Steve Jobs' much publicised claim of a two-to-three-times speed increase in the new Intel-based iMac is bunkem, according to the latest benchmark tests.

The new Intel-based iMac G5 is not much faster than its PowerPC predecessor, Macworld has discovered.

Macworld benchmarks show that the iMac G5 running an Intel 2.0 GHz Core Duo gained only between 10 to 25 percent in performance compared to the IBM architecture, far lower than the doubling in throughput widely claimed by Apple.

The magazine carried out a series of tasks using two Apple applications, iMovie and iPhoto, discovering that performance boost ranged from worthy to absolutely nothing, depending on the application function tested.

For Apple applications that aren’t yet Intel native - running using the Rosetta emulator - the performance is only half what it would be on the PowerPC architecture that preceded the switch to Intel.

This is potentially calamitous for the average Mac user because until Apple applications arrive that have been compiled to run on the Intel chips natively, they will be forced to use Rosetta and see performance drop compared to their PowerPC machines.

The uncomfortable irony in this is that after years of rubbishing everything to do with Intel for its association with mainstream PCs, Apple performed a remarkable u-turn once it decided to move to the Intel architecture.

The company’s website projects the performance gain expected from the new chips with as much hyperbole as it can muster. “This revolutionary bit of technology is actually two processors built into a single chip, giving iMac up to twice the horsepower it had previously. So the wows will come faster than ever before,” it purred.

Now it transpires that the two processors add up to more like one-and-a-bit processors in extra zip, about what you’d expect given that this is the latest dual-core design.

The magazine puts forward an explanation for the gulf between what is claimed and what has been found to be the case - biased benchmarking.

Apple generated the spectacular doubling in performance by using what are known as "synthetic benchmarks", programs designed to test chip throughput using raw approximations of how applications behave. Long a controversial subject in chip testing, these benchmarks do not necessarily correspond to the actual performance users will experience with real-world applications.

Vendors are assumed to exaggerate performance hikes from new chips, so it is not a surprise that the move to Intel has not yet generated the returns claimed for it. But this is an unusually sensitive time for Apple. It is in the process of moving its loyal user base to a new hardware platform and needs them to keep paying premium prices for its hardware.

In particular, it is apparent from the disappointing scores from the Rosetta emulator that the key to this historic move will be the availability of new, native applications such as Adobe Photoshop and not simply the underlying chip platform.