Future high-performance GPU nVidia has many names. He is known as GF100, GT300 (on the basis of the present nomenclature), in addition, recently it became known that nVidia gave it to chip cards internal code name "Fermi", after the Italian physicist Enrico Fermi, one of the founders of nuclear physics. Given this, it is not surprising that the board itself, according to available data, will be called the "reactor" (reactor).
To date, we know the following parameters GT300/Fermi (of course, unofficial data, and the final specifications may differ slightly from those listed):
• The number of transistors than 3 billion.
• The chip is made on 40 nm process technology at the factory TSMC.
• 512 universal processors, nVidia might be called "CUDA cores", ie "core CUDA".
• For 32 cores in a cluster.
• 384-bit interface GDDR5 memory.
• The capacity of the L1 cache is 1MB, the capacity of the L2 cache is 768Kb.
• All the memory capacity of up to 6GB, the video for the ordinary user supposed to embed 1.5Gb of memory.
• Built-in support for C (CUDA), C + +, Fortran, support DirectCompute 11, DirectX 11, OpenGL 3.1 and OpenCL 1.1.
In addition, the source said that today will be a press conference nVidia, which will first be presented to this mysterious graphics chip.



Reply With Quote
Copyright Techfuels
Bookmarks