Intel needs to think along parallel lines

More development for parallel processing could help Intel's new chief to revitalise the ailing giant

Intel’s Paul Otellini has followed the pattern of many incoming chief executives by deciding to reduce overall staff numbers. The reason is the usual one of a few disappointing quarters of financial results. But Otellini is looking in the wrong place for big improvements, because Intel’s troubles will not be solved by firing a few staff.

Among the targets for the axe are senior marketing staff, involved in market development and advertising, because Intel has more managers in these positions than its arch-rival, AMD. However, this is likely to be a bad strategy, as the redundant staff could join the opposition. That would give AMD even more clout in the marketplace, where it has already stolen a probably temporary lead in multicore processors. If Intel isn’t careful it may slip gently into the role of permanent also-ran.

Meanwhile, rumours also suggest that Otellini plans to rebalance the workforce by recruiting more customer-facing sales staff, suggesting that he is aware it is an increasingly uphill struggle to sell Intel processors. But adding more sales staff to a firm where, for many years, the processors almost sold themselves will, in a tough market, just equal less profit.

Intel’s troubles cannot be solved by cutting overall staff numbers and adding a few to particular departments. The real problem is that Intel is suffering the traditional hardships of any company that has pushed a technology too far and too hard. The x86 architecture has passed its sell-by date.

For some this idea may seem heretical – particularly the armies of IT managers who have built empires on the back of x86-based servers. Well, they needn’t worry: x86 servers are not going to disappear overnight. But some experts are now seriously suggesting it’s time for chip makers to move on to a new generation of technology using parallel processing.

Steve Prentice, vice-president and distinguished analyst at Gartner Group, recently noted the trend to increase processor and system performance with parallelism and more built-in redundancy on the chip.

Of course, we already have dual-core x86 chips, and IBM sells four-core Power processors, and in a few years we will probably be able to buy Xeon and Opteron chips with eight or 16 cores.

But Iann Barron, once leading light of famed UK-government backed startup Inmos, suggests that if we can get more parallelism into computer software then faster systems will be possible. Given that some software would benefit from a rewrite to optimise it for 16-way hardware, there’s an opportunity to rethink software architectures and possibly move to alternative processors.

So Otellini firing a few staff won’t save Intel’s bacon. To counter what is coming and not just be left with a dwindling market, Intel needs to make drastic changes in product direction. And that will need very motivated staff. Hatchet jobs rarely achieve that.