iXBT Labs - Computer Hardware in Detail






2000 Hardware Digest
(part II)

January 4, 2001


The main subject of the year in this field is opposition of the Direct RDRAM and the DDR SDRAM. Such wording is not rather correct, though. We'd rather talk about Rambus activity and DDR grow.

The Rambus company attracted attention not from the technological side, but from juridical one. Everything started right in January when the company announced their claims for possession of patent pool for the DDR SDRAM/SGRAM, PC100, PC133 SDRAM, requiring license payments from all manufacturers who use this specification. There of course no any people to agree, that's why after some estimations they chose Hitachi to claim to.

After that the conflict was growing on too quickly: in March they brought an action in the Federal Trade Committee making demands to ban an import of Hitachi chips and products on their base into the USA (e.g., game stations Sega Dreamcast), in April - the same thing in Germany. Hitachi counter-claimed, but taking into account interests of their clients, e.g. Sega, they had to admit Rambus patents and to license all the aforementioned technologies. Especially considering that a few weeks earlier Toshiba did it. After that came Oki Electric, Mitsubishi, NEC, Samsung. In the very end of the year Elpida Memory announced licensing - a joint company of NEC and Hitachi, they are becoming one of the largest company in the world in memory production.

So, Rambus now can have an easy time. Commissions from Samsung and Elpida must be enough for a small company to exist. Though, they are only two out of five that agreed to pay. The other three - Hyundai, Micron, Infineon, on the contrary, accused Rambus according to the antimonopoly law and of violation of intellectual property. Rambus's response was the same counter-claims. What will happen further we can find out in 2001 with the court examinations.

Now we can only say that a vivid activity of Rambus has yielded fruit: in January they announced starting developments of the Advanced DRAM Technology that is based on the same principles as the RDRAM, but it is more open, cheap and efficient. Plus, do not forget the DDR II. In general, we should admit that Rambus has already lost in the market field. Even Intel don't doubt about it, yet at the spring IDF'2000, in February, it announced that they were going to move 100 percent middle level PCs to the RDRAM by the end of 2001. At that time they didn't yet spoke about the Brookdale.

However, in spite of such an activity of Rambus, the RDRAM was still developing - what was left for Samsung who had invested a lot of money for it. Besides, it was unclear what to expect from customers, that's why nobody decided to refuse that memory. And it is no surprise that many greatest manufacturers were going on developing that direction: in 2000 many of them created 2-3 generation of the 144/128 and 288/256 Mbit chips of the RDRAM, what for the most of them meant 0.15-0.17 micron technological process. And even Hyundai was producing that type of memory having received quite attractive terms from Rambus.

Besides, despite a simple move to a smaller geometry the company tried to decrease the prime price of the RDRAM by other methods - from new methods of packing to a simple reduction of requirements to the boards for the RIMM modules, like Intel that decreased the cost at the expense of the quality. It of course bore fruit - say, Kingston, the largest manufacturer of RIMM modules, reduced the price by 35% in average explaining it by decreasing the price of Samsung chips.

However, it was not enough to come out into the mass market: in the middle of the year Intel hold a meeting with the DRAM manufacturers trying to persuade them to decrease the price for RDRAM chips, let it be even at the expense of decrease in income. It was useless. Moreover, the largest manufacturers of RDRAM started to return a part of the lines for SDRAM production, since it was bringing more income at that time and they were expected increase in price in the second half of the year.

But that was a mistake. That's right, price increase started yet in the end of the winter, by the middle of the year the price for the 64 Mbit PC100 exceeded $7, and for the 3-4 quarters the manufacturers expected it to be $9-10. Well, first, the prices were in accordance with the forecast reaching $8.5 by end of June. During July the prices for the PC133 modules increased by 10%, Micron set for their 64 Mbit chips a contract price in $8, though a few months ago it was only $6.

In August the prices stopped growing and by the end they even fell a bit, though the manufacturers were still promising shortage of the 64 Mbit chips in the 4 quarter. And in September we saw sheer fall in prices by more than $2 reaching approx $6 per chip. And again, the DRAM manufacturers said that in October the prices should go up. But neither in October nor in November they were going to grow: in November the price constituted less then $4 per 64 Mbit chip SDRAM.

The matter is that this summer the market took a wait-and-see attitude: usually in summer the prices are falling, but this time we saw nothing alike - they were very stable and then it was understood that in autumn with the splash in demand the prices should go up. But in September no price leap could be seen (due to different reasons - and the general decrease of demand for PC, and differed demand as a result of expecting DDR SDRAM), the stock houses turned to be overcrowded.

But this time the memory manufacturers went easier through it than it was in the previous year. During the first half of the year they at least managed to move to 0.18 micron (what gives around 40% decrease in the prime cost as compared with the 0.25 micron), not mention such Titans as Micron which was already producing 0.14 micron DRAM chips, that's why this time $3 per chip appeared higher than the level of the prime cost for the majority.

But into the picture comes a 0.13 micron process - Mosel licensed it from Infineon, Winbond with Toshiba announced a conclusion of an alliance in order to develop 0.13 micron DRAM chips, a bit later, together with Fujitsu, starting developing 0.11 micron chips. But Samsung appeared to be in front, they announced a 0.12 micron 512 Mbit chip of the same size as a standard 256 Mbit, and 0.10 micron technology, that had been already announced by the company in April, should ensure the first place in the race for 1 Gbit DRAM chips. However, they are capable to do even more - at the ISSCC Samsung demonstrated 4 Gbit DDR SDRAM chips made on the base of this technology. According to the company, it will be the first case in the world when 4 billion transistors will be made on one chip.

But it's not a single method. There is for example a transition to 300 mm wafers. For example, in this year Infineon on their pilot line of DRAM chip production out from 300 mm silicon wafers reached 90% valid output for 64 Mbit chips - a level of a good commercial production. And such a transition is another 30% decrease in the prime cost as compared with usage of standard 200 mm wafers. So, considering that the first 300 mm factory for DRAM (0.13 micron) chip production will be open by Elpida Memory in the nearest years, and then they will be coming more and more often (e.g. Winbond announced investing 7.5 billion USD by 2008 in building of two such works), we can expect decrease in the prime cost of the DRAM chips by more than twice during 3-4 years.

Especially considering that by that time 64 Mbit chips won't be produced anymore. This year the Koreans increased production volume of the 128 Mbit chips by 50% , a bit decreasing production volume of the 64 Mbit chips. Nevertheless, 64 Mbit chips today are still leading. But, if considering memory size… Besides, it is Micron and not the Koreans that sets the mood with their 43% of the 128 Mbit SDRAM chip market.

But in any case, first, in 2003-2004 we will use not 128 Mbit chip modules but 256 Mbit, and secondly, with the time production of one 128 Mbit chips will get cheaper than two 64 Mbit etc., what gives another factor of price decreasing. Plus new technologies.

But all these will come together only in 3-4 years when we will see at the same time and 300-mm, and 0.13 micron, but until that time the demand will be growing faster than the supply. The only advantage there will be moving to production of chips with higher density, what is possible now - in the beginning of the year Smart Modular offered the first PC133 SDRAM module on 512 Mbit chips with 2 GBytes capacity.

Production of 256 Mbit chips DDR SDRAM (on a par with 128 Mbit) was started by the majority of DRAM manufacturers. However, 256 Mbit were seen mainly as patterns, but by the end of the year samples of the modules on their base have already started shipping. But not all companies started mass production of DDR SDRAM. Taiwan will begin mass production only in Q2 2001.

It seems that PC1600 won't live long on the market, and it will be PC2100 to become a standard. Micron promises to start delivering DIMM PC2600 modules in Q2 2001, as well as Samsung. JEDEC has already been discussing PC3200 specification half a year.

We should note a perspective direction in the memory field - nonvolatile memory on the base of magnetic field effects. There are a lot of companies working out there that reached great results. The working prototypes were already shown by Micron (SyncFlash), IBM (MRAM - Magnetic Random Access Memory) and Motorola (MRAM - Magnetoresistive Random Access Memory).

The SyncFlash is being used in commercial products. In the end of the year Philips released a series of chips for game stations supporting this interface. IBM and Motorola have already reached great success here, so we are expecting end products from them as well.


In this year NVIDIA is keeping their rash grow shipping their products exactly on time. 3dfx struggled against failures. ATI was late with the release of their products, S3 declined their road map...

In the beginning of the year we had practically a complete set: spring generation cards continued being produced, and this means GeForce 256, and Viper II on Savage 2000, and ATI Fury MAXX. 3dfx have already obtained the first sample of the VSA-100. But it turned to be a slow 100 MHz chip of A0 version, but it was enough to demonstrate at the CeBit a slide-show allowing to estimate the image quality.

At the CeBit you could see a prototype of the card on the KYRO. With the time it turned out that Videologic and STM took the way of ATI: an interesting chip equal to the competitors in speed but with the higher price. Plus, a weak support on the drivers' level. But the market preferred the GeForce 2 MX to the KYRO.

Now some words on Matrox. The company didn't demonstrate any game solutions. However, the solution for video editing RTL2000 on the G400 don't have analogs in most cases, a series of business Millennium G450, and a successful Marvel G450. The company showed niche solutions. It seems that on the market where you can find today's NVIDIA this is the only right solution.

If 3dfx had released VoodooTV series in the beginning of the year and had announced a couple of similar products, the company would have lived up to the Rampage release. It's of course very interesting to watch the events if 3dfx hadn't bought the GigaPixel. On the one hand, it's necessary to develop, on the other hand, without giving out the money the company could have released the Rampage.

3dfx pinned hopes also to the VSA-100, but the chip to be a serious competitor against the GF256 was too late. Moreover, it appeared even later than the GeForce 2 GTS - the cards on it started appearing in April. The Voodoo5 5500 appeared on the market only in June.

The Voodoo5 6000 was to become a super-product in the assortment of 3dfx, but it was developing so long that NVIDIA managed t release at that time the GF2 Ultra which ensures the performance of the same level in games but costs less. So, the result is that the famous manufacturer is pulled out of the market.

S3 went through it even earlier than 3dfx. They sold their graphics department VIA in the end of April.

And remember Number Nine that announced their shutdown, and NeoMagic that went from the mobile graphics market leaving their sector for monopolistic use of ATI (and now for NVIDIA with their GeForce2 Go).

In fact these two companies were leaders during the whole year. And the leadership concerns NVIDIA more than the other. During the year the company managed to released 4 versions of solutions for different price niches (not taking into account the combination SDR/DDR) - GTS, MX, Pro and Ultra. (Plus two professional solutions from the Quadro 2 series).

The keen manufacturers sometimes release funny products: the 16 MBytes MX from Asus or the 64 MBytes MX from PowerColor! Or even ASUS V7100 Deluxe Combo appeared in the end.

By the way, the market of the devices converting PC into a digital VCR in 2000 became one of the fastest developing: NVIDIA and 3dfx (VoodooTV) joined ATI and Matrox.

Undoubtedly, ATI decided to support their reputation in this field and released an extension of the All-in-Wonder series on their new chip Radeon. Unfortunately, the chip didn't appear in time. If it had happened in summer at the same time with the GeForce2 GTS, the state of affairs of ATI might have taken another way. The chip Radeon is just excellent but it's poor realized by ATI programmers.

As a result they got unprofitable quarters. However, the company has a traditionally strong direction (without competitors): the Mac, where the company turned to be a monopolist after 3dx had left (despite the fact that NVIDIA wants to fight in this sector), the market of business solutions (the Radeon VE is absolutely non-game product) and graphics for notebooks. By the way, in October ATI became the first company demonstrating the AGP 4X mobile chip - Mobility M4 which started winning in OEM designs.

Storage devices

The whole year is a non-stop evolution without any shakes. At least, at the level of products. As for technologies the year was quite serious, but they ought to be realized.

The year started from 7,200 rpm of ATA/66 HDD on the 10 GBytes discs with up to 30 GBytes capacity, such as Seagate Barracuda ATA II or Quantum Fireball Plus LM. And it ended up with 7,200 rpm of ATA/100 60 GBytes Maxtor DiamondMax Plus 60 or IBM Deskstar 60GXP on the 20 GBytes discs. (The record was set by Maxtor with their DiamondMax 80 (5,400 rpm!)).

This led to twice HDD volume increase. The speed 10,000 rpm remained to belong to the High-End models; transition to the ATA/100 from the ATA/66 isn't worth mentioning.

Let's consider the High-End. The beginning of the year - the Quantum Atlas V with the maximum size of 36 GBytes and 7,200 rpm, the end of the year - 7,200 rpm 180 GBytes Seagate Barracuda 180 and the 15,000 rpm 36 GBytes Cheetah X15. That means the volume increase five times (!), speed increase - 1.5 times. Moreover, interface has also changed here: the Seagate demonstrated in the end of the year a HDD with Ultra320 SCSI interface, and Adaptec announced the first controller FibreChannel 2 Gbit, which gives the maximum bandwidth in 400 mbps in duplex mode.

But the more realistic product in the beginning of the year will be the 73.4 GBytes HDD with 10,000 rpm and Ultra160 SCSI interface.

By the way, the tendency of double increase HDD size a year completely confirms to IBM observations; and the company assumes that by 2004 the 150 GBytes HDD should become ordinary for PCs. Besides, applying this rule again, by the end of 2001 we will have maximum size of HDD 120-160 GBytes.

According to forecasts of the largest HDD manufacturers the speed 15,000 rpm should become a standard for servers during 2001-2002, and for user machines it will be 10,000 rpm.

Since we have intensive experiments now in this field, let's see what is there going on:

  • For the last year the record in writing density has grown from 45 GBytes per inch squared to 60 GBytes per inch squared, i.e. the maximum potential size of a standard 3.5'' disc increased from 63 to 80 GBytes - twice higher that's required for the nearest year. According to the forecasts of manufactures, 100 GBytes per inch squared for commercial use will be reached by 2003. Today the record density in average consumer HDD constitutes around 15 GBytes per inch squared: 15 GBytes by the end of 2000, 30 GBytes - 2001, 60 GBytes - 2002, 120 GBytes - 2003.
  • 300 GBytes per inch squared is considered the potential for such technologies.
  • Read-Rite announced reaching 1 GBytes data rate with the usage of their universal write/read GMR heads, and TDK created a magnetic head allowing to reach 5 times higher data rate as compared with today's analogs.
  • The SerialATA, which was announced to start developing at the IDF Spring, by the end of the year reached a test version 1.0. However, we will hardly see in 2001 this interface in mass PCs. But in 2002 we will enjoy 1.5 Gbps 1X variant of this interface. And in front there is still 2X and 4X. Now everything depends on chipset manufacturers.
  • The noise-suppressing technologies are of vital importance too. All manufacturers try to contribute in noise decrease. This became a reason for a new extension of ATA specification - the Automatic Acoustic Management (AAM) which allows HDD to switch a low working mode decreasing a bit a search speed. And the AAM has already found a use in new models - remember the U5 from Seagate.

As for companies themselves, we should note a foundation of two largest manufacturers. First, Komag bought HMT Technology at 125 million USD, what must develop into the greatest manufacturer of discs for the corresponding storage devices, costing about 0.5 billion dollars and working with such companies as Western Digital, Maxtor, Seagate, Samsung, Conner, Iomega.

And secondly, pooling of Maxtor and Quanutm making the largest HDD manufacturer. According to the official data such a giant will produce around 50 million HDDs a year.

Now let's move to optical discs. Here we can see the same evolutional development. Exactly one generation of CD-RW drives has changed - at the Cebit there were prototypes of the 12/8/32X drives coming into production in summer, at the Comdex'Fall you could saw prototypes of the 16/10/40X drives coming out on the market in winter. But it could have been a little better: Cirrus controller ensures read speed up to 48X, and Elantec Semiconductor - 20X write/rewrite. But isn't worth it? 10X CD-RW discs appeared only in the end of the year.

There appeared two interesting technologies in this field:

  • Calimetrics licensed from Polaroid a new optical technology allowing to increase a volume of storage devices based on light reflection effect (CD, DVD) by 2.5-3 times. And if adding their proprietary multilevel recording technology, it will yield up to 5 GBytes per CD-RW disc and up to 25 GBytes per DVD-RAM. By the way, in the end of the year TDK announced support for MultiLevel Recording. Today this technology allows recording CD-RW discs up to 2 GBytes at 36X, and it's necessary to add only one little chip to the today's CD-RW drives.
  • Sony and Philips announced the Double Density CD-ROM/-R/-RW specification, that twice increases the size of these discs from standard 650 MBytes to 1.3 GBytes. The back compatibility works there as well.

Besides, there is the FMD from Constellation 3D which has already demonstrated a prototype of a video player on the base of their 10-layer disc with capacity up to 50 GBytes. FMD drives on the base of a red laser must read today's CD discs.

There are as well miniature 500 MBytes optical discs from DataPlay, a coin in size. The price should be $5-10 per disc. Such companies as S3 and Toshiba are interested in developing this technology.

The 3DR (three dimensional recording) from Siros Technologies provides writing and reading of multilayer optical discs, up to 18 GBytes per layer (36 GBytes totally). This was successfully implemented in practice.

Now about DVD. You should admit that DVD+RW was again beaten by DVD-RAM. Again we saw nothing except promises and prototypes. Yet at the Cebit there were shown recording and playback of video in real time mode on a prototype of a drive from Ricoh, and the next date of coming out in the market was planned for the end of 2000. Where is DVD+RW? Nowhere. At the Comdex there were only three prototypes of drives from HP, Philips ands Ricoh, with promises of conquering the market in 2001.

The DVD-RAM camp has demonstrated the second generation of DVD-RAM discs housing 9.4 GBytes. And in the second half of the year they shipped the real products: quite available drives, video cameras using these discs for recording.

An active work is carried out on increasing the size of DVD discs: in September Hitachi announced a new technology (a combine of a new laser and CDs), in theory it allows reaching 5.25'' CD discs 100 GBytes in size, and Sony has reached more modest results, but in practice - in October the company demonstrated a prototype of their drive on the base of GaN laser that is capable to work with DVD discs housing 22.5 GBytes per one side. This is only a prototype - potentially we are talking about discs with the size more than a hundred GBytes.

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.