Original URL: http://www.theregister.co.uk/2004/01/22/amd_quietly_drops_universal_chip/

AMD quietly drops universal chip speed metric plan

Other chip makers not keen, apparently

By Tony Smith

Posted in Channel, 22nd January 2004 11:39 GMT

AMD's attempt to define an industry-standard performance measure for modern microprocessors has been quietly shelved, according to the executive in charge of the initiative.

The appropriately named Hal Speed, head of AMD's True Performance Initiative (TPI) told The Tech Report that the TPI no longer has such a goal.

The TPI was put in motion when AMD introduced its Athlon XP performance-derived numbering scheme, back in October 2001. At the time, AMD's Pat Moorhead told The Register that the '1800+' naming scheme was simply a "bridge metric", to be used until a scheme could be developed that other chip vendors could adopt.

The result: processors could all be rated using a measure as standard as clock frequency but more relevant to today's superscalar, multi-threading CPUs. Finally, the 'megahertz myth' could be laid to rest.

Such a measure would surely appeal to the likes of Apple, whose Motorola and IBM-made chips offered more processing horsepower than their clock speeds suggested. Perhaps even Intel might approve, looking to a future where a 1.6GHz Pentium M was able to outperform a 2.4GHz Pentium 4.

Moorhead wouldn't comment on attempts by AMD to discuss its plan with such vendors, but Speed suggests that Intel at least was approached. Whatever, feelers put out by AMD appears not to have yielded anything, and with the Athlon 64 still being marketed with performance ratings, it does indeed seem likely that AMD is no longer pursuing its original TPI plan.

In any case, said performance ratings appear to have been accepted by the market, if not as a measure of performance against Intel's ongoing use of clock frequencies but at the very least as a guide to which AMD chips are faster than other AMD chips. For vendor-to-vendor comparisons, many folk rely on benchmarks we reckon - though these are becoming equally unreliable as chip architectures diverge. ®