Processors

The main thing to be said about this year is that it was quite still and stable. There were of course some considerable events that will tell upon in the future. But in general, the year 2000 appeared to be a successor of the tendencies that had been started the year before than it gave a way to new ones.

The main tendency that was put a finish to is a clear leadership of the company AMD. It's hard to say who should be given the credit here: AMD that worked hard or Intel that helped them. Non-stop problems with deliveries in the first half of the year, recurring technological problems (a put-off of the PIII 550 in January, errors in the MTH Timna), a clear fall behind AMD in the field of increase of clock speed of Coppermine, and as a result, Intel started announcing that "a processor will be released in two months" instead of "a processor is already debuted" in order to catch up with AMD which was releasing a completely real analog, cancellation of the Timna and Itanium, a put-off of the Pentium 4 for a month…

However in the coming year the companies should exchange their places, and by the end of the year Intel should become a clear leader: in terms of speed potential the Pentium 4 is higher than the Athlon which has already played its last trump card in the form of copper connections, while the Pentium 4 is only going to start their 0.13 micron technology. Besides, we should consider that a number of optimized for SSE2 programs will be increasing, and in certain time (that is needed to improve the platform) we will see a clear hit considering an aggressive price policy started by Intel regarding the Pentium 4 from the very beginning of its birth.

AMD spent the last year with dignity, and as a result it received quite a good money (several quarters with record incomes, it was first time when sales volume exceeded a billion dollars) and change of opinion about this company. Look just at the Gateway that passed to AMD.

Nevertheless, the both companies have managed to surpass 1 GHz - an old core P6 has enough strength for it. But when taking up 1.13 GHz it failed to fix it and they decided on another trial in the middle of 2001. But the pause lasted not so long - in the end of the year the Pentium 4 was debuted, it could easily compete against the old models of the Athlon which were too tough for the Pentium III.

On the Low-End market Intel appeared even worse. The Celeron that had been a leader quite for a long time here lagged behind the Duron from AMD by the end of the year. AMD has three advantages here: price, clock speed, performance. The spring born Celeron based on the Coppermine turned to have a decreased FSB frequency to 66 MHz what caused a considerable fall in the performance. Intel can be understood - otherwise the Celeron started competition against the Pentium III, but on the other hand, the Duron with its 200 MHz FSB is doing with the Celeron what it wants.

In January this should be corrected, the Celeron will get 100 MHz FSB. But with the difference with the faster Duron by 1.5 times. And such state will last the whole 2001 year: Intel is not going to pay much attention to the Celeron and of course they wouldn't move it to 0.13 micron technology in 2001. At the same time AMD is releasing the Morgan, which should bring benefit at the expense of low price, though with some decrease in performance. So, while in the High-End market everything is unclear for the next year, in the Low-End AMD is an evident leader.

Talking about these two companies, we have forgotten to look at VIA. Well, they were a bit unlucky - the Joshua that was produced by National Semiconductor, and the Samuel produced by TSMC. And the 0.15 micron Samuel 2 turned to be no better in terms of performance, although it's beyond any competition in the prime price. But all this has already passed, and the first thing developed by VIA (to be more exact - by Centaur/Cyrix developers) is the C5X core that will serve a base for the Ezra processor, it can really change the situation.

It seems like it will be more successful on the market of game stations and notebooks. And the mobile version Samuel 2 which was released this year is aimed exactly at this sector. AMD hasn't done their best here. The company could offer nothing better than the K6-2+/K6-III+, updated old processors which can not compete against Intel neither in clock speed nor in performance. And considering the tendency of customers to buy only High-End, AMD told us a sad story: it couldn't debut mobile Athlon or Duron. Nevertheless, it must take place in the nearest future, and then we will see how it can change the situation on the mobile market.

Today the greatest fight (I mean technology-marketing sector) is hold by Intel and Transmeta with its Crusoe. I must admit that the newcomers could attract a vivid attention of notebook and game station manufacturers. However, they change their opinion quite quickly: the greatest notebook manufacturers (IBM, Compaq, Toshiba) first announcing their prototypes of notebooks on this processor, quickly called them back. In the end of the year Transmeta was poured over with a shower of not so pleasant news: manufacturers of notebooks based on the Crusoe came across failures in its work.

Nevertheless, the company didn't give up - it managed to create a good base in Asia, and it has quite perspective models in their roadmap. However, it should compete against the 0.13 micron Ezra which also has quite good effectiveness together with performance, mobile processors from AMD, and of course Intel. And the latter means firm business on the mobile market: from the release in January of 600 MHz Coppermine with SpeedStep support, the clock speed of it has nearly reached 900 MHz, and in this year Intel promises announce the 0.13 micron 1.2 GHz processor version.

AMD has far not brilliant state of affairs in the field of servers and work stations. While Intel in 2001 will have two solutions: the Pentium III Xeon with 1 and 2 MBytes L2 cache and the Foster (512 KBytes and 1 MBytes L3), AMD will have only the Athlon working in dual-processor configurations. The Mustang has canceled.

In general, Intel has spent quite an interesting year: on the one hand, the Pentium III Xeon at last got a decent cache L2 - 1 and 2 MBytes (before they offered only light versions with 256 KBytes L2 - in fact, a usual Pentium III in the Slot-2 case) and with decent multiprocessor support, on the other hand it got stuck at 700 MHz. The 800 and 866 MHz processor versions which were to be released Intel has canceled "at the instance of public". And as a result, the next update (the last one) of this series will be only in the second quarter of 2001 - 900 MHz Xeon with 2 MBytes L2. And why to hurry? There is no any competitor.

At least, in this class. In general, the year turned to be quite scanty for server processors. In fact, we could see only the Sun UltraSPARC III with not 600, but 900 MHz. And HP announced their new generation of 64-bit processors - PA-8700 which use all the latest IBM developments in technological processes: 0.18 micron, copper connections, silicon-on-insulalor (SOI). The company keeps their PA-RISC architecture since Itanium is still not of benefit.

The processor was announced to work at 800 MHz, though in reality several demonstrated in August systems showed only 733 MHz, and Intel was ashamed to give out SPEC results for the Itanium saying that they were not so important. Nevertheless, in the last quarter Intel started deliveries of systems on this processor shipping around 20 systems to some important customers. In fact, the Itanium turned to be just a test system for IA-64 architecture.

One of the greatest competitors Compaq couldn't boast of its brilliant carrier either. The Alpha hasn't reach the planned 1 GHz - in the very end API announced only a 833 MHz processor version. However now, when IBM conclude an agreement with Compaq on production of these processors, the situation is likely to change.

There is more than a year left until a release of 64-bit AMD processors based on x86-64 architecture announced the last year, so today we can not take this into account.

Motherboards

The progress reached appeared quite solid, we have seen nothing similar since 1998 when the 440BX was announced. By the way, it managed to live up now. However, it had to use some kind of make-up, like external ATA/66 and then ATA/100 controllers, though some imperfections like absence of AGP 4X support couldn't be overcome. Nevertheless, the chipset is not going to give up - in the end of the year the company Standard Microsystems Corp. (SMC), which received it from Intel, integrated there ATA/66 support naming the product the Victory66.

Now we are witnessing an interesting situation - the 440BX is still alive, and the i820 managed to exist only a bit more than a year: from the end of 1999 to the beginning of 2001. It's no a surprise, though. First comes the story with its release, then an error in MTH and a call back of all boards on it (and the majority of i820 models were right with the MTH - nobody desired to mess up with the Rambus), that's why when Intel tried to reanimate the i820 with the help of announcing the i820E, it attracted no attention.

On the contrary, VIA was occupying a huge sector of the market: the i820 is not in demand, the 440BX is but it's produced in small quantities. And as a result the Apollo Pro133A based boards were produced … Especially considering dual-processor models which started being produced by the middle of the year. However, by the end the chipset became out of date, but the cosmetics came to help here as well: first, as external controllers, then we saw the 686B with ATA/100 support that allowed mainboard manufacturers to produce their old models in a new enclosure.

There is one more trump for VIA - this is of course popularity of AMD processors. Which was underestimated, though. In the very beginning of the year they started shipping the KX133 which turned to be the worst chipset from VIA for the last time. It undoubtedly lived a short life, though there were announced many boards on it due to absence of alternatives. But the AMD-750 was still more reliable variant for Slot-A. And with release of the KT133 in summer, there was nothing left to do: in fact it was the only chipset for newly released Socket-A Athlon and Duron. And VIA underestimated the potential of the latter being late with the integrated KM133 and damaging AMD business.

Another pleasant thing for VIA: a ban on sale of chipsets for Intel processors got canceled (yet in February ITC started investigation on accusations of Intel against VIA) : in June the companies announced a cross-licensing. Interestingly that it took place right in the following month after the Computex showed a complete disinterest of the industry in the i820E.

However the announced in the same June i815/i815E was accepted warmly. It was waited for a long time, and when it appeared nobody paid special attention to unnecessary integrated graphics. There we saw what we waited for - the platforms on it outperformed the i820 in speed being at the same time cheaper at the expense of the PC133 SDRAM.

And when in autumn this chipset started undergoing update, we could say that it was going to live long. There appeared the first announcements of dual-processor boards on the i815E, the i815EP with cut-out graphics, the successful mobile i815M that came in handy considering that on notebooks the i820 didn't appear (there was used an old 440BX). And the i815 position was strengthened by the Almador in favor of the i815EP with a new stepping providing support for the 0.13 micron Pentium III. That should last long.

However, as for integrated chipsets - nature adhors a vacuum - there are enough players. And SiS, that spent the year for nothing mainly due to the problems with production - the SiS630 based boards started being announced yet the previous year, but in fact they were too little on the market. In the end SiS, that carried great losses, accused UMC of all the troubles which couldn't produce enough chipsets for them. UMC responded a cancellation of their production and brought an action against SiS. Maybe now, with the SiS730 that is produced at their own work, the company will be more lucky. Especially considering that this integrated chipset for Socket-A really deserves paying attention to.

And VIA showed not the best side concerning integrated solutions. Of course, in the second half we saw both the PM133 and the KM133. But there practically were no boards for them. This year should be more successful - VIA appreciated perspectives of the Duron, and the i810E is easier to compete against than the i815E, moreover with the release of the KM266 and the PM266.

This will be just a middle level, though. On the top we will see the integrated chipsets of the newcomers - NVIDIA and ATI. The first product of ATI, taken from ArtX S1-S370 TL, appeared not very good and even couldn't reach the retail market. It was just a trial step - a real competition should start with the release of the ArtX2. NVIDIA will be near with its Crush11/12 and Crush17/18, it will give an easy time to nobody.

The latter is actually started preparation for the release officially announcing their south bridge MCP. The characteristics sound more solid: 800 MBytes/s LDT, powerful 3D sound, ATA/100, all possible communications. If the north bridge with the integrated graphics core GF2 MX will be well matched, the product will be just excellent. We still don't know the price, though.

As for Intel, it wasn't successful with the new south bridge - the ICH3 is put off until the first half of 2002, so several new chipsets disappeared from the roadmap of the company for 2001 (like the Tehama-E which should have become i850 with ICH3). The i850 after a month delay was released in the end of the November and then started to live without much trouble - all 4 beforehand announced motherboard manufactures released the products on its base; no claims for them, ahead there comes the second wave of boards on the i850 from other manufacturers.

The main new competitor is RRC (renamed ServerWorks), with its GrandChampion-HE. By the way, the company started the attack quite firmly - the boards on their chipsets appeared in the retail market, and from the grands like Tyan or Asus. Plus an actual blessing from Intel for representing in the server part of the market and conclusion of the 5 year agreement with IBM, and analog technological partnership with Adaptec.

A bit lower in the same field as the i850 are the PX266 and the SiS 645 which are to be released by the end of the 2001. They are coming with the DDR SDRAM support. But in the end of 2001 - beginning of the 2002 Intel should release an analog chipset - the DDR SDRAM version of the Brookdale coming in the third quarter of 2001. ALi didn't show anything interesting, though. But they said that they intended to debut a chipset for the Pentium 4.

But ALi is unexpectedly turned to be a leader in another field, being the first to release a chipset with DDR SDRAM support. Yet in July at the Computex they demonstrated patterns of platforms based on the AMD-760 and the ALi MAGiK1, and a month later Ali announced 4 first chipsets with DDR SDRAM support: ALi MAGiK1, Aladdin Pro5, and their mobile variants. And in August they demonstrated the first board on the ALi MAGiK 1 - Iwill KA266-R. and... that's all. In reality on the market we can't see anything: mass availability of the boards on the chipsets from Ali starts only in January, as well as the boards on the October announced Apollo Pro266 and AMD-760. Simultaneously announced KT266 will likely appear even more later.

But for the beginning the AMD-760 is enough, especially considering that the chipset turned to be successful. Although it doesn't have multiprocessor support - such opportunity will appear only in the AMD-760MP which can work now (judging by the demonstration by AMD at the last MicroProcessor Forum), but it was postponed until the second half of 2001 - for AMD it is still the first coming out onto the market of products for work stations and servers, the company wants it to be very smooth, so they prefer debugging it another half a year. Responsibility grows since the only company which had been going to make multiprocessor chipset for Athlon, HotRail, in May announced leaving this business and therefore stopping work on this chipset.

There can help only Micron. In 2001 the company decided to enter mass market of chipsets with some of their products. The Copperhead under the Pentium III, the Shogun and the Skimitar under the Athlon, and the Mamba with 8 MBytes integrated cache of the third level should become the second (and maybe the first!) server chipset under the Athlon.

To be continued...

Write a comment below. No registration needed!