Cell processor yields looking dismal
IBM appears to be having a tough time making a sufficient number of working copies of its Cell processor, perhaps causing further problems for Sony's PlayStation 3.
IBM Vice President of Semiconductor and Technology Services Tom Reeves was quoted in an article on Electronic News last week saying that initial yields for the Cell chip are somewhere between 10 percent to 20 percent. The company thinks it can improve that with logic redundancy techniques, but even 20 percent to 40 percent is still a shocking number in the world of high-volume chip manufacturing.
Yield is an extremely important metric in the chipmaking industry. It's a measure of how many working chips can be cut from a single silicon wafer. There are always going to be a few defects on the wafer that render some processors useless, but Intel, Advanced Micro Devices, Texas Instruments and others spend billions trying to get as close as they can to perfection.
Trying to get any of those companies to comment on their specific yields, however, is like trying to pin down Barry Bonds on his workout regimen. But they are definitely higher than 40 percent, a number that would probably make Intel Chief Financial Officer Andy Bryant faint.
The lower the yields, the more expensive it is to produce large numbers of chips. And with some analysts concerned already concerned about the profitability of the PS3, this news won't make things easier for IBM and Sony.
But generating high yields on brand new processor architectures is tricky, said Dean McCarron, an analyst with Mercury Research. Often times chip makers will struggle with the initial yields of a new design until they go through a transition to a smaller manufacturing process, after which they have more room to improve areas that were prone to defects, he said.
UPDATED - Electronic News later added a note to their article that said Reeves had clarified his statement, saying he was referring to yield problems with large processors in general, not quoting specific numbers for the Cell processor. But it's hard to imagine how else he came up with the specific numbers he used in making the comparison. An IBM spokesman said he could not comment on yields for the Cell processor.
Here's the text of Reeves' quote, from the article linked above.
Electronic News: What's the defining factor that makes some chips better than others?
Reeves: Defects. It becomes a bigger problem the bigger the chip is. With chips that are one-by-one and silicon germanium, we can get yields of 95 percent. With a chip like the Cell processor, you're lucky to get 10 or 20 percent. If you put logic redundancy on it, you can double that. It's a great strategy, and I'm not sure anyone other than IBM is doing that with logic. Everybody does it with DRAM. There are always extra bits in there for memory. People have not yet moved to logic block redundancy, though.