14 DAY TRIAL // Ever since NVIDIA first announced its next-generation GPU architecture, codenamed Fermi, there has been a series of speculations made on the likely release date for Fermi-based graphics cards. Reportedly planned for release before the end of 2009, Fermi-based graphics cards are now expected to become available later this year. On that note, the Santa Clara, California-based graphics chip maker recently provided several journalists with a fresh set of details on what the new Fermi architecture is going to deliver.
Although previous rumors indicated that the graphics chip maker was planning the launch of its first Fermi-based products for late 2009, the most recent reports suggest that the company will be releasing the new GPU sometime in March, with availability slated for the end of the month. In addition, we now have some more details regarding the architecture, which the company announced during the Consumer Electronics Show, but only to a select number of journalists, which have now reported on their findings.
Our collegues from AnandTech, Tom's Hardware, Hot Hardware and PC Perspective, have all published an article featuring the various bits and pieces that NVIDIA provided them with, during these media sessions. According to them, we should expect a 40nm GPU, with support for DirectX 11 and a number of new features. Unfortunately, things like clock speeds, die size, power consumption, performance and pricing details have all been left out, as NVIDIA is looking to generate the right amount of hype, before the cards are on the market.
According to available details, the new GPU will feature 512 CUDA cores, up from the 240 available in the GT200. Around 3 billion transistors will be enabled on the new GPU, compared to the 1.4 billion that were available in the GT200. The GF100 also features 16 geometry units, 4 raster units, 64 texture units, 48 ROPs and a 385-bit GDDR5 memory interface. The GPU will bring more new stuff to the table, but as far as performance goes, when compared to AMD's DirectX 11-based GPUs, we'll have to wait until the formal release.