ED#76 : More Trouble For AMD In 2008?

AMD continued a losing streak for the fifth quarter in a row when they posted a massive net loss of 1.772 billion dollars in the 4th quarter of 2007. Of course, they blamed most of it ($1.675 billion) on their acquisition of ATI. Without that, ATI said it would have been a much smaller loss of only $97 million.

It certainly doesn't bode well for AMD's future in 2008. Even their launch of the long-awaited Radeon HD 3870 and 3850 graphics cards could not help them wrest the performance leadership in 3D graphics from arch-rival, NVIDIA. NVIDIA has since launched an even faster GeForce 8800 GTS 512MB graphics card as well as two lower-cost SKUs, the GeForce 8800 GS 768MB and 384MB graphics cards, to nullify AMD's new graphics cards.

With the long delays in getting newer and faster models of their new Quad-Core Opteron (Barcelona) workstation/server and Phenom desktop processors out into the market, AMD is faced with a daunting year ahead. George Ou of ZDNet has this to say about AMD's processors :

It does look like the Phenom 2.4 and 2.6 GHz have been delayed till Q2-2008. There's no word on the server chips though and at this pace, we may see Barcelona 2.6 coming out the same time as Intel Nehalem.

Originally when Barcelona was going to be going up against Intel Clovertown 2.66 in the dual-processor space, Barcelona was looking like a champ especially when you factor in SPECfp_rate2006 performance. Then with the first delay, we were looking at Barcelona 2.6 versus Intel Penryn 3.2 on the Stoakley platform where AMD at least comes within the ball park and can command some decent margins but that is not to be.

If we're looking at Barcelona 2.6 versus Nehalem in Q3, then Barcelona becomes nothing but a low-cost second-tier alternative. Supposedly AMD's 45nm Shanghai will be out by then but that won't be competitive against Nehalem as it lacks the execution engine enhancements and AMD will go from a massive memory bandwidth advantage to a 33% deficit because Nehalem is reported to have 3 channels of DDR3 versus Shanghai's 2 channels of DDR2. Nehalem's massive execution engine which improves on the already-great Core 2 chip with super fast memory access will propel it to stratospheric performance levels.

Of course this assumes AMD will deliver Shanghai on time which is kind of hard for me to believe. You have to remember that it takes Intel a year to go from first silicon to commercial product and Intel has a better execution track record. AMD won't even deliver first silicon until this quarter and even that's not a foregone conclusion that they will deliver. This would put production Shanghai parts in Q1-2009 if everything goes right.

On the desktop platform, Phenom quad-core 2.6 will generally perform slower than Intel's Yorkfield quad-core 2.5 GHz which forces AMD to sell quad core chips at less than $250 at the end of this quarter. Unfortunately for AMD, 2.4 and 2.6 will be delayed till Q2 so AMD will only play in the barely profitable sub-$250 market. 2.4 to 2.6 GHz quad-core Opteron variants of Barcelona will be able to command better margins because of the memory bandwidth factor in dual-processor servers if they can launch before Nehalem.

If George's predictions are correct, then AMD looks set to continue losing money throughout much, if not all, of 2008.

Questions & Comments

If you have a question or comment on this editorial, please feel free to post them here!