ITotals - December `2002

Processors

C3, with its slightly lower price and far much lower performance has hardly been regarded a serious contender to Celeron up till now, but the news of Intel's intention to cease producing the last of the Socket-370, Celeron 1.4 GHz, (the last shipments of the box version are scheduled for no later than 11 July, with the last OEM shipments to be effected before 12 December 2003) is about to break the spirit of PC assemblers (at least in Russia): to date they don't see other alternatives to Celerons as "the cheapest PCs ever" that have been popular with most value users - the Socket-370 platform is still holding much favor in this sector.

This time, they seem to be forced to turn to another alternative - the C3 which used to go unnoticed but now you've got no other choice than turn your eyes to that. VIA seems to cause joy rather than grief to them, at least recently: C3 is indeed increasing its turnover and doubled the sales volume for November alone; the 1GHz version of the processor saw the light in Tokyo early in December, which is more than enough for the Low-end sector.

Apart from the announced demise for Celeron 1.4 GHz, Intel didn't bring any other news in December. Mind you, why and where should it come from? All what could be shot out was released still in the late third or early fourth quarter, so last month the company had nothing to do but reap the fruits: as per Intel's forecasts, its sales volumes for the fourth quarter should surpass all expectations and probably exceed the $6.5-6.9 bln target by $100-500 mln.

Anyway, don't think that only Intel is doing well because of its victories over AMD. The latter is expecting to gain $700 mln of sales volumes in the fourth quarter. Sure, that's much lower than Intel's, but it's 35% higher than in the previous quarter! As a result, although AMD is unlikely to reach profitability, nevertheless it will get as close as possible to the make-out boundary.

And it did that - no doubt. Unlike Intel, it did that in December as well. Mind you, compared to other news, the release of Athlon MP 2400+ is not so big deal to get excited about, but nevertheless.. Keep in mind that in line with the release of this processor AMD reduced prices for MP 2000+ and 2200+ by $153 and $208, respectively. Also, rumors had it 2400+ would be the last of the Athlon MP line using the 266MHz system bus and 2600+ the first to use the FSB333.

Athlon MP 2400+ is based on the 0.13 mk process technology like all the recent AMD processors. Moreover, the company is still migrating older and slower models onto the technology as well. For instance, on 11 December AMD migrated the XP 1700+ and 1800+ into the 0.13 mk process technology, which will allow reducing the company losses, and - you never know - maybe retrieving the positive dynamics in the second-third quarter of 2003. In fact, AMD is not confining itself to low-end models - in the middle of the month Athlon XP 2700+ hit the retail shelves in Japan, with first restricted shipments of Athlon XP 2800+ started simultaneously.

There are some other interesting news items regarding the mid-end Athlon processors, although they don't quite relate to real products or even their announcements. It is a question of first test results on Athlon XP 2500+ Barton (0.13 mk, FSB 333, and 512K L2 cache). In fact, they all proved quite predictable - until the release of Athlon 64 due to the evident and vivid superiority, AMD has gained some stock in both the performance (the 512K L2 cache is off jokes) and the clock speed (mostly due to the raised system bus speed, so the performance boost is not likely to be that essential): for now, no higher than 2800+ is seen on the horizon, although no earlier than in the second quarter.

As for the Athlon 64, remember that fresh results of its tests appeared in December - this time it was an evaluation unit running at 1.2 GHz clock speed. Well, the results are somehow encouraging. That evaluation unit exhibited performance approximately the same as that for 2.2 GHz Pentium4 on average (keep in mind its 1MB L2 cache versus 512K in the Pentium 4). By the time the Athlon 64 is factually released, Intel is expected to offer 3.4 GHz, therefore, to stay competitive, this AMD's new offspring has got to offer a clock speed starting at 1.7 GHz at least, so things are not that bad and the expected clock speed is going to be around 2 GHz on the date of its release.

Since the 'on-paper announcement' of the families is likely to take place somewhere in March-April (presumably, at CeBIT), the dates of factual release are getting more certain, and these processors won't appear in mass quantities earlier than the second half of the summer. That's when AMD is going to do much better in terms of finances especially in view of the fact that the Opteron is being earnestly positioned as a "Xeon killer" and AMD seems to have some serious grounds for that. 64-bit addressing, integrated DDR controller, > 1MB cache sound impressive. This results in a solution no weaker than Xeon plus the competitive price AMD has always been riding high at. What really matters is that AMD has to do much to conquer the market - the Athlon MP does not seem to have gained much success.

In any case, this year the Opteron should increase from 1.4 up to 2.6 GHz and will be quite on par with Xeon at performance, especially if AMD does implement its response to the HyperThreading in the renewed cores of further lines. And AMD really is capable of doing that - at least the respective patents of its own are there and the new K9 line seems to have something for that already planned. Again, if AMD, as it promises, has enough force to overcome the 0.09 mk process technology this year, then Opteron will have every chance to measure swords with Xeon. However, let's be realistical about it - AMD is unlikely to cope with the 0.09 mk process earlier than the year 2004.

Mind you, it's not only the 'Intel vs AMD' confrontation that rules the PC manufacturers market. This market is also represented by the dual-processor PowerMac driven by the PowerPC G4 even though it features merely 1.25 GHz and lags well behind both Athlon MP and Pentium4 systems, let alone Xeon systems. There is also the Alpha EV7 promised for the coming January, the last but one in this family offering 1.2 GHz and integrated dual-channel RDRAM controller. And last but not least - the IBM PowerPC 970, scheduled for the second half of this year, which is in fact a cut-down version of the Power4 and aimed at the desktop end of the market: it offers 64-bit, >1.8 GHz clock speed, multivector multithreaded command execution. This is really going to be a good chance for the Apple, although it'd be better off migrating to the x86 processors.

Motherboards

Gigabyte 8INXP (Granite Bay)

Next are Asus P4G8X and Iwill P4GB.

Asus P4G8X (Granite Bay)

Iwill P4GB (Granite Bay)

Meanwhile, the official names for further Intel chipsets aka Springdale and Canterwood have been announced. Basically, the names are quite evident: the Springdale line will be dubbed as the i865 (G, P, and PE), with the Canterwood, a workstation version, heir to Granite Bay, dubbed as the i875. The latter will be produced in the only version - P. As for the Springdale, we can't help mentioning a spicy stuff for those into overclocking: this is the first time Intel solves the overclocking restrictions in its i865/i875-based motherboards!

One more pleasant albeit predictable thing is the coming price reduction for the company's chipsets expected already in this quarter. To be more precise, on 30 March, at the very end of it. The 845 line will go down by $1-2 clearing the way to new-generation chipsets. Not a big deal though, but remember that these chipsets went down in price as much on 30 December.

Clearly, VIA chipsets will be at any rate cheaper, that's why the Soltek SL-85FRV-R built on the P4X600 chipset was just announced in December. But it's curious that the chipset itself hasn't been announced to date. In any case, even though it's still unknown how much a motherboard based on the dual-channel PC2700 chipset from VIA will cost, but we can stay assured it's going to be far much cheaper than the similar solution based on the Granite Bay chipset. The P4X600 chipset might be the last to be codenamed this way. Next will be the PT600 (the former P4X600A) and the PT800. The company decided to change the scheme of designating chipsets for Pentium4, and made a good job of that - the names turned shorter and handier.

Nevertheless, the change of names is unlikely to help evade legal proceedings from Intel. For that, VIA has its own attorneys: no sooner had the British Court of Appeal rendered decision at Intel's suit late this month to ban sales of VIA chipsets and motherboards on VIA chipsets in favor of VIA and initiated investigations on Intel's development patenting activities than in just three days VIA lodged a complaint to the Taiwan Trade Commission blaming Intel of unfair competition. That was much more serious than out there in Britain or Germany where VIA had previously lost the legal procedure.

For AMD it seems like no new platforms are required after the emergence of KT400, and Pentium4 chipsets are under strict legal pressing. So it's no surprise that in November VIA's sales volumes hit the $66.8 mln mark, which is 14.2% lower than a year before. On the other hand, there are other causes like seasonal recession because of which the November results reported by largest motherboard manufacturers got worse compared to the previous months, while SiS preserved its positions primarily due to the initiated mass supplies of the Xabre 600.

And things with it are going to the better. It's no secret SiS can't cope with the demand for its chipsets despite the fully loaded factory capacities. That's why the company decided to resort to the help from an OEM partner. And the choice fell upon UMC! That's the UMC against which SiS used to bring suits.

First, rumors had it there was sort of a rapprochement taking shape, but late in December there was announced both parties set up cross-licensing relations and established an alliance! What is more, late in December UMC contributed $18 mln to double its share in the SiS capital to make up 4.46% of the total stock. So things with SiS are going to the better. What makes us think so is the prospective dual-channel SiS655FX chipset with support for PC3200 and FSB800 as well as the fact that SiS has implemented HyperThreading support for all of its chipsets starting with SiS645DX (albeit of revision B that is just about to produce).

All is really promising unless Intel interrupts this flight for SiS who is facing problems with the FSB800 licensing. At least, through hearsay, Intel won't give in this version of the bus easily, which is logical: Intel's recent chipsets proved to be pretty successful, and where SiS can compete Intel is the price, of course. Then why not to correct the disparity at the expense of license taxes?

Maybe that's why NVIDIA once decided not to take the trouble of getting the license from Intel but restricted itself to producing chipsets for AMD, which is what they have still been doing up till now. Quite recently, nForce2 came into play, so it's no surprise motherboards on its base are still being promoted by Biostar, Chaintech, FIC, MSI...

But ATI, unlike NVIDIA, contrived to get a license for the Pentium 4 system bus much earlier than others. This can be already used as some sort of a criterion: Intel certainly favors to the weaker market players and as soon some of the players stands on the way trips it up at once. That used to be the case with VIA, NVIDIA, now it seems to be the turn for SiS: Unlike them, ATi has a lot of room to climb up - this year the company is planning three new chipsets for Pentium 4 and two for Athlon 64.

Meanwhile, AMD is already announcing a list of its partners for the production of Opteron motherboards based on its own chipsets this year. Among the lucky stars are MSI, FIC, Gigabyte, Tyan and Arima. A very strange choice, isn't it? At least it's absolutely not clear how the list was made and why Asus is missing in it.

The second-level manufacturers are fond of producing rather weird motherboards. No one can tell the reason why Soltek produced two new Granite Bay and nForce2 boards in December all gold-plated from head to foot, AOpen is tampering with boards equipped with vacuum lamps illuminated by specially mounted LEDs, Elitegroup keeps on reinventing motherboards with integrated graphics chips onboard - Gameunion 746, as well as with Xabre 200 featuring 64MB of local DDR. But this tampering always has a definite plus - such novelties are worshipped by the computer press which keeps hyping them with lots of photos for dessert!

Indeed, it's time manufacturers took care of increasing sales volumes a bit: the November totals show that some of them should have raised the alarm upon looking at the shattered sales volumes! This primarily concerns Gigabyte and MSI, since their competitors exhibited quite positive trends. Among the most successful competitors is ECS that stays as the leader at sales volumes (in quantities of pieces sold), with ASUS running closest and reluctant to give in easily. In this quarter ASUS is about to start producing new Springdale motherboards for Intel, which should add 3-4 mln boards more for the year 2003. MSI is also doing pretty well - rumors have it this year it's going to become an OEM-partner with Dell.

Memory

On the other hand, this pressure might be smoothed out due to another factor: it's becoming more fashionable with manufacturers to gravitate towards OEM partnerships and fulfill guaranteed orders at guaranteed prices, so it's no surprise more and more chips manufactured by DRAM suppliers immediately leave for the OEM line bypassing the open market. And in some way the production growth might level off this process. So the most reasonable forecast will be the price reduction albeit not that sharp as could be expected. It's no doubt the prices will go down, to which the following fact points: as analysts say, putting into operation of all factories scheduled for 2003 alone should increase DRAM production volumes by 66%. This is despite the last year's DRAM production growth by merely 36.7% and not because the manufacturers could have increased production facilities. This market traditionally limits the demand, not the supply.

Meanwhile, already in December there appeared some preliminary annual totals. Samsung is definitely taking the lead, with Micron at the second a bit unstable position. Third comes Infineon headily soaring up to newer heights. Let's see how this Table of Ranks will change in the upshot: chances are the second and third places may reverse even though Infineon finally broke up with ProMOS, abandoned membership in that company and in January onwards will cease purchasing chips made by it. (They are already flooding the open market, cutting down the prices). Well, assume Infineon lost 10-15% of the total DRAM chip sales. But there is Nanya, whose first cornerstone into its joint factory was laid down on 2 December, there is also the Chinese SMIC which is shipping 0.14 mk memory modules to Infineon following the "chips in exchange for technologies" formula. There is also the guaranteed market - its 5-year-long $2.5 bln supplies agreement with Kingston.

As for ProMOS surpluses, there is a chance they might get directly into Elpida's hands - at least, negotiations on that are underway. We wish it the best of luck, although such volumes can't change the situation radically, judging by the goals it posed - during the year it has got to increase the production volumes by 6-10 times!

Good ambitious plans, aren't they? I wonder what goals Micron set for the previous quarter. At least the quarter was completed at -$316 mln with the $685 sales volume. It looks like a paradox - from the technical viewpoint the company is doing well after migrating around 40% of its production lines to the 0.13 mk process technology, but anyway for several quarters the company has been deep down. Traditionally, Hynix was appointed scapegoat, and Micron's quarterly report point-blankly accused it of damping. In some way, Micron is right and Hynix once again confirmed the accusation while receiving another ring-buoy in January in the form of $1.6 bln converted permanent debt of the company stock from creditors and another $2.5 bln issued for the technical re-equipment of production. So the decision of U.S. Ministry of Trade to introduce barrage dues against Hynix may affect it in a bad way.

It's a pleasure to see that the companies are fighting on the technological ground as well. In December, Hynix presented fastest DDR chips ever - the 128 MB chips offer 500 MHz clock speed, which is equivalent to the resultant 1000 MHz! In fact as powerful as DDR II, but about 10% cheaper and having the older wiring. Micron responded to that by presenting 0.11 mk 1GB DDR SDRAM chips, thus holding two records at a time - the 0.11 mk process suitable for commercial production plus the 1GB for DDR.

As for the finished products, the memory modules, an interesting thought is arising: the same situation as with motherboards is taking shape. Trying to increase sales volumes, manufacturers are coming up with new solutions trying to expose those small user groups who ignored them before. Among the good samples that hit the retail early in December in Japan are merely the 2GB register PC2100 modules built on Infineon 512 MB chips (that used to go to the OEM channel completely) and the second-generation OCZ "overclockers' modules" in which due to the better selection of DRAM chips and finer dressing of the PCB the company succeeded in reducing the CAS Latency down to two cycles in the "PC3500" and "PC3700" modules.

In fact, it's not a long way to go from PC3200 to PC3500, and JEDEC at its December session finally made up the decision to make the PC3200 standard, thus skillfully wagging to Intel who made a U-turn in its direction. We must admit though the decision made by JEDEC is simply a de jure confirmation of the currently established situation. A good example of that is Samsung who in November showed 10 mln PC2700 and PC3200 chips per month in the 128 MB equivalent. This is despite that Samsung's DDR share in the totally produced memory modules is much lower than that of Nanya. Now that Intel and JEDEC officially "recognized" PC3200 the way is cleared, so by the end of the year the PC3200 should take over almost 100% of the DDR modules market.

There is an option to further increase the bandwidth without any raise of clock speed - take for instance the notorious QBM by Kentron. Upon approval of the PC3200 the company suggested the option of QBM modules built on its base, the PC6400. That's great because it is an ideal fit for prospective FSB800 Pentiums, however, the cost of a QBM module appears to be around $15 as expensive than for the regular module of similar capacity. It's definitely not an option for the mass market, so the following scene is taking shape: among the chipset manufacturers who support QBM modules, are only VIA (and maybe SiS, theoretically), and among the modules manufacturers it is only PNY. The idea is good, but hardly viable.

Video

First, there are two turncoats who catch the eye: the newly established ELSA, who early in December declared the change of orientation towards ATI, promptly released as many as five video cards built on ATI chips, from Radeon 7000 to Radeon 9500 Pro, with Visiontek joining in the middle of the month. Interestingly, in both cases the change of directions occurred immediately after the two companies found themselves standing on the brink of a precipice. What was that - an attempt to catch at the last straw or the augmented ability to assess the market on getting into a scrape? Look at Gigabyte's experience - the company was so concerned by the direction that decided to detach its video cards manufacture into a separate division, headhunted the former C.P Technology (PowerColor) vice-president and appointed him director.

The list of video cards produced covers quite a range from Radeon 7500 and Radeon 9000 Pro up to the classical samples based on Radeon 9100 and Radeon 9700 Pro. The former case is most notable for the fact that PowerColor and Sapphire presented boards built on the latest ATI's Low-End chip, while the latter is remarkable for the Sapphire video card, unique of the kind, where the Radeon 9700 Pro is cooled by passive radiators only!

Anyway, ATI consumers and partners are having an exciting time: The immense demand for Radeon 9500 has caused its shortage, not to mention the Radeon 9100 whose demo samples have just been released for public awareness; and that's not yet all - the company is currently at the cutting-edge RV350 whose pre-production models have already been received from ATI. The situation with this new generation of ATI chips is quite confusing: the higher-end R350 is going to be manufactured using the well-tried 0.15 mk process technology, while its lower-end brother, RV350, a 0.13 mk version of today's R300, will be manufactured by TSMC using the 0.13 mk process technology. Interestingly, both these chips will be released simultaneously - this coming spring. Therefore, we are drawing the conclusion - TSMC is indeed having problems with the 0.13 mk process technology, and even if it solved them, its capacities would not be enough for ATI this spring.

In fact, we can stay assured things with ATI are going to be better in this quarter than in the previous: the $5 mln net profits (you won't believe it, but in the previous quarter it ran into $32.2 mln losses) are indicative of improving positions because the line currently offered by the company looks balanced enough, and the backing from partners is strong as never before. This remains to be seen though: for any IT company, the first quarter of the year is not a bed of roses.

This fully applies to NVIDIA either for whom the recent quarters have not been as successful as before. In this quarter, the company may revert to the positive, which is not yet certain though: mind you, the Ti4200-8X is of course a hit but not enough to keep the sales afloat. Apart from it, nothing else is there: the yet undeclared Ti4800 and Ti4800SE are in fact AGP 8X versions of Ti4600 and Ti4400, and by no means are suitable for being a hit. There is also the MX440-8X, but the AGP 8X support is unlikely to add much to the good old and popular MX440. All this is expected in January-February, and in February-March first GeForceFX-based video cards from MSI, Leadtek and 3D Club will start hitting the retail at prices around $600-700. Doesn't look like a successful experiment, does it?

A good thing for NVIDIA is the release of DirectX 9.0 including the high-level shader language HLSL by Microsoft, and soon we can expect first DirectX 9.0 benchmarks to run smashing demos under GeForce FX: albeit a sham, isn't it time we had DX9 application accelerators? If you look at the synchronism at which the DX9 and NVIDIA's Cg Compiler 1.0 were released in December, you are getting thrilled to see what will come of it in about half a year or a year from now.

Enough with the major players - let's turn to the minor contenders of the graphics accelerators market, among them SiS with its Xabre 600 (a 0.13 mk version of Xabre 400) whose supplies started in November, which resulted in Elitegroup's release of Camelot Xabre 600 based on the same chip; and S3 Graphics with its so long-awaited Columbia core renamed into Delta Chrome X9. Judging by the declared specifications, it's going to be some sort of today's Radeon 9500/Radeon 9700. That is, S3 Graphics is definitely not a contender for the major league. Next comes Trident who is most unlikely to aspire to the tops especially after publishing its XP4 tests: the performance is several times lower than in Radeon 9000 Pro. Despite Trident's promises to keep the prices for XP4 video cards under $100, Radeon 9000 Pro cards are just in that price range already:

On the whole, nothing has changed: ATI and NVIDIA, plus the hopeful SiS. Petty players don't count.

Data storage

As for hard disks, the past month was not notable for any bright news. The U.S Trade Commission approved the Hitachi-IBM bargain that resulted in establishing Hitachi Global Storage Technologies, of which nobody had any doubts. A really exciting news is that at last first Serial ATA hard disks by Maxtor appeared in stores in Japan, albeit at transcendental prices. Although we have already received a 120 GB Seagate SATA V specimen of our own, these disks are unlikely to be offered for mass sales soon.

More realistical are the new Barracuda lines having renovated markings - 7200.7, 7200.7 Plus.

Interestingly, the new hard disks are coming to take the place of Barracuda ATA V, but where has the sixth generation gone? The modifications differ in the cache size - 2 or 8 MB and in the capacities: 40-160 GB for the former, and 120-160 for the latter. There is also a Low-End offering - Barracuda 5400.1! I wonder why on earth Seagate decided to deface the trademark by introducing a model offering less than 5400 rpm? Nevertheless, that is a fact - meet the heir to the no less famous Seagate U series, an Ultra ATA/100 40 GB hard disk measuring as small as 2 cm in height, the average seek time 12.5 ms.

Of prospective solutions, TDK declared new shockproof magnetic heads for hard disks able enduring 1000G shocks during the operation! This is almost the level of flash memory storage. Let's wait for the second half of the year when the company starts the mass production. By the time, the newly awake iVDR alliance who promised to demonstrate first CES disk samples might start manufacturing these products - 2.5" SerialATA-based cartridges aimed for computers and household appliances.

As for the optical storage media drives, things with them are in full swing. For the past month alone, as many as 20 new models of CD-ROM, DVD-R, CD-RW and DVD/CD-RW drives of various makes and modifications were presented! There were also new chipsets like the new Sanyo controller with support for the HD-BURN that allows doubling the CD recording density. Then go new documents sort of newer DVD+RW and DVD+R specifications issued by Philips, and finally new disks like 40X 800 MB declared by TDK.

In the coming spring, the Sony/Nichia alliance is starting to manufacture blue lasers, with Sanyo and Sharp as the would-be competitors. We're in for flood of news on optics as soon as new alliances start fighting each other in earnest and the DVD-RAM, DVD-RW versus DVD+RW confrontation will be no more than childish sports.

This will give new causes for wars which are being prepared just nowadays. Next year Matsushita and Ricoh are about to start creating a new technology of optical disks that allows storing 1.5 TB data on each, which is an implementation of the 3D 10-layer optical media principle, something close to the holography. And this niche is abundant with fortune-seekers.

For instance, InPhase, a well-known division of Lucent, received a $2 mln U.S.Government grant for the development of new holographic media materials, so the joint work with the partner, Hitachi-Maxell, will gain momentum. Last month Aprilis, a Polaroid's "daughter", also specializing in holography, acquired 21 patents to do with this field from Manhattans Scientific. Up to now, Aprilis' most outstanding achievement has been the 200 GB WORM-disk, so let's see how far it's going to advance.

The stakes are high, so the contenders are desperately striving to the finish as fast as they can because the first to finish will create and rule the standard, and who knows - maybe press the grandees of today. Another reason to hurry on is new discoveries, not the holography alone. The universities of Juta and Ohio are doing research in polymeric photomagnetism: as investigations of this unique phenomenon have shown, these materials if magnetized through light induction create structures which are in a way similar to three-dimensional fractals on the nano-level. Every year brings several discoveries like these which could be made use of for creating future-generation data storage media.