Meanwhile its chief competitor NVIDIA has been struggling to prepare its first DX11 part. The company previewed its Fermi architecture in October, but details have been few as the manufacturing problems and yield issues at the Taiwan Semiconductor Manufacturing Company still plague NVIDIA's 40nm process.
The first Graphics Fermi 100 (GF100) chips started production in January, but it wasn't until February that we learned that they would be sold under the GeForce GTX 480 and GeForce GTX 470 monikers. Some specifications finally showed up last week, and today we have the final details confirmed.
NVIDIA GeForce GTX 480
The die size of the GF100 chip is massive, measuring 23mm x 23mm for a total of 529 square millimeters. By comparison, the Cypress chips used in the Radeon 5800 and 5900 series cards come in at a more moderate 334mm², making the GF100 almost 60% larger.
However, performance is much stronger, with the GTX 480 and GTX 470 beating the Radeon HD 5870 and 5850 respectively, according to internal benchmarks shown to us by NVIDIA. The tessellation engine in particular is exceptional strong, and the near-linear scaling bodes well for those enthusiasts considering a GF100 SLI setup.
NVIDIA GeForce GTX 470
The powerhouse chip consumes an extraordinary amount of power. NVIDIA recommends the use of a 600W power supply with the GTX 480, but that is just a minimum. Enthusiasts who want to prepare for a possible SLI setup should have a 1000W PSU at the minimum, with a 1200W PSU not out of the question.
The most disappointing news is that after all of this waiting, NVIDIA has told DailyTech that readers should not expect volume availability until the week of April 12. There are supposed to be "tens of thousands" of GF100 cards at launch, but production is slow since all of the cards that will be sold during the launch are reference boards built by NVIDIA.
Somewhat surprising is the lack of support for DisplayPort, the next-generation computer display standard set to replace VGA and DVI for desktops and laptops. NVIDIA states that the GF100 supports DisplayPort, but it will be up to its board partners to support it in their own designs.
Instead, the reference design has two Dual-link DVI ports and an almost useless mini-HDMI port. NVIDIA is keen to tout its three interfaces and support for its 3D Vision Surround technology, but users who wish to use more than two monitors at the same time will be required to use a second card. This contrasts strongly with ATI's Eyefinity technology present on all 5000 series cards, which support the use of three monitors at the same time using a single video card.