Exclusive: The Xbox 360 Uncloaked -- Part 2
- May 15, 2006 16:33 PM PST
- Email this!
Here's part two an excerpt from Dean Takahashi's newly released book The Xbox 360 Uncloaked. Read why the Xbox 360 was originally code-named Trinity.
At the start of 2005, Bill Gates delivered a keynote at the Consumer Electronics Show. While he highlighted some Xbox games, he didnt say a word about the new console. He did say that Microsoft expected to move 20 million consoles by July.
The Xbox team showed a demo of Forza Motorsport during the keynote. But in the middle of the demo, the game froze. Conan OBrien, the talk show host who joined Gates on stage, cracked, Whos in charge of Microsoft anyway? The speech reminded everyone of the buggy software that Microsoft had produced over the years. Gates did note that Microsoft and MTV had struck a strategic partnership, but the significance of that announcement wasnt clear to Xbox fans at the time.
Gates did announce a series of portable media centers aimed at taking some of Apples iPod customers. But the announcement was received with considerable consumer skepticism. Apple had taken the music player market by storm in less than two years.
At a silicon review meeting, Todd Holmdahl opened with, Happy New Year everyone. Welcome to launch year. There were groans around the table. The statements gravity sunk in with Bill Adamec.
Fortunately, IBM had completed its final tape-out on the PowerPC microprocessor on Jan. 31. IBM was ready to start debugging its factories and the chip itself in preparation for high-volume manufacturing. It was just 17 months after IBM signed its contract. The chip program was in good shape, but it had little room for error. After Waternoose went off to the factory, the IBM engineers started working on the cost reduction of the silicon design for the second year of production. Jeff Andrews had some scary moments, but he breathed a sigh of relief when his chip was done. The final IBM chip would come back just before E3 in May.
The ATI graphics chip, however, wasnt finished yet. Hagens team made changes to fix the known bugs in the graphics chip. They sent it back to the Taiwan Semiconductor Manufacturing Co. factory and they got the chip back in January. They had to run a huge suite of tests, using a few hundred computers, to make sure they didnt introduce a new bug. But they had not yet found a solution for the most difficult bug. The chip locked up and failed to work under certain conditions. Although the ATI team could extract data about every piece of the chip at the instant that it froze, the information didnt help them. They had to start writing tests to zero in on the problem. On daily calls with Microsoft, they referred to the problem as a phenomenon.
Then the engineers isolated the bug. Two logical functions happened in sequence, one after another. Most of the time, they worked fine. But once in a while, the clock in the chip fell victim to electrical noise. It would jitter, or pause long enough to cause a delay. That delay was long enough to insert itself between the two logical functions. When that happened, the chip failed. Hagen said it was hard to find because the jitter was an analog problem, occurring a couple of levels of abstraction below where the digital engineers were looking for it. In about February, ATI submitted the solution for the flaw. They revised the logical data in the design and drew up plans for another chip. TSMC began working on it.
