Intel released their 2GHz P4 (Willamette core) on August 20, which was significant not only because it gave Intel bragging rights for the first 2GHz processor, but because it also put Intel back in the running with regards to desktop performance. This will be the highest speed processor using this core. At the same time, the 478-pin socket was introduced in parts from 1.4GHz on up. So far, there are no good explanations for the change from 423 pins, but perhaps we will see some of them utilized for features that have not yet been activated. Interestingly, since the intro of the 478-pin parts, the 423-pin parts have been flooding the market, going for as much as $25 less than the same speed 478-pin parts. This is likely due to OEMs dumping unsold inventory, or possibly Intel getting rid of same?
There have been many detractors of the P4 architecture, with most focusing upon the instructions-per-cycle (IPC) or ‘clock-for-clock’ performance versus the Athon. The real issue, of course, is how far Intel can push the clock rates and what features might still be disabled. Intel claims they will have a 3.5GHz model in 2002, and they have admitted that the Xeon parts have Hyperthreading (or Multi-threading) circuitry already on-chip. In addition, benchmarks of P4 DDR chipsets reportedly show an improvement over the i850 chipset by as much as 30%. It would appear, therefore, that Intel’s greatest mistake with the P4 was not the long pipeline, or any other architectural issues, but instead their insistence of matching it with DRDRAM instead of DDR.
Next month, the Northwood core is expected to debut at 2.2GHz. It will not only be produced on a smaller (.13u) process, but will have an increased L2 cache size of 512MB. The faster speed, larger cache and ability to run it on DDR chipsets from SiS and ALi (and possibly even Intel) will very likely give the P4 detractors something to think about.
The Tualatin desktop processor (256K cache) was introduced in at the end of July at 1.2GHz, and will be the end of the line for the desktop PIII. It will continue to be marketed for small form-factor systems, however the DIY market will not see much of it. A variation of this part has been relabeled as a Celeron, with a 100MHz FSB and with the data prefetch feature disabled. The PIII processor-M, based upon the same core but with 512K cache, is also being offered in both 478-pin and 479-pin packages.
Reminiscent of the now infamous Intel stumbles of the past few years, a bug was discovered in the 900MHz PIII Xeon processors earlier this this year forcing them to stop shipments, which Intel admitted to in July. They also pushed back the introduction of the P4 Xeon (code named Foster) due to validation issues. Since no other x86 manufacturer competes in this space and the PIII Xeon has already been fixed, it has little impact on their market, unlike the previous problems with desktop parts. Intel has also claimed that they have made improvements in the McKinley design that will allow it to perform up to 50% better than the Itanium
After a year of pushing the envelope on processor clock rates, and being the first to 1GHz in 2000, AMD is being exceedingly stingy with the clock speeds this year. After introducing the 1.4GHz part this spring, no faster parts have yet appeared, though a 1.2GHz MP part was introduced. Reportedly, the fastest Athlon that will be released this year will be at around 1.6GHz, which also is rumored to be close to the fastest that the part will operate using a .18u process. Assuming what appears to be a standard clock rate improvement of between 40% and 50% for a die shrink, the move to .13u next year could allow the Athlon to get to 2.5GHz or thereabouts.
On August 20, the 1GHz Duron was released, which utilized the new Morgan core that uses less power has some additional performance enhancing features over the older Spitfire core, such as data prefetch and SSE. More recently, the 1.1GHz Duron was released, likely to combat the 1.2GHz Celeron. Next week the Athlon XP is expected to be announced, which also uses a newer core (reportedly identical to the Athlon MP) with similar performance enhancements. Several speed grades will be introduced, including 1.5GHz, and will apparently quickly replace the ‘older’ TBird core parts.
AMD supporters were a bit shaken after IBM, Micron and Gateway announced they were ‘standardizing’ on Intel processors, mostly due to cost cutting measures in this difficult market. A bit of analysis revealed all of these OEMs continue to offer some product lines with AMD processors, and estimates have been made that the changes affected less than 5% of AMD sales. Discussions with motherboard makers indicate that Athlons are still the most popular processors in the DIY market and growing, and the comments from AMD that their shipment numbers remained about the same as Q2 seem to validate this. Somehow, AMD has to figure out how to leverage this strength into the home consumer and business markets.
In one of the most controversial moves of the year, AMD has decided to market their parts using a new performance rating that is supposed to give consumers a better idea of how the Athlon parts match up to Pentium 4 parts. The 1.53GHz will be known as the 1800+, meaning that it performs as well as a P4 1.8GHz, while slower parts will get numbers such as 1700+, etc. Obviously, AMD had to do something to counter the ‘MHz equals performance’ misconception (which they helped to propagate last year in the MHz race), particularly since a potential 2.5Hz Athlon vs. 3.5GHz P4 could be problematic for AMD. Unfortunately, though some are staunchly defending this move as appropriate, what it appears to do is to validate the concept that the Pentium 4 is the ‘benchmark’ processor. I don’t think that this is the impression AMD really wanted to give. Possibly, if the HP/Compaq rumor mentioned on the first page is true, there will be a branding effort to get the name recognized on its own rather than in relation to the P4.
VIA seems to be working hard to get its C3 processor (based upon the Cyrix and IDT processor cores) recognized in the market. The processors are not in the same league as Athlon/P4 or even Duron/Celeron, but they do have a clear market segment to target. With the much faster clock rates of the desktop lines from AMD and Intel have also come greater heat dissipation, which in turn means more sophisticated and noisier cooling solutions. VIA has realized that there is a growing market for low-power and low-noise systems, which they believe they are in a good position to satisfy. Their own C3 webpage emphasizes this, and recent reviews on both Lost Circuits and this site indicate that they have done a good job of targeting this market. It remains to be seen if this market is really large enough to justify the effort, but with planned speeds over 1GHz they could grab some market share.
Discuss (15 comments)