by Andy Yaschenko
04/21/2004 | 09:34 AM
The hottest news of March was surely Intel’s voiced intention to use a new rating-based system for marking their processors. The three-digit index will depend on the frequency of the chip, but also on other parameters that affect performance in some way (the FSB frequency, the cache size, support of various technologies and instructions and so on). By the latest unofficial info, the existing Pentium 4 will get a rating from 520 to 760 points where 5xx are reserved for the ordinary Pentium and 7xx for its Extreme Edition. The Celeron with its 2.53-3.33GHz frequencies will occupy a small stretch from 325 to 350.
<%BANNER[article]%>This news and the new rating system are definitely good. At least, the proposed system is more articulate than the spontaneously-appearing suffixes behind the frequency (A, B, C…). Today we have four versions of the Pentium 4 2.8GHz (one model has no suffix at all and three others have those alphabetical markings). Theoretically, this approach could lead to a fifth version with the same frequency (EE for Extreme Edition).
As a result, Intel chose to borrow the well-known and proven labeling system from the manufacturers of cars, planes, trains, TV-sets, Opteron processors and so on. Customers from other industries don’t complain about it – there’s no inconveniency for them. The rating system will start to be implemented since the Pentium M and Intel’s mobile processors will have their own independent numbering system: 3xx for the Celeron M, 5xx for the Mobile Pentium 4, and 7xx for the Pentium M.
By the way, don’t you recall the indignation of some critics about AMD’s denouncing the megahertz and offering their new rating system some time ago? Now, Intel found itself in a situation when they could double the L2 cache in the 90nm Prescott and Dothan processors compared to 130nm Northwood and Banias but couldn’t push the clock rate up that high. They have now to explain the customer that the frequency alone is not the only thing that matters.
Intel isn’t only borrowing, it’s also loaning ideas. Thus, AMD intends to implement the SSE3 instruction set into its future processors – SSE3 has proved to be really useful for games and multimedia applications. After that, Intel will only have one ace up its sleeve – Hyper-Threading – since AMD doesn’t want to take it up. It’s too difficult to implement this technology in the 64-bit CPUs AMD is now manufacturing.
As for the delay with DDR2 support, AMD is going to make up for it no sooner than the second half of the next year. So far, they have limited themselves to using memory modules on DDR500 chips. AMD manifests no desire to support the upcoming BTX form-factor, proposed by Intel. AMD says the heat dissipation of their processors fits perfectly into the ATX framework and will fit in the near future due to the transition to the 90nm tech process, scheduled for the second half of this very April, although mass production of such chips will only start in June.
In March, we heard the announcement of the highest-performing x86 processor of today, the Athlon 64 FX-53. Some people around the world could even see it in retail stores.
And without any announcements, the Athlon 64 2800+ appeared on pricelists, with 1.8GHz frequency, 512KB cache, and an official price of only $178! It is for the Socket 754, of course, since the Socket 939 is delayed again. They set the launch of the Socket 939 Athlon 64 CPU to the end of May as yet, so we’ll hardly see many such processors in shops in the Q2. At least, in what may be called “mass quantities”.
On the other side of the battlefield, Intel doesn’t feel too well, too. By the latest rumors, the LGA775 platform is postponed to June. Processors and mainboards for/with this socket are expected in this month, although mainboards are already sampling. Processors will probably be late, as usual.
There are many versions explaining the reasons for the delay, including the notorious problems with the 90nm tech process. At least, the 3.4GHz Prescott, announced back on the 2nd of February, won’t be here until April. As a kind of compensation, Intel rolled out the 2.4GHz model priced very moderately, something like $120. Such chips may be defective 3.4GHz models, by the way.
Overall, Intel is busy making things right with manufacture. Besides the 90nm process, the 130nm Pentium 4 EE and the Xeon processor transition from six metallization layers to eight, probably to permit a higher chip yield. Meanwhile, a 65nm line is being prepared at the experimental D1D – Intel traditionally remains the leader in mastering new tech processes.
The Xeon also received some attention in March. The 90nm Nocona core with 64-bit addressing (the technology is likely to be named EM64T), also in the LV version, with frequencies up to 2.8GHz, appears by the end of this year. The Pentium 4 with those 64 bits is unlikely to show up soon. The technology will most likely appear in the Xeon first, which family welcomed new members in April, including the 3GHz Xeon MP with 4MB of L3 cache.
Getting done with the big two, I’d like to mention the agreement of Intel and Intergraph as they ended their six-year patent war. Intergraph won the day completely with trophies amounting to $675 million.
Transmeta, on the contrary, keeps low and doesn’t sue anybody. The company gets closer to its client, NEC, licensing technologies for reducing the chip’s power consumption. Now NEC will use this knowledge in its 45-90nm tech processes and Transmeta will have its licensing fees – every extra cent counts for them!
Gigabyte is the hero of the last month in this market. First, the company unveiled the Gigabyte RZ product series, the answer to the ASUS X-Series (the companies both use the same approach offering cheaper versions of already produced mainboards), then at the very end of the month they agreed with Foxconn to mass-produce such mainboards. Gigabyte also announced the opening of their new fab in the mainland China. The company has already climbed to the world’s third place in sales volumes and if it keeps on the tempo, it will have a chance to challenge Elitegroup for the second.
Interestingly, Elitegroup made quite an opposite move in March, offering the Extreme mainboard series, positioned as high-end. That’s logical to expand into a non-occupied-yet territory, but on the other hand, Elitegroup will have a hard time trying to overcome its image of a low-end manufacturer. Image is a powerful thing.
CeBIT was a place where one could improve on the image of the company. Models on the i915/i925 chipsets constituted a majority of the presented mainboards and many of them were for the LGA775 socket, including (quite unexpectedly) LGA775 models on the i865PE, which may be the most optimal variant for today. On the other hand, the mixture of the i915 with ordinary DDR SDRAM looks appealing, too.
What’s funny, it transpired after CeBIT that Intel will probably shift the start of shipments of LGA755 processors to June-July, so all the glorious things exposed at CeBIT are likely to remain just showcase samples for the next quarter. So far we can see only E7210-based mainboards appearing in stores (the i875P with a PCI-X-supporting South Bridge), while the prices of the i845GL/GV, 848P and 865G are going down by 1-2 dollars, as usual. A similar price reduction of their family members is expected by the start of the summer, before the launch of the new series.
This pause of course helps VIA and SiS who can now prepare their counterarguments better: PT890 and SiS 656. These two were also present at CeBIT. The SiS 656 was officially announced there, thus becoming formally the first chipset with support of the PCI Express bus. VIA only unveiled two new integrated chipsets before CeBIT, PM800 and PM880, although it’s not quite clear why should they boost the performance of the solution as concerns the system and memory busses, but leave the ancient graphics core that doesn’t need such power!
Mainboards on the PT890 were more adequate offers, as this chipset can support both DDR and DDR2 SDRAM, PCI Express and AGP 8x interfaces. The CN400 chipset for VIA’s own C3 processors was announced where integrated graphics was more to the point than in Pentium 4 systems. Lastly, they also released a new miniature mainboard on this chipset, the Nano-ITX form-factor EPIA-N. On 12x12cm area and with passive cooling, it has everything to build a full-fledged computer. Now VIA seems to have finally found its niche!
VIA didn’t forget about another direction where it’s been traditionally strong, although less now than before – platforms for AMD processors. The K8T800 Pro, much emphasized at CeBIT, is for them, differing from its “non-Pro” mate in the support of the 1GHz HyperTransport bus, in addition to 800MHz. That’s going to be the way the Athlon 64 will develop in this year: a 25% increase of the bus clock rate here and a higher frequency of the supported DDR SDRAM there. Plus a new socket – the K8T800 Pro supports the Socket 939.
The SiS 756 chipset, announced at CeBIT, supports the 1GHz HyperTransport and the Socket 939 and the PCI Express bus and various minor things like a Gigabit Ethernet controller integrated into the South Bridge. We’ll probably see mainboards on this chipset in May-June or thereabouts.
Somewhat to my surprise, first mainboards on the nForce3 250 appeared in March (considering the speed at which NVIDIA moves on in this field lately, we might as well never see them at all). This time they did it: we saw the ill-fated nForce2 MCP that supports SerialATA and Gigabit Ethernet. The nForce3 250 differs from the first generation (nForce3 150) only in support of these peripheral functions and in the 1GHz HyperTransport/Socket 939 combination, of course. Overall, there’s nothing new against the competitors, but NVIDIA seemed ready to throw in such chipsets in mass quantities in March already.
We will of course be watching the turn of events, but, if nothing seriously changes in the next two-three quarters, VIA will increase its market share, while NVIDIA will lose. SiS will keep its slice of the pie, in all probability. I guess there won’t be the second renaissance as in the times of the SiS 645 and SiS 650.
Prices went crazy in the last month! They skyrocketed up since its beginning and were growing without any pauses until its end. As a result, in March only, the price of a 256Mb DDR400 chip went up from $4.43 to $5.53 in the spot market. In other words, 25% per month! The same tendency will probably rule in April, too.
Why so? In fact, it was clear that the beginning of the spring would be ameliorative. A long decline is sure to be followed by a period of some growth – the increasing demand from major OEM makers gave effect at last. So there were natural reasons for the prices to go up, but not that high…
Seeing the demand, the memory makers rushed into the sweetest contract market, working in the spot one just to show their presence there. That was a fertile soil for deficit to spring from. Considering that many of them had swung their “PC memory – flash” pendulum to the lucrative flash, we get the well-known situation: growing demand meets limited supply.
The traders have long been waiting for a growth phase, now they were willing to uphold this new trend, pushing the prices even higher. By the way, the contract market, much more voluminous and thus more adequately reflecting the ongoing processes, was more reserved in March – its prices grew only by 10-15% compared to those 25% in the spot market.
The spot market was feeding on rumors and everything was thrown into the heap: promises that the arrival of i915/i925-based mainboards will make an average PC use 512MB of memory. Add the commotion about the presidential elections in Taiwan and the arguable forecast of Samsung about memory deficit in the second half of the year.
The forecast was really arguable. Today, there are causes for a deficit as the sine curves of demand and supply coincided. But by the end of the year, there should be an opposite situation: average demand and high supply. Contrary to the Koreans’ prognosis, supply will grow: the Taiwanese manufacturers are putting 300mm facilities into operation before the schedule and the supply of DDR chips will be definitely higher in Q3-Q4. If no extraordinary events happen, there’ll be no significant price growth (like 10% or higher) in the fall. Moreover, the manufacturers will be transitioning to 512Mb chips as the “raw material” for mass memory modules. This fact can only reduce the prices, too.
The artificial nature of this monstrous price growth is confirmed by another fact – just compare it to the near-zero growth of mainboard and notebook shipments in March against February. I can believe that the average amount of memory in PCs grew in that time, but not so much! Thus, “waking up” in April-May to notice no i915/i925-based mainboards in retail, the finished transition to the 0.10-micron tech process by all major companies and feeling the smell of the approaching dead season, the memory prices will most likely start to plummet down at the same rate.
Closer to the end of the year, it’s going to be very fascinating to watch the companies change their positions in this race. Someone will ride the wave, but others will drown. Samsung will probably find itself among those lucky companies who end this year with profit. This is the more important for the Koreans as they lost some share of the market in the last year. According to the refined data from iSupply, by 3.6%. At the same time, Samsung is still on top, having 28.6% against the second company, Micron’s 19.1%.
By the way, this share reduction doesn’t mean anything – Samsung simply transferred some of its facilities from DRAM to flash. In this year, however, they put new lines into operation and finish the transition to the 0.10-micron tech process (and start moving towards 70nm), which is going to regain the percent back to the company.
Micron, notwithstanding the last-year growth of 1.1%, will probably reduce its share this year by the same reason: the company has recently announced its intention to take up flash seriously. Well, they should have done that before as the company rolled back to a loss of $28 million in the last quarter with sales declining (although the memory prices were very good!).
Infineon grew from 12.7% to 16.3% in 2003 and followed Micron closely, but may not overcome it in this year: the company seems to have problems with the new 011-micron tech process and is going to start mass production of such chips only in the end of 2004 – beginning of 2005. Lately, the company also reduced its memory chip shipments by 50%… Overall, Infineon will probably be on the losing side by the year results, notwithstanding its activities in the mainland China.
Thus, the unbreakable Hynix has a chance of moving up to the second or third position by the end of the year – it doesn’t have to do much with its existing 14.7%. They just need to implement the 0.10-micron tech process on their production facilities (it’s ready and being introduced) and keep on the invasion to China where the company is negotiating about building their own fab.
It would also be interesting to watch the fate of the DDR550 initiative proposed by Hynix. They decided to position such chips as a cheap alternative to DDR2 and this sounds quite reasonable.
You may recall that Hynix was the hottest proponent of PC133 against RDRAM once. So if rumors about AMD’s going to overclock the memory controller of the Athlon 64 to support these 500-550MHz are right, the perspectives of DDR2 in this year will look even paler than now.
Hynix is the originator, but it’s clear that nearly each big memory maker can produce 550MHz DDR SDRAM chips in mass quantities as the numerous announcements of appropriate modules at CeBIT suggest. On the other hand, the same manufacturers also showed DDR2 products, PC-3200 and PC2-4300, just in case. But why should you buy them for crazy money if you can get cheaper PC4200-4400 modules that work right now, with the current i865PE, for example?
Like with mainboards, CeBIT brought much news here. Well, the manufacturers are mostly the same. Of course, the expo might have been simply outstanding if NVIDIA and/or ATI presented their new generations of graphics chips there. The arrival of the NV40 and the R420 was in the air in Hanover. There was hardly any new info, the industry fed largely on rumors.
In a nutshell, we were promised something monstrous called GeForce FX 6xxx (6800) consisting of a couple of hundred of million of transistors (that’s more than in a Prescott or in an Athlon 64), very hot (two 12V power connectors onboard), with up to 256MB of GDDR3 in the standard version, with an architecture equivalent to 16 standard pixel pipelines and with an undefined performance. The official announcement of the chip was scheduled to April 13, but it would be “on paper” – we’ll have to wait for a couple of months to see real graphics cards on the new chip.
Overall, it looks like NVIDIA went for a big risk trying to regain its technological leadership in the PC graphics market. The new architecture looks daringly fresh, and the chip – too clumsy and hard to manufacture. I do hope, though, the new chip won’t have the unlucky fate of the NV30 (GeForce FX 5800) – NVIDIA will live through another mistake, but with heavy losses.
The NV40 will have the time-tested AGP 8x interface, although there will be versions for the PCI Express implemented with the help of the AGP-PCI Express bridge that we already know. To avoid misbalance in the bandwidth of the two busses, NVIDIA decided to overclock the AGP interface to 4GB/s bandwidth, only twice less than that of the PCI Express. However, manufacturers of graphics cards already complained about this approach: the external bridge makes the PCB more complex and costlier by about 5 US dollars. ATI has no such problem with its integrated bridge, buy the way.
On the other hand, there’s less info about the R420 than about the NV40. The number of pipelines is uncertain even. Some say 12, others 16. In the equivalent of the existing architecture. 16 pipelines seem more probable and ATI’s product will have a somewhat higher performance – at least, that’s my opinion.
XGI and S3 didn’t sparkle at CeBIT, too. The former brought a standard collection of cards on its V8 and V5 GPUs in different versions. The Volari V3(V5?)XT is the most interesting device, which seems to be a cut-down version of the V5. In other words, a solution of the same level with the GeForce FX 5200 – not bad for XGI, if the chip hadn’t been promised to come out in the second half of the year only.
S3 showcased long-promised, but never selling DeltaChrome cards and their development with an integrated PCI Express interface. The product may appear in the market under the name of GammaChrome. They also mentioned the OmniChrome family, which is something like an All-In-Wonder on the same graphics cores. Everything sounds true to life, with snapshots of the cards and all, but it looks very much alike to what they’ve been feeding us all these years.
So, only NVIDIA and ATI have real products now. In March, card manufacturers were mastering the latest product from NVIDIA, the GeForce FX 5700 Ultra chip with GDDR3. Few really offered something, while the performance remained the same. Only the self cost and the heat generation of the cards went down.
In fact, it seems like the GeForce FX 5700 (Ultra) is the most popular among graphics card makers today – mainstream is always selling best of all. This time the graphics mainstream embraces all cards on processors from 5200 to 5700 (even with “Ultra”). Thus, they offered several models on the “pure” GeForce FX 5700, including on the fresh GeForce FX 5700LE and GeForce FX 5500. The last of them could become a hit as it is exactly in the middle of the sector, priced somewhat lower than the meaningful barrier of $100.
Other cards launched in March adjoin this segment. On the one side there’s the MX4000 and the 5900XT is on the other. Well, there’s nothing else to release new products upon. In the intermediary season like now, passions subside and the users’ demand focuses on the golden mean. The manufacturers are trying to expand their assortment anyway and NVIDIA showed two new Quadro chips: FX 600 PCI and FX700, value graphics cards for workstations.
Yes, I’m not mentioning ATI. The Canadians didn’t show up in March, but promised to answer to NVIDIA’s attempts in the low-end sector by releasing the RADEON 9550 (a slowed-down RADEON 9600). Notwithstanding the delay, ATI’s offer will probably look better: with a lower price than the GeForce FX 5500, but with a higher performance.
So ATI has another bright quarter ahead. Through excellent work in the last quarter, the company increased its sales volume by a half and also enlarged its market share. Since the rumors about MSI’s going to release graphics cards on ATI’s chips are coming true, the future of the company looks cloudless. Of course, if they don’t fail with the R420.
The beginning of the spring was a kind of Renaissance, at least there were many new monitor models revealed. We saw nothing extraordinary from the technical point of view, but on the other hand, all new products at last complied with the latest tech norms. That is, the pixel response time was typically 12-16msec, the brightness of 17” models was about 300 candela per sq. m, the contrast ratio of 450/500:1.
Well, among “pure” monitors, without an integrated TV-tuner, there are more big-diagonal models (like 19-20”), which are going to rule the market in this year. NEC offered its MultiSync LCD2060NX, Philips rolled out its Brilliance 190P5 and Iiyama showed the ProLite E511S.
There’s some divergence here. Someone tries to be appealing with a response time of 16msec, but with average brightness and contrast ratio, others stress exactly these two latter parameters, others still emphasize service functions like an automatic adjustment for the ambient lighting.
Overall, there’s some room for agreeing upon common standards. I hope the best qualities from all models will make them up: 16msec response time, 300 candelas per sq. m brightness, 700:1 contrast ratio, and numerous service functions like application-specific presets, creation of onscreen zones with different brightness/contrast settings and so on.
Of course, we can’t avoid those models with integrated TV-tuners – they will only be getting more in number as the components become cheaper. In March, Iiyama unveiled its 17” ProLite E430T-S and Samsung rolled out the 17” SyncMaster 710 MP-R and the 15” SyncMaster 510 MP-R. The new products have standard characteristics, but feature a nice and functional design.
As for future products, Mitsubishi Electric showed a prototype of an LCD monitor for designers and other such people who need good color reproduction, brightness and so on. This 1900x1200 device has a brightness of 600 candelas per sq. m and can display the entire Adobe RGB color space. After this thing comes into production, you won’t have a chance to say that LCD monitors are not for graphics artists.
By the way, about production, in March two major companies, Sony and Samsung, joined in the new S-LCD venture. They agreed upon it back in October, though, so there’s no surprise here. Fabs for producing LCD panels of the seventh generation seem to be too expensive even for such giants. On the other hand, S-LCD has little to do with computers, since the founders say that the new company will produce panels for LCD TV-sets.
Regrettably for computer monitors, the market of LCD TV-sets is of a higher priority for the manufacturers. As a result, deficit is felt from time to time, reflecting on the prices. There’s no guarantee that the manufacturers will be able to meet the ever-increasing demand in this year.
Capacities are growing; speeds are high, while the size is reduced. In March the Hitachi Deskstar 7K400 notched the record-breaking 400GB capacity, while the Fujitsu MAT3xxxNC/NP drive for server applications is only 300GB, but rotates the spindle at 10,000rpm with all the consequences like an average access time of 3msec.
The areal density is growing, too. The Japan-based Alps, a leader in the market of magnetic heads, announced their new version, providing an areal density of 120Gb per sq. inch against today’s 90Gb. This is a potential opportunity, of course, expected to come about no sooner than the next year and only if the platters and other circumstances permit. In ideal, Seagate promises as many as 50 Tb per sq. inch if you heat the platter up with a laser, but again no sooner than the year 2020 or thereabouts.
The data-reading speed is growing along with the areal density, because the angular velocity of the platter remains the same. This growing traffic must be somehow processed. Now Adaptec starts shipping the first of-the-shelf Serial Attached SCSI controller with a peak bandwidth of 4.8GB/s and Fujitsu shows prototypes of its SAS drives with an average bandwidth of 3GB/s. These are low-format devices, by the way, and this is another market tendency. We’ve got used to 2.5” HDDs and the next Travelstar model from Hitachi – the 40GB 4K40 – looks quite familiar.
Smaller models are in fashion now like 1.5-1.8” form-factors: the MicroDrive is so popular nowadays that Hitachi hardly meets the demand. Well, flash is getting more expensive, while hard disk drives are cheapening.
Microsoft, however, seems to have chosen flash. At least, they officially said it in March: M-Systems, a well-known manufacturer of flash drives, will supply its produce. Of course, they won’t be shipping 90GB models like the one they announced along with that declaration, but something like that. That’s a strange choice, if flash has been really preferred to the mechanical HDD. Let’s wait for a clarification.
The calm season ends for the optical drive market. At CeBIT they showcased a lot of 12x models that are going to come to shops in May. Some companies even showed 16x DVD+R/RW drives! 12x models will probably be announced through the second quarter, followed by 16x ones. As for sales, 12x drives will be more popular as they are most optimal. The 16x speed is too close to the physical limit, especially for dual-layer discs, although Philips is ready to offer an etalon design of such a device (16x with support of 8.5GB DVD+R).
Meanwhile, there’s already the next generation of optical drives in the market, at least theoretically. I mean the notorious Blu-Ray technology. Sony is even set to offer the second generation of such devices with support of two-layer 50GB discs, capable of reading BD-ROM, CD and DVD. That’s all very nice, save for the price, starting from $2800. Otherwise, that’s an excellent technology.
The Centrino spring is unfolding. Although there has been some time since the announcement of a new processor for this platform, the manufacturers are quite happy with the existing ones to go on throwing out new models, although those look ever more similar to the previous ones. The tablet Portege M205 from Toshiba differs from its original version of half a year ago only in the new 60GB HDD and support of the 802.11g protocol. The corporative models from iRU (Stilo 3414 and 3514) and HP (Compaq nx5000) appeared: the latter is a boring, but cheap Centrino-based notebook, while the first two are smarter with their integrated web-cams and unusual color schemes.
The next month this calm market will be blasted up with the announcement of the Dothan core that will transform into the Pentium M 715-755 (1.5-2.0GHz frequencies) and Celeron M 370 (1.5GHz). The last processor will only appear in the Q3, rather than in May. Then, Yohan/Napa and then Merom/Santa Rosa will get closer. The last combination seems fantastic, of course: a multi-core architecture, up to 4MB of L2 cache, various technologies, including the 64-bit addressing of course, PC2-6400 memory, PCI Express x16, DirectX 9-compatible graphics core and wireless network like 802.11n. Today, not every server can boast such a processor and such memory!
So far, if you want to have a non-compromise notebook at any cost, without looking at its weight and heat generation, you’ll probably have to use an analog of an ordinary desktop CPU or even the desktop processors itself. For example, that’s what Dell did in its Inspirion 9100 where an ordinary Pentium 2.8GHz (Prescott core) is the central processor. The device works for 1.5 hours on its batteries and weighs 4.1kg. This is a classical replacement of the desktop PC as all its mobility permits just moving it from one desk to another.
Of course, there are less extreme variants from the same field like the Inspirion 1150 on the Mobile Pentium 4 or Celeron that weighs 3.5kg or the HP Compaq nx9110 on the Mobile Pentium 4 weighing just 2.6kg (although the screen diagonal is 15” like in all models of that class). Well, HP also offers the Compaq nx9105, the full analog, but on the Mobile Athlon XP-M. That processor is alive and developing: the Athlon XP-M 2100+ has been released for sub-notebooks. Cheap sub-notebooks, considering the price of the CPU - $97.
Meanwhile, the 64-bit processors from AMD are in vogue today, even if those 64 bits are not very necessary for anyone. Anyway, the DTR Athlon 64 3400+ processor appeared in March – it belongs to the same league as the 2.8GHz Prescott, even higher-performing…and hotter.
But if there’s supply, there must be demand. As a result, two new models on the Athlon 64 were announced in this month only by the same HP: Pavilion vz5000z and Presario R3000Z. They are of the same desktop-replacement kind. The perspectives of mobile AMD CPUs are bright due to the attention of Asian brands like ASUS and Acer who like the price/performance ratio of those processors.
Again, CeBIT has brought some news to the notebook market, too. Exhibitions are not only about today’s products, but also about upcoming technologies. Thus, the Hanover show was remarkable for the demonstration of fuel elements that are going to permit the notebook to work with a cheap easily-replaceable cartridge for about 10 hours. Another interesting technology is ExpressCard based on the PCI Express interface – it is in fact a new generation of expansion cards for notebooks, after PCMCIA cards. Express cards have a smaller size and work faster. That’s exactly what we need from new technologies!
