The Inquirer-Home

Prescott to be slower than P4 Extreme Edition?

Marchitectural conundrums of our time
Tue Sep 30 2003, 18:42
IT'S BEEN TWO YEARS since Intel cheese Paul Otellini told an IDF that megahertz no longer mattered. Back then he said that rather than simply getting faster, chips should get smarter. He was talking then about Intel's Hyperthreading technology, and suggesting that the chips of the future would be multithreaded and multi-cored, making a slice of silicon 'smarter', rather than just cranking up the clock speed.

We're now about to see a fifth-generation Pentium, which we don't expect to have more than a single core, but should at least sport some clever multithreading, or at least some improved Hyperthreading, as well as some smart 'Louis XIII' extensions.

The chip will debut at 3.2 and 3.4GHz, we learned, with later chips coming in at both higher and lower clock speeds.

Prescott chips will replace P4s quite rapidly, we learned, except, perhaps, for the sore thumb, the supposed Athlon 64-buster and the "the fastest thing that would be released by any company this year", the humbled Pentium 4 Extreme Edition.

There can be little doubt that this chip was simply an AMD-buster. Indeed, when Aces prematurely posted its benchmarks for the chip, they were forced to withdraw them having discovered, they said, that the Intel spinner who passed them it forgot to mention that benchmarks should not be posted before September 23rd, coincidentally, of course AMD's 64-bit day.

Intel truely believed the P4EE would prove faster than AMD's 64-bit offerings and we can imagine the Intel Overclocking Lab guys fiddling with a 64-bit Athlon, searching for a quick way to beat it on speed. "Try whacking a few Megs of extra cache on a Xeon..? Hey guys, we might have something here."

Of course, when we published our benchmark comparison that showed a 64-bit Athlon trouncing the Extreme Edition, Intel was suffering from a bout of red faces all round. The company sought to point us at tests that showed their baby in a better light, but we were happy with the fairness of our testing and found the results quite reliable enough.

So the Athlon 64-buster failed to bust the Athlon 64.

Indeed, the Pentium 4 Extreme Edition is becoming a millstone round the behemoth's neck and the spoiler tactic may already be one it regrets.

Worse, Intel announced the chip, so will have to deliver it. Customers paying three grand for a Xeon are wondering what value there is in a chip that, souped up under a different label, will cost, let's say, around a quarter of that price, and the target market - those pesky, performance-hungry gamers - will be happy to plumb for the faster solution - and one sporting 64 bits no less.

Just to add to Intel's woes, The P4 EE is likely to be quicker than the opening Prescott offerings.

Prescott will be an impressive chip. Don't get us wrong. Hyperthreading will help. Straining the silicon will give those hapless electrons more room for manoeuvre, and no doubt Lou Burns' mysterious extensions will toss something useful into the pot. It'll have a Meg of cache too. But will it compete in terms of raw speed with a P4 EE, let alone a 64-bit AMD offering? Probably not.

But the architecture will scale, of course. And the move to a 90nanometre process will help Intel cut costs and deliver the chip by the shed load.

In the meantime, the chipmaker will suffer the embarrassment of delivering a new chip that may be demonstrably slower than one it already has on the market, which in turn has already been 'trounced' by 'the competition'.

And since Intel has already been banging the "megahurtz don't matter" drum for a couple of years, without anyone really noticing, how will they explain away the performance gap?

What will be the message come Prescott day? The spinmeisters and masters of megahurtage are sitting round tables right now, trying to work it all out. There's one economic argument that always works fellas - if you can't beat 'em on speed, beat 'em on price. µ

See Also
Megahurtz don't matter, says Steely 3.5GHz Otellini


Share this:

blog comments powered by Disqus
Subscribe to INQ newsletters

Sign up for INQbot – a weekly roundup of the best from the INQ

INQ Poll

Heartbleed bug discovered in OpenSSL

Have you reacted to Heartbleed?