The big news, of course, is the official launch of the Pentium 4 (code named Willamette) at speeds of 1.4GHz and 1.5GHz. There was a lot of anticipation about this launch, because it was considered a crucial event for both Intel and AMD. Intel has suffered of late because of their inability to ramp the PIII beyond 1GHz using their current production process, as well as the rejection of the i820 chipset and other perceived bungles. For AMD, the concern was whether this new core would leapfrog Intel into the performance lead again.
Based upon the reviews of this processor, it appears that initially AMD might have gained the most in terms of PR, but it is also apparent that Intel has once again introduced a core that will carry them a long way into the future. These reviews show that the P4 provides no significant performance benefit over the Athlon or PIII using current mainstream applications, however, the synthetic benchmarks show that there are plenty of opportunities for improvement. There are also opportunities for a new class of applications that can take advantage of the memory bandwidth and floating point performance it provides.
Recent reports seem to indicate that Intel has decided to scale back their aggressive plans to ramp the P4 early in the year, most likely because feedback from OEMs was not positive. Fast ramps make systems obsolete quicker, and it is anticipated that the initial demand for the P4 will not be huge. I would expect that once the Northwood part becomes available, we will see some fast ramps as Intel once again goes for the performance crown.
With the introduction of the 1.2GHz Athlon and 800MHz Duron processors in late October, AMD appears to have made their final speed upgrade for the year, however they should be releasing a 266MHz FSB version in early December, with speeds of 1GHz and 1.2GHz. Late in Q1, we can expect to see the Athlon running on multi-processor capable motherboards
Early in Q1 we should see the 1.3GHz parts for both 200MHz and 266MHz FSB speeds. Since Intel is apparently not pushing the GHz race right now, AMD appears to be satisfied with staying just a few steps ahead in performance. Recent reports indicate that the 1.5GHz part may not be available until close to mid-year. Sources close to AMD have indicated that with the Mustang core at .18u, it is not beyond reason to expect the Athlon to reach 2GHz with some voltage tweaks, so the slower speed ramp seems to be in their favor. In Q4, AMD will switch to .13u, allowing even faster operation at lower power (and cheaper cost)
Because of the huge potential with the Pentium 4 architecture, AMD appears to have only about a year left in the K7 core before Intel is able to run away in raw speed. For this reason, AMD realizes that their future now depends upon the Hammer line of processors. The AMD roadmap provided at Comdex shows the Clawhammer (single and dual processor capable) will be sampling in the fouth quarter of 2001, with production shipments anticipated in the 1st quarter 2002. Sledgehammer, which will be 4 and 8-way capable, should start sampling early in 2002, with production shipments the next quarter.
According to AMD, this core will contain ‘architectural performance innovations’ and the ‘pipeline supports aggressive clock speed scaling’. Sounds somewhat like the direction Intel took with the P4. AMD, however, says that performance per clock will be better than the Athlon. The one benefit AMD has over Intel offerings with this core is that it should run both 32-bit and 64-bit code with equal efficiency. This means that existing applications should suffer no penalty with ‘legacy’ code (as happened with the Pentium Pro processor), so no porting effort is necessary just for better performance. The early indications are that the Linux community is eager to take advantage of the 64-bitness of this new processor, with Sun also apparently showing interest.
VIA introduced their VIA/Cyrix chips without much hoopla. Initially offered at 500Hz to 600MHz, they are targeted directly at the low end Intel and AMD have left behind. However, the speed ramp appears to be very aggressive, with 1GHz speeds anticipated by April. A weak floating point unit will prevent this chip from being used in high-end systems for gaming or 3D graphics work, but VIA fully intends to take as much of the low-end market as they can, and it seems they might have an eye on pushing into the higher-end space with future offerings.
Be the first to discuss this article!