Re: addition the k5 and cyrix all were way slow on MHZ but had better IPC than Intel.
We had the infamous PR rating fiasco.
In April '98 when cyrix was 233/300 and the top PII was 350, Cyrix was competitive.
But by March of '99, cyrix was only at 250/366 and PII had been at 500 since February of '99. Cyrix wasn't just behind PII in mhz, they were behind in performance - and the 250/366 should have been rated at 250/300. Meanwhile Intel was pretty much giving away 300MHZ CPUs by then. Add to this many reports of problems with the MII, and the story is much more than high IPC not being significant.
I see something similar to the PR rating disaster for Intel if their IPC is actually as bad as postulated by some. They would be presenting their 1.5GHZ "rated" chip, as a 1.5GHZ chip. If AMD's 1.3GHZ chip proceeds to substantially outperform the Intel 1.5GHZ in many benchmarks, P4 could fall prey to the same market contempt that doomed MII and nearly killed the original Celeron - and Intel can't simply put back the cache on P4 they way did on the original Celeron.
Here's an example from another market. I bought a Mustang Cobra Convertible last January for several thousand dollars below the invoice price. The Cobra is a limited production model of the Mustang that Ford hand builds each year - it has a different engine and suspension than the regular Mustang. Cobras are much less common than regular Mustangs (they built only 4,050 convertibles in 1999) and they usually sell for more than list price, let alone invoice. 1999 started out as a great year for the Cobra because it was the first year for its new independent rear suspension (IRS - and it was only IRS ever offered on a Mustang). But when the owners started testing their cars on dynamometers (Cobra owners do such things!) they found out that most actually had closer to 300hp than the advertised 320.
Sales for the car pretty much ended, and Ford had to provide huge rebates on a model that was normally sold at well above list price. The early buyers who typically paid $3,000 to $4,000 over list were pretty disappointed.
Eventually Ford recalled all the cars to perform $1,800 worth of engine work (I have my copy of the invoice to Ford) to bring them above the advertised 320HP.
The point I'm trying to make is that an flagship performance product that doesn't perform quite up to expectations can be harshly rejected by the market. When buyers pay for a premium product and only get a good product, they are very disappointed.
The difference between the 300HP IRS Mustang and the expected 320HP IRS Mustang was trivial - but made the product almost unsaleable in the marketplace.
If those early performance numbers are accurate, the same thing could happen to P4.
Dan |
