When Advanced Micro Devices (NASDAQ:AMD) announced its next-generation R9-290X graphics chip, the consensus was that it was neck and neck with NVIDIA's (NASDAQ: NVDA) $650 GeForce GTX 780 and could even hold up well against the green team's $1,000 GeForce GTX Titan. Now, while the extremely high-end portion of the discrete GPU market isn't the big volume/revenue driver, having the highest performing, top-of-the-line "halo" card is important in this market segment from a marketing/sentiment standpoint. While AMD seemed to have the spotlight with its R9-290X, NVIDIA's GTX 780 Ti decisively takes back the crown.
Interestingly enough, AMD was first out of the gate in early 2012 with a 28-nanometer GPU. In this market, being able to cram as many transistors into a given area and power budget is one of the key performance drivers, and being the first company to process new geometries often leads to very tangible competitive advantages. With its Radeon 7000 series cards, AMD had a head start -- if gamers wanted a high-end graphics card between January and March of 2012, it was AMD or bust. These things were flying off the shelves faster than AMD could make them (although part of that was due to TSMC's yield issues at the 28-nanometer node).
However, when NVIDIA struck back with its Kepler GPUs, built on the same 28-nanometer node, the competitive landscape was turned on its head. NVIDIA, unlike AMD, stripped out much of the non-graphics "compute"-oriented portions from its highest end consumer GPU. This meant that NVIDIA could offer the same or better gaming performance as AMD, but it could do it with a die size of 294mm^2 against AMD's 352mm^2, and at lower power levels.
While NVIDIA quickly gained share on the desktop, the notebook GPU scene was even better for the company. Thanks to NVIDIA's "Optimus" technology, which allowed notebooks to seamlessly switch between Intel integrated graphics and NVIDIA's GPUs to maximize battery life, and thanks to a more efficient GPU architecture, NVIDIA cleanly swept both the Ivy Bridge and Haswell generation of Intel notebooks.
Hawaii, NVIDIA price cuts, and GTX 780 Ti launch
After several quarters of bleeding market share, AMD began to strike back, at least in the desktop GPU segment. The company began to offer it's "never settle" bundle in which the company threw in a pile of great 3D games with each of its graphics cards. This began to reverse the share losses, and although it brought back some lost revenue, operating profit in the company's graphics division remained anemic. AMD needed more if it wanted to compete profitably. That's where Hawaii comes in.
With AMD's Hawaii family of cards (R9-290 and R9-290X), AMD stunned the world by offering excellent products at extremely competitive prices. For $399, AMD was willing to sell customers 6.2 billion transistors, and performance roughly equivalent to NVIDIA's much higher priced GTX 780. It's a gutsy move that could help to win back share, but it's probably still very difficult from an operating profit perspective.
NVIDIA, in a bid to counter this move, lowered prices on its GTX 780 and many of the cards below it. To secure the high-end crown, it launched the GTX 780 Ti, a fully enabled version of the GK110 chip that powered both the GTX 780 and the GTX Titan. Of course, the chip, unlike Titan, sees its double precision floating point performance artificially handicapped (to protect sales of the high end Tesla cards based on the same silicon), but it secured NVIDIA's leadership at the top of the gaming enthusiast heap.
Foolish bottom line
At the end of the day, NVIDIA tends to do what it takes to secure single-GPU leadership. It is interesting to note that while the Hawaii chip sports 6.2 billion transistors, the GK110 from NVIDIA comes with 7.1 billion transistors. NVIDIA is able to justify such large die sizes because the "big" GPUs (NVIDIA does two lines -- the "small" mainstream/high-end one and the "big" ultra-enthusiast/professional one) are designed to be sold primarily as workstation/HPC products for multiple times what a top-end PC gaming card will fetch.
AMD has been putting up a good fight and the game console wins help strengthen ties with game developers. But, NVIDIA wasn't going to take any threats to its leadership lightly. It's nice for NVIDIA shareholders to know that all the company had to do to regain its status as the vendor of the fastest graphics chips was to enable a block on chips that it had been shipping for a year. It will be interesting to see what the market share trends look like over the next couple of quarters.
Ashraf Eassa owns shares of Nvidia and Advanced Micro Devices. The Motley Fool recommends Nvidia. Try any of our Foolish newsletter services free for 30 days. We Fools may not all hold the same opinions, but we all believe that considering a diverse range of insights makes us better investors. The Motley Fool has a disclosure policy.