Nvidia

The next generation CUDA architecture, code named “Fermi”, is the most advanced GPU computing architecture ever built. With over three billion transistors and featuring up to 512 CUDA cores, Fermi delivers supercomputing features and performance at 1/10th the cost and 1/20th the power of traditional CPU-only servers. At least that’s what Nvidia wants you to believe. True to form, the hype machine is in full blast as the countdown to Nvidia’s new graphics chipset is winding down. Which is understandable. The company’s primary competitor in the performance graphics card arena, ATI, has enjoyed a comfortable lead in the wallets of consumers for some time now. ATI’s last generation of chipsets, based on the 4000-series, have been lauded for their cost-to-performance ratio. So much so that Nvidia has had to slash its prices for their competing 200-series cards to even merit a second look from buyers. In an article by Charlie Demerjian (2009) on Semiaccurate.com titled “Nvidia kills GTX285, GTX275, GTX260, abandons the mid and high end market”, he outlines the current financial woes of Nvidia, as the company can no longer produce three of its product lines because they can no longer turn a profit, not with ATI cards of similar performance at a lower cost. These are cards which are considered high-end, but the future for their midrange cards are also under dark clouds.
As if being clobbered in the value arena by ATI wasn’t enough, last September ATI launched their new 5000-series of graphics cards with but one intent. Beat Nvidia in the raw performance categories too. Mission accomplished. Nvidia’s current king of the hill, the GTX 295, must now bow down to ATI’s big dog, the Radeon 5970. The value-for-money crown, and now, the “fastest single card on Earth” titles are now comfortably wrapped around ATI’s red waist. Nvidia’s engineers must be feeling the pressure. And the PC enthusiast crowd has been waiting with bated breath how Nvidia will answer.
And the answer is Fermi. Judging by the official Nvidia presentation of the Fermi architecture, which can be perused at http://www.hardocp.com/article/2010/01/17/nvidias_ fermi_gf100_facts_opinions/, the chipset does look impressive on paper. And if the screenshots of actual renders, as shown on http://www.tomshardware.com/news/Nvidia-Fermi-CUDA-GPU-GeForce,8968.html prove to be accurate then there might be hope for Nvidia yet. The architecture is chock full of improvements. Suffice to say everything that could be done to make the old generation faster and more advanced Nvidia’s engineers did, and then some. But the trip from the engineering department to sales is not as smooth as it should be. Fudzilla.com (Abazovic, 2009) reported way back in November that Fermi will not debut for that month or December 2009 as was originally planned by Nvidia. The article projected January 2010 as a more realistic target. It is now February. In an article on Digitimes.com (Chen, 2009) sources state that the official launch will be in March this year. Nvidia’s woes continue.
Meanwhile, ATI continues to churn out cards unopposed. They have just launched the Radeon HD 5670 and 5570 for the mainstream market. No card from Nvidia can stand beside them from a technology/feature set standpoint.
All of this competition spells well for the consumer as it drives down prices. Here’s hoping Nvidia recovers and gets back into the fray. Nobody wants to see a one-sided fight.