Teardown of the Nvidia GTX 480: The hottest card ever made by Nvidia
rkliman
Entry ID #: 3845
Created: Wed, Jan 11, 2017 6:21 PM
Nvidia’s high-end Geforce GTX 480 graphics processing unit was designed for performance. Released mid-2010, a painfully long time after ATI’s insanely powerful 5000 series flagship lineup, this card could not afford to compromise – and would become infamous in the PC gaming, enthusiast, and video-editing communities for its astoundingly high heat and power draw. The members of 12A are such enthusiasts, having built PCs for gaming and media editing, and had heard much about this powerful card despite entering the enthusiast landscape several years later. When we finally got our hands on a “Grillforce GTX Fry-80,” we were excited for the opportunity to discover the components that make up a GPU. We were also interested to see what exactly made this powerful card run hotter than even double-processor cards of the era (for example, ATI’s 4870x2). The first thing we noticed was the surprisingly massive GF100-375-A3 chip – the 480’s Fermi architecture graphics processor. Running at 700 MHz, it is the main processor in the card. Surrounding it are 12 GDDR5 128 MB video memory modules, responsible for rapidly delivering data to the processor for computing. To the right of the main chip are cylindrical capacitors, voltage regulators, and power delivery modules, responsible for regulating power delivery to the memory and processor. To the left are chips for encoding display for output, connecting to the card’s DVI and mini HDMi display output ports. Below the main chip is the PCIe 2.0 16x bus, which communicates with the rest of the system through the motherboard’s PCIe slot. On the top-right of the card are the six and eight pin power connectors which when under maximum load supply around 475 Watts to the system. Above the circuitry is an aluminum heatsink, with thermal pads placed above the power delivery modules and memory and a fan that blows air through the processor’s heatsink out the back of the card. Finally, a rear I/O shield and plastic shroud cover the rest of the card. Though nothing was marked Texas Instruments, due to the vital innovations TI is responsible for in this field, many components certainly contain vital patents and designs created and licensed by TI. The heatsink is not small, but was inadequate given the card’s power draw - ATI’s Radeon HD 5970 consumed 50 fewer Watts despite higher performance. Our biggest surprise was the size of the GF100 chip, and upon further research we discovered this size was responsible for the card’s high temperatures – to compete with ATI’s tiny microprocessors, Nvidia’s engineers used a larger chip to fit the same amount of processing at a high cost of power and heat. Beyond giving us a much clearer picture of the many different chips located on a sophisticated piece of technology, our teardown ended up teaching us a great deal about the difficulties and compromises in designing electronics. We saw how the demand for performance led to major design compromises, and how sometimes these compromises are unavoidable given market pressures.
