iXBT Labs - Computer Hardware in Detail






September '2001 digest


In general, AMD began the month "not bad", having released its Duron 1 GHz based on the Morgan core for sale. Though it was hard to call it cheap as it had appeared for more than $100 in Japan. But this will things sort themselves with time, there's no doubt.

However, we are now interested not in real Duron/Morgan, but still nonexistent Athlon XP/Palomino to be released on October, 9. Is sad to realize that since the very long time AMD was compelled to use methods of those who can't follow up clock rates - instead of real megahertz users will be offered a system of ratings where 1,33 GHz processor will be referred to as Athlon XP 1500+, and 1.5 GHz - as Athlon XP 1800+.

No comments. With the release of Pentium 4 AMD has seriously lagged behind in both architecture, and clock rates of its processors. Now until Hammer AMD should join Apple proving that a megahertz is a megahertz, and speed is speed.

And Hammer will not be released soon as it is still planned to the second half of 2002 only, notwithstanding its today's samples. Though there are some doubts about it. Probably, AMD will "speed up" nevertheless and the release of Hammer will happen not in two years, but in the first half-year of 2002 (at least in samples). It would be rather logical.

However, today this situation can be somehow corrected by transition to 0.13 micron technical process - such Athlons with clock rates over 2 GHz will be released already in the first half of the next year simultaneously with 0.13 micron Durons.

It is especially necessary to hurry up taking into account that fortune avoided AMD last months together with the largest PC manufacturers: Micron, IBM, Gateway - all of them refused to use AMD processors in their PC lines. Certainly, market recession and the necessity of lines consolidation are reasons as well, but still the most probable is that it was not done without Intel's money.

Speaking about Intel it is necessary to mark that the company met the month with another price-cutting just as like AMD, however. It will be enough to quote prices just for a number of CPUs: P4 1,3 GHz - $112, AMD Thunderbird 900 MHz - $64; P4 2,0 GHz - $500, AMD Thunderbird 1,4 GHz - $123, Celeron 1,1 GHz - $105. It is amusing that the lowest Pentium 4 costs today only $7 more than the highest Celeron.

In such conditions Pentium III line obviously does not have a place to live as orders for this processor will be accepted by Intel till December, 7 only. Then there will be Pentium 4 and Celeron. By the way, it would be very interesting to look what will happen to prices on the release of Celeron 1.2 GHz at the end of October.

Speaking about mobile processors Intel wasn't noticed in September, but right at the beginning of the month it started to recruit developers for Banias to be released in 2003 and today just being a set of specifications for a future processor not even completely authorized. And speaking about mobile Pentium 4 for the first time it will appear for sale in February featuring 1.50-1.6 GHz clock rates.

Now only Intel's server processors remain to be mentioned. But the situation here is really and actually sad. We won't ever see 2 GHz Foster as a server processor due to great errors of increased processor cache mismatch at working in a similar mode. (Though it has been released as a variant for dual-processor workstations with 256 Kbytes of L2 cache.) So, actually a new architecture will enter the market in the beginning of 2002 only together with 0.13 micron Prestonia that at the moment of release will work at the same 2-2.2 GHz with its "own" Plumas chipset accordingly.

However, the release of McKinley samples will be the most interesting thing in this area for this quarter and its official release will happen together with i870 announcement. Closer to summer McKinley should become completely commercial taking a place in real servers. Actually the real life of IA64 will begin with this moment as well.

A society column features just one worthy event this month - in their struggle for P4X266 VIA and Intel have "exchanged" court actions with each other, doubting the legality of both Pentium 4, and VIA C3. Nevertheless, lawyers fight and managers sell, and all this didn't affect either Pentium 4, or C3 sales.

At the end of the month a "gift" variant of 800 MHz C3 in a tin can (like that old ones for sugar candies) came into the market. And this quarter should bring us samples of the following generation - Nehemiah with 1.3 GHz clock rate and 256 Kbyte L2 cache four time larger than C3's.

Only exotic remains now: PowerPC G5 from Motorola is to be released soon. This processor generally emphasizes clock rate like Pentium 4, and Motorola has a chance to keep the distance after Intel if not to reduce it. G5 is expected to have about 1.6 GHz clock rate. It is interesting to mark that FSB of this new processor will be 400 MHz. A chipset for it will support up to 16 Gbytes of DDR SDRAM, USB 2.0, and Bluetooth. Everything is up-to-date in general. Volume production will most likely begin in the first quarter.

And the second thing to mention here is the announcement of the latest generation of PA-RISC line coming to a logical end: PA-8700. This processor meets the latest technologies: copper, SOI, 2.25 Mbyte cache, however, taking into account that McKinley will come to the market shortly, and HP will start to actively advance systems on it, there's almost no doubt that PA-RISC like Hewlett-Packard's Alpha will end bad.


This section was the most eventful this month. And the general event was the announcement of i845 blockbuster chipset from Intel succeeding 440BX. There's no sense in describing its features as wholesale price of $42 is enough to make it the most expensive of all volume chipsets for Pentium 4 at the moment of announcement. (M1671 - $31, P4X266 - $34, SiS645 -$18-20.)

As usual, all motherboard manufacturers were very enthusiastic about the announcement from brand giants like Asus, MSI, Elitegroup, SuperMicro to tiny like PCPartner, or Totem. And enthusiasm of some was they nearly announced a following generation of this chipset with DDR SDRAM support. For example: Acorp's 4D845A, or PCPartner's 845AS4-237 and 845MS4-243. it's clear that today's i845 perfectly supports DDR SDRAM, but these are hardly boards on it, most likely marketing departments have just hurried and lifted the veil over information on boards on newly updated chipset. So, it's like we have to wait for the New Year, but not longer as warehouses will be filled with these boards awaiting sales. And now…

… Now Intel will try to brake other chipsets' appearance on the market that might be of any potential rivalry. And nobody will doubt that P4X266 is such a rival. Add to this traditionally "friendly" relations between Intel and VIA, and understand that it is no wonder that relations between companies started to meet the a two-year-old script - three days prior to the announcement of i845 Intel started a court action against VIA accusing it in the violation of five patents.

VIA reacted instantly and accused Intel in the violation of two patents at the day of i845 announcement, and even Pentium 4 became an object of accusation! Intel decided to punish such impudence and in two weeks started the second attack that affected not only VIA, but even its large clients like Elitegroup in particular. However the latter didn't experienced any difficulties hoping to make good money selling P4X266 boards especially if it enters the market first, and as VIA has promised to compensate all legal costs.

Though "first" may become doubtful as already at the beginning of the month Chaintech, Shuttle, and Tyan officially announced the deliveries of P4X266 boards having complete box packages instead of white packing. Elitegroup and ACorp joined them during VTF showing their boards on this chipset. And there was some information on similar products from AOpen, SUMA, ABIT, DFI… All of them unanimously confirming that boards will considerably surpass the speed of i845 boards with PC133 SDRAM support and thus will cost much cheaper.

And this quarter (in November) VIA starts volume production of P4M266 with the integrated Savage 4 core that was shown working already in the beginning of the month, and before the end of this year we'll witness deliveries of the next P4X333 generation with PC2700 support being much more attractive for long-term investments as PC2700 modules are incompatible with today's PC1600/PC2100, and pinout of tomorrow's DDR II modules will be compatible with PC2700.

In addition, deliveries of P4X266A pre-production models with FSB 533 MHz support for future Pentium 4 CPUs should start already now. So, personally I wouldn't advise to purchase P4X266 boards just in a very urgent case.

But these are plans for future, and today VIA's leader is KT266A (this chipset should have been named more seriously as it surpasses its predecessor in productivity very much), volume deliveries beginning in the middle of September simultaneously with announcements of boards on its basis by Soltek, ABIT, EPoX, AOpen, ASUS, Chaintech, SOYO, FIC, MSI, Shuttle, Gigabyte.

In general this chipset promises to be successful: preliminary tests show that it's productivity should be at the level of nForce 420-D unreleased yet, and even surpass it in some cases. And have in mind the short time before the release of KT333, and production of nForce boards will start only in December, and sales will be limited.

Full-scale attack on Socket-A continues, let's remember AMD as well, whose 760MPX is already at hand. November bring the production of Gigabyte's GA-7DPXDW - a complete dual-processor board for Athlon, with 64-bit PCI slots, AGP Pro slot, built-in IDE RAID from Promise; Asus prepares its A7M266-D; MSI - its K7D-MASTER-L… Taking into account that Intel didn't offered right dual-processor solutions of Low-End/Mid-End level for a very long time already, Athlon now has all chances to enter the market of dual-processor workstations.

SiS, certainly, does not pretend for this yet, but it has its own niche for successful work - SiS 735 motherboards are being manufactured at full speed, names of such September newbies as JetWay and Chaintech are enough, ABIT will join shortly as well. But SiS 735 is old, the peak of style is SiS 650 released at the end of the month. It's a chipset for Northwood with DDR SDRAM support (up to PC2700) and a built-in SiS 315 graphic core, moreover, $39 for chipset that will start to be produced at the end of October is the smallest price in comparison with same i845 (taking into account chipset's capabilities!).

Unfortunately, SiS slightly lags behind with Athlon chipsets. Only at the end of October SiS 740 and 745 will appear in volume amounts, the former is integrated, but supports PC2100 only, and the latter is classic, but supports IEEE 1394, and both bridges are on one chip worth just $29.

Speaking of ALi - it is limited to plans only. Generally, to plans concerning the following M1671 chipset for Pentium 4. FSB 400 MHz, PC2100, AGP, PCI… In general, having once sparkled with the announcement of the first chipset with DDR support, ALi moved to steady rear guard again. We won't seriously consider new M1535D+ southern bridge of this company with ATA/133 support added, will we?.

In general, as it was said above, this month was very rich with events. Even Micron (!) on VIA Technology Forum in the beginning of the month showed a chipset. This time it was already with PC2700 support, and the company as usually announced that it's not going to produce it, but mean it for "simplification of development and progress of new memory products" only. Like the Samurai in due time.

Though not everything's so easy with the latter… Copperhead (dual-processor Samurai) takes part in Intel's plans till present, and Gigabyte on World PC Expo'2001 at the end of the month already showed GA-6MXDR dual-processor Socket-370 board on the basis of this chipset! Two sockets for Tualatin, built-in UltraSCSI 320, two Ethernet controllers, four 64-bit PCI slots and one PCI-X slot, built-in ATI RAGE 128… Just buy and install in your server!


This month was rather quiet, but we'll mention a couple of interesting tendencies below. And, first, we'll look into memory prices. That, in general, adhered to the usual way this month: go south.

If in the beginning of September on TICE the cost of 128 MBit PC133 chip made $1.36, and just $0.72 for its 64 MBit colleague. At the end of the month these values changed to $1.20 and $0.60, having dropped for 12 and 16.5 percents for this month. Accordingly, market prices dropped for 12-15 percents. For example, 128 Mbytes dropped from $12.1 to $10.3, and 512 - from $49.5 to $43.25. A noticable, but rather weak reduction.

Among the main themes: most important is the overcoming of $1.30 limit for 128 MBits. The thing is Taiwan DRAM manufacturers promised to suspend deliveries when prices will drop below this level. Like it would be senseless to sell chips for these prices, and a general manufacture suspension will rise market prices again. Not at all: prices were attached to this barrier during all month, but when they dropped right at the end, there was no apocalypse - warehouses were full, manufacture should be stimulated, sales were minimal, so, it was like take what you are given.

The second important thing is actual and official death of PC100. PC66 disappeared from markets a long time ago, and at the end of September the same TICE officially stopped quoting PC100, having removed it from auction. Time's not to stop, PC100 has really come to its end. But this is not about 64 Mbit chips as they will continue to make serious market profits at least until 2002.

But let's get back to DDR SDRAM prices this time. As usually Crucial remains the most noticable indicator, having pushed prices for PC2100 modules to $19.8, and to $33.3 for 128 and 256 Mbyte modules at the end of the month, and it is necessary to mark that prices for 256 Mbytes drop much faster. In the beginning of the month the price difference between was more than double!

However, such dumping had rather straight effect on Micron's activity results. No one in their office was glad to sum up the quarter finished on August, 30: 575.5 million dollars loss in comparison with 726 million profit of the last year. Sale volumes from the angle of finance dropped from 2.31 billion to 480.3 million this time, but Micron's market share hardly narrowed this time, it might have even got wider: one should blame the general reduction of market prices along with the furious dumping of Micron itself.

However, Micron feels just great in comparison with Hynix - a new hero of the month that replaced memorable Rambus. Huge debts, lasting losses due to general market conditions, close period for 4 billion debt payout… Despite active protests of the same Micron, brought even to WTO, the government of South Korea added this "Titanic" 390 million dollars through controlled banks. In addition, a new credit line for 1.17 billion dollars was opened for the company that allow it to delay funerals, and to engage in restructuring and today's business.

Such as releasing 0.13 micron 512 MBit PC2100 DDR chips that fully meets the favourite tactics of memory manufacturers - to release niche solutions for server or mobile markets where prices for products of almost the same prime costs are higher than on the volume market.

Japanese Toshiba, another suffering manufacturer, frankly announced the intention to completely leave the DRAM manufacture business, having sold the appropriate division to an interested party. Unfortunately, September didn't brought one: Elpida decided to not deal with it after thinking it over, Samsung hesitated because of its own problems, and the third potential buyer, Infineon, has not gone beyond negotiations yet. "Yet" as there's no doubt that sooner or later Toshiba's business will be sold to either Infineon or Samsung, buyer having 51% of shares only.

And now Samsung like Hynix seeks niches to save itself. In particular, having started to release economic (1/10 power consumption of standard) 256 MBit SDRAM chips for IMT-2000 phones, at the end of the month having announced the deliveries of 1 GHz 576 MBit RDRAM chips that, by the way, despite all Rambus's problems is the wise decision, taking into account some factors leading to Pentium 4, however.

These are: popularity of this processor, and increase of i850+RDRAM bundle sale volumes due to this, and the fact that 4-bank RDRAM still seriously occupies Intel's plans - Pentium 4 strictly requires two channels for normal perfomance, and RDRAM is the cheapest variant in case of two-channel circuit. It's a question of chipset and motherboard realization though RDRAM itself becomes rather inexpensive in case of 0.12 micron technical process used by Samsung.

By the way, more about Rambus. Its business are rather interesting. Both good, and bad at once. On the one hand, in the beginning of the month all was so bad that the company asked federal court to move the trial of its action to Micron for more than a year to 2003, when its appeal against the decision of judge Payne who acquited Infineon will be considered.

And those good news: in the middle of the month Intel and Rambus revised their relations: Rambus allowed Intel to release i845 with DDR SDRAM support in the beginning of 2002 without any sanctions and, in general, to release any chipsets supporting this architecture in the future. In exchange Intel has engaged in paying the company up to 10 million dollars quarterly within five years that makes sum about 200 million dollars total. In addition, Intel seriously considers the question of embedding of RDRAM interface in its network processors. Probably, this aspect is included in the agreement as well.

So, now Intel actively exploiting this freedom is variously engaged in DDR progress, having added 22 pages of amendments to PC2100 specification guaranteeing stable work of appropriate modules with their DDR chipsets, so, already the first i845 will support not only DDR PC1600 as it was promised initially, but also PC2100 that is far more pleasant.

And in May JEDEC itself adopted PC2700 chip specification, so in September SMART announced the release of 184-pin modules on their basis with clear conscience. Prototypes of chips for prototypes of modules were presented by Nanya nurturing dreams of DDR market leadership till present.


It's fun that despite ATI was the hero of the month the majority of news was about NVIDIA. That is not very surprising, however, even without Titanium released on October, 1 (next month this section will be far longer!).

In September NVIDIA itself did nothing in general, but was as usually noticed in connection with actions of its numerous partners among whom only Creative sparkled something out of the ordinary deciding to exclusively engage in manufacturing and distribution of Personal Cinema - a remotely controlled solution for recording and playing back TV broadcasts - in Europe. Ah! Taiwan doesn't care a straw about it, and we'll get cards on the basis of MX400 for $100-120 in the near future already.

And others… Each in his own way: GF2 Pro cards from Gainward and AOpen overclocked to the limit, a whole line of standard cards from eVGA - from TNT2 M64 to GeForce3, 64 Mbyte MX400 card from MSI, and at once two cards from UNIKA and AOpen with 128 Mbytes worth about $100-110. A product for videocard fans.

ATI, having begun to sell cards already in the middle of September, remained the "interest leader" (it's clear that ATI gave up the habit of making half-year gaps between announcements and market sales, as the company was obviously moving along its technological schedule). The provisional price of Radeon 8500 was varying between $290 and $349, and the price of 7500 made about $189 that is already much lower than initial prices announced by ATI.

But it was only the first pleasant surprise (or two, taking into account the speed of cards' appearance for sale). The next one were news that speed of Radeon 8500 will be increased to 275/275 MHz in October; and then ATI's partners helped with prices again. In the middle of October Power Magic announced the deliveries of Radeon 8500 and 7500 cards (and also a great lot of Radeon VE models) without prices, however specified by Raising Rich: these products will cost about $190 for Radeon 8500 with 64 Mbytes of DDR, and just $125 for 7500.

Bravo, ATI! Such a fruitful return to sources (namely Taiwan manufacturers): the buyer should not pass such offers by. At least, there's no doubt that already profitable company will get stronger in the near future.

Speaking of other market participants - they were generally silent this month. It is worth to just mention the release of three new boards on Kyro and Kyro II by the same Hercules, and also Gainward's sales of SiS 315 boards - 32 Mbyte cards with TV-output cost about $50.

Storage devices

As usually, CD-RW rules the section. Like NVIDIA's leadership in the Video section independent to its activity.

Omitting new alike 16/10/40 models, we shall emphasize just three things worth of attention. Speaking about usual "classic" CD-RWs - new 24/10/40 drives are entering the market, both internal, and external with USB 2.0 interface - from Sanyo, Ricoh, Yamaha, Acer. And lower manufacturers won't allow to overestimate prices, dumping them to $150 level at once.

The second group consists of various tiny external CD-R/CD-RW drives, capable of being used as independent MP3-CD players. No sense in listing all manufacturers that announced similar models this month, it's enough that the average speed formula varies quite widely from 8/4/24 to 12/8/24X. The average price is about $300. Sales of both the former, and the latter group start in October.

The third group is made of various DVD drives without any "normal" one to appear in September. Either NEC presents an ultra-thin DVD/CD-RW drive for ultra-light notebooks, or Panasonic starts volume sales of external DVD-RAM/R drives with IEEE 1394 interface.

As an "optical drive highlight" we shall mark TDK's MLCDRW1000 on World PC Expo'2001 - the first drive with ML (MultiLevel Recording) disks support with an opportunity of recording and re-recording, and also with the capability of performing the same operations on standard CD-R/CD-RW disks though at a rather average 12/10/40 level, whereas ML disks are operated at the possible maximum of 36/30/40X. Not mentioning DVD read support. Unfortunately, it will be released only in February.

To finish this theme, we'll just mention that at the end of September Philips announced Orange Book part II Vol2 v1.1 specifying 32x CD-R disks that enabled CD-R manufacturers to immediately start their production to throw them out on markets in October. Drives designed for this record speed will appear prior to winter, and 40X ones are at hand as well - these disks are expected in the second quarter.

HDDs are traditionally bad, i.e. few. Still these few are good. Well, the release of Samsung's inexpensive and silent SpinPoint V40 might be omitted, though why should we omit 20-80 GByte HDDs with just 3 bel noise level of active work mode, taking into account the recent considerable progress of Samsung in increase of this production's reliability.

But the second and last chance of this section to draw attention to itself this month is the release of DiamondMax Plus D540X 160 GByte ATA HDD. Right you are - at ATA's limit of 137 GBytes. But BigDrives with 48-bit address space being one of two ATA's latest updates allows to overcome this barrier. Fans of HDD video storage may enjoy quite a large volume along with storage ratio of 0,04 cents per megabyte - this 160 GByte drive costs $400.

But the speed of above HDD is 5,400 rpm, so another Maxtor's drive - D740X - looks little more balanced. On the one hand - just 80 GBytes, but 7,200 rpm and ATA/133 on the other. Though the price for one megabyte is a little higher as 80 GByte model costs $250.

Write a comment below. No registration needed!

Article navigation:

blog comments powered by Disqus

  Most Popular Reviews More    RSS  

AMD Phenom II X4 955, Phenom II X4 960T, Phenom II X6 1075T, and Intel Pentium G2120, Core i3-3220, Core i5-3330 Processors

Comparing old, cheap solutions from AMD with new, budget offerings from Intel.
February 1, 2013 · Processor Roundups

Inno3D GeForce GTX 670 iChill, Inno3D GeForce GTX 660 Ti Graphics Cards

A couple of mid-range adapters with original cooling systems.
January 30, 2013 · Video cards: NVIDIA GPUs

Creative Sound Blaster X-Fi Surround 5.1

An external X-Fi solution in tests.
September 9, 2008 · Sound Cards

AMD FX-8350 Processor

The first worthwhile Piledriver CPU.
September 11, 2012 · Processors: AMD

Consumed Power, Energy Consumption: Ivy Bridge vs. Sandy Bridge

Trying out the new method.
September 18, 2012 · Processors: Intel
  Latest Reviews More    RSS  

i3DSpeed, September 2013

Retested all graphics cards with the new drivers.
Oct 18, 2013 · 3Digests

i3DSpeed, August 2013

Added new benchmarks: BioShock Infinite and Metro: Last Light.
Sep 06, 2013 · 3Digests

i3DSpeed, July 2013

Added the test results of NVIDIA GeForce GTX 760 and AMD Radeon HD 7730.
Aug 05, 2013 · 3Digests

Gainward GeForce GTX 650 Ti BOOST 2GB Golden Sample Graphics Card

An excellent hybrid of GeForce GTX 650 Ti and GeForce GTX 660.
Jun 24, 2013 · Video cards: NVIDIA GPUs

i3DSpeed, May 2013

Added the test results of NVIDIA GeForce GTX 770/780.
Jun 03, 2013 · 3Digests
  Latest News More    RSS  

Platform  ·  Video  ·  Multimedia  ·  Mobile  ·  Other  ||  About us & Privacy policy  ·  Twitter  ·  Facebook

Copyright © Byrds Research & Publishing, Ltd., 1997–2011. All rights reserved.