure on NVIDIA.
Now Nvidia has finally decided to come in action with the overwhelming approach of ATI.
The picture of NV 100 GPU has been recently introduced on BSN site showing the desktop version GF 100.
While many fans and readers believe that the picture introduced by the BSN site is fake, executives of Nvidia have proved its authenticity in Guangzhou (China) at the opening ceremony of experience store of Nvidia.
Drew Henry the senior director of platform products shows the same picture of GF 100 GPU with his cell phone running the DirectX11.
The next generation CUDA technology has been given a name FERMI, is the most advance GPU architecture that has been ever built.
Fermi features many important innovations:
1. 512 CUDA cores
Fermi cuda cores are optimized for both exceptional performance and accuracy. Double precision performance up to 8 times faster than GPU architecture.
2. ECC support (Error Checking and Correction)
Fermi is the first GPU architecture to support error checking and correction (ECC), this technology detects and corrects soft errors before they affect the system. Registered files, Cache and DRAM all are ECC protected making the most reliable GPU.
3. Giga thread engine
Giga thread engine supports high kernel execution which allows different program function to execute in parallel which improves speed and efficiency for all application. Fermi also includes dedicated bidirectional data transfer engine that intelligently manages 10s' and 1000s' of threads.
4. Parallel data cache technology
Fermi is the first GPU architecture to support cache hierarchy.
L1 cache is of 64KB (kilobytes) which improves bandwidth and reduces latency for a wide range of GPU computing applications.
L2 cache is of 768KB (kilobytes) which improves coherent data sharing across the GPU.
It contains 6 pin and 8 pin powers that means 300W will be consumed by GPU. The clock and performance are said and assumed to be very high that means you will have healthy over clocking headroom.
It is expected to be equipped with 3.0 billion transistors
Fermi makes CPU and GPU the complete spectrum of computer applications.
Besides gaming it will also make parallel programming easier and accelerates the performance of many applications, including high scientific computing, sorting, search algorithm, acceleration in ray tracing etc.
Fermi is a 49nm GPU just as rv870 but as an advantage it has 40% higher transistor count.
The price will be a valid and important concern. But at this point it is not safe to make any more assumptions. So let's wait for the real thing before we begin stuffing forums with rumors.
No comments:
Post a Comment
We shall share what we think is shareable!!