NVIDIA’s new ‘Fermi’ graphics platform may have suffered a succession of annoying delays, but newly released specifications for the GPU specialist’s latest architecture suggest the technology may be well worth the wait.
After using the recent Consumer Electronics Show (CES) in Las Vegas as a springboard to highlight the Fermi GF100 graphics processor, which has been developed through the 40nm manufacturing process, NVIDIA is now looking to shine the media spotlight just a little brighter in terms of core details.
Specifically, the next-gen graphics processing unit (GPU) will carry in excess of three billion transistors and 512 CUDA cores, which NVIDIA claims will significantly boost the GPU’s capabilities across various technical aspects such as ray tracing, physics, finite element analysis, high-precision scientific computing, sparse linear algebra, and search algorithms.
With NVIDIA’s new graphics beast set to offer plenty of muscle thanks to 48 ROPs and a 384-bit memory interface running on GDDR5, other contributing elements likely to test the temptation glands of gamers everywhere include 16 Shader Multiprocessing cores (consisting of 32 CUDA cores each), while each of those 16 cores has 16/48KB of dedicated L1 cache, four texture units, and a Polymorph Engine.
While related price points have not yet been divulged by Santa Clara-based NVIDIA (don’t expect it cheap), the company’s proposed release schedule indicates that the first boards equipped with the GF100 chip will begin appearing towards the close of February, with more widespread availability in March.