GRAPHICS CHIP OUTFIT Nvidia has had a real devil of a time getting its next generation of GPUs codenamed Fermi out of the fab and shipping in graphics cards.
Envisioned by the Green Goblin's CEO Jen-Hsun Huang as the realisation of his soaring ambition to take over high performance number crunching in GPGPUs, as well as deliver a knockout blow to Nvidia's graphics rival AMD-ATI, the Fermi chip was designed to be both massive at more than 500mm2 and internally agile to do both HPC and GPU work.
The cost of Huang's ambition became obvious as soon as the prototype wafers were produced, as initial chip yields were reportedly abysmal, in the low single-digit percentages.
Wafer yields are said to have improved after a time-consuming re-spin, but they're still rumoured to be less than fully satisfactory.
The chip was also reported to be very power-hungry and extremely hot, pushing or even exceeding the PCI power limit of 300W.
Graphics cards based on Nvidia's upcoming line of Fermi GPUs are expected to hit the shelves some time in the next few months, certainly within the first half of the year.
While some of the company's critics are predicting that Nvidia's line of Fermi GPU based cards will fail to overwhelm the competition, run too hot for sane deployment and be outrageously expensive, we're keeping an open mind.
But what do you think? We've got an INQUIRER readers' poll up about Nvidia's Fermi GPU prospects, and that's here. µ