As to processors July appeared predicted, but still a bit unexpected: Intel has released a number of previously announced processors; both Intel and AMD have announced quarter results precisely in time, which became a surprise for many people. And, naturally, everybody tried to plan the future.
Intel was the best to finish July in the means of production, it has presented processors for each segment of the computer market. Did you say that AMD is close to 1 GHZ in the Low-End sector? As you wish - here's the 900 MHZ Celeron (with 850 MHz mobile variant at the same time). Though, perhaps, we should stop here, as the core voltage makes 1.75V already. 950 MHz or 1 GHz are still all right, but we should pass to Tualatin core, if we want to increase Celeron frequency any further.
By the way, in July Tualatin has appeared for sale in Japan, that is usual. The most junior 1,26 GHz processor variants have hit the counters. The price, certainly, had typically Japanese amendment, but it's awfully high even with it - about $315-340. However, it's not a surprise, and Tualatin is not sold, for example, in Russia, because dealers just do not supply it for such price, assuming the appropriate demand.
The reason of similar expensiveness is obvious as well: not to allow Tualatin to compete with the beloved brainchild - Pentium 4. Intel is compelled to raise Tualatin prices to make buyers prefer Pentium 4. Thus, Tualatin has modest enough role of a server processor for Socket-370 platforms, and, first of all, the role of a notebook processor.
Sony, not waiting for the official Tualatin (Mobile Pentium III-M) announcement on the July, 30, has unofficially announced a new model on its basis - Bio Note GR, that is rather advanced, by the way: 1.13 GHZ Pentium III, 830MP chipset, Radeon Mobility. Some time after that, right at the end of the month, when 866 MHz - 1.13 GHz processor has been released, the majority of grands (Compaq, Dell, Gateway, HP, IBM, Sony) has immediately presented their notebooks on it.
And it's not a joke about server application: there are fine ServerWorks Socket-370 chipsets that just need new processors if Intel does not want to hand this market over to Athlon MP. As a result there are such products, as Dell PowerEdge 2550 - dual-processor rack-mount 2U-server on the basis of ServerWorks HE-SL. The conclusion (taking into account 512 Kilobytes of L2 cache on Tualatin) is obvious: such platforms should exist even if Intel wants Pentium 4 to occupy this place.
The favourite itself develops fast and it has grown up to 1.8 GHz already in the beginning of July. But it's still about the same 0.18 micron Willamette, as 2 GHz 0.13 micron Northwood is expected in November. Is there any sense in waiting? Perhaps, yes. According to our tests, P4/SDRAM systems concede to P4/RDRAM ones in productivity at the same frequency, but the difference in price nicely compensates productivity loss. That is in case of productivity fall, price/productivity ratio will not get worse.
It's far more confusing in the sphere of server processors, as here are the favourite OEM Tualatin, Pentium 4 (Foster's server appearance), actively pushed forward by Intel, and Merced, that's based on millions of dollars and thousands of man-years. And all of them have to be placed on single not very extensive market.
900 MHZ Pentium III Xeon (2 Mbytes L2) has left the stage in July, due to a processor mistake, that has compelled Intel to suspend its deliveries until August, but the first Xeons (Fosters actually) with 256 KBytes of L2 cache and 1.4 and 1.5 GHZ clock rates have appeared for sale in the same Japan - actually they are the same Pentiums 4 in another form-factor. $400 for 1.5 GHz is cheap. And incomparable with Itanium for sure: even the most junior 733 MHz variant of it costs $1,177. But Itanium has 2 Mbytes of cache and incomparable productivity. It's just another level. New Zealand SGI, hesitating between Pentium 4 and Merced for a long time, has eventually chose the second, having completely refused the further development on 32-bit processors.
Intel's financial parameters will undoubtedly raise, when full-scale sales of Pentium 4 begin, but still aggressive processor pushing to the market brings the company, well, afflictions. Sale volumes has lowered by 24 percents in comparison with results of similar quarter in the last year, the decrease of net profit has made 94 percents and Intel has earned only 196 million dollars in the second quarter of the current year.
The company does not surrender despite this and continues to increase capacities like other largest players of this industry. According to hearings, in year 2002 Intel is going to proceed to 300 mm on the Californian D2 factory that releases chips with four 0.18 micron lines. Thus, it's already seventh 300 mm projects of the company, but the truth is that at least three of them are already suspended. (And the first 300 mm factory to be engaged in production is Fab 11X in Albuquerque, New Mexico. It will be started already in the second half of the next year.)
In the middle of July Intel and AMD have announced reduction of prices on the processors. The truth, if Intel has modestly thrown off the prices for PIII 900 and four senior Celeron models (and appreciable enough on all mobile PIII from 750 MHZ up to 1 GHZ), AMD has more than appreciably reduced the price for all processors. Duron 800 has fallen from $64 to $50, Athlon 1400 - from $253 to $176. And there will be 0.13 micron Northwood in the fourth quarter that AMD will have to answer accordingly. So, price wars continue.
And exactly the same as Intel, despite financial losses (985 million turnover - lower by 17 percents than the last quarter, and the decrease by 10-15 percents is expected in the following quarter; the profit in comparison with the last quarter has fallen from $124.8 to $17.4 millions), AMD needs to be engaged in the manufacture expansion - the company is already selecting a site for the future Fab35, which construction will cost for 3-4 billion dollars.
AMD still tries to continue without extreme measures: pilot production of 0.13 micron copper chips under SOI technical process began in the middle of July on Fab30 that is the following generation of AMD's technical process (by the way, escalating Fab30 up to this level by 2003 will cost AMD 2.3 billion dollars). AMD tests Isonics silicon-28 for a long time already - powerful means of reduction processor heat-evolution.
Unfortunately, all this will not be required soon - only Palomino and Morgan cores, that are updates of Thunderbird core, are planned this year by AMD. It's August, but only mobile Athlon 4 (which AMD will overclock up to 1.3 GHZ at power consumption about 30W by the end of the year) and semi-virtual Athlon MP (announced, but still not for sale) were created on the base of Palomino. Though AMI and 3DLabs have announced the support of this processor in the BIOS and videocard drivers.
VIA was not very active this month, but, nevertheless, it may prepare for glory: we'll not see Pentium III on the basis of Tualatin core in mass sales, but there are lots of motherboards released for it. And they will ideally suit the following version of C3 - Ezra-T, designed under 0.13 micron technical process and electrically compatible with Tualatin. 800 MHz - 1 GHZ variants of this processor should be released by the end of the year, probably, with the preservation of the same price range - 50 dollars for the senior model. We also await the first Nehemiah engineering samples in the end of the year, that will be the following generation of VIA processors with 384 Kilobytes of cache, SSE and multiprocessor configurations support - an ideal variant for a not too expensive, but still powerful dual-processor home PC under Win XP.
Server manufacturers have already started to deliver C3 to clients, suffering from an energy crisis in California. But the business is spoiled with the absence of dual-processor configurations support in C3. Nehemiah should resolve this problem, and, probably, in the next year the amount of servers on VIA processors will appreciably increase.
Two other pieces of society column deserve a separate mention. The first is the doubtless success of ARM, winning new allies and keeping old ones. In July Samsung has announced the licensing of ARM processor cores, Intel and TI have prolonged license agreements, Epson has announced it's system-on-a-chip product.
The second event is the final transition of Transmeta to TSMC. TSMC will begin to produce 0.13 micron Crusoe TM5800 that will allow to reduce their price from $200 to $60-70, and in due course after the expiration of IBM contract, TM5600 and TM3200 manufacture will proceed to TSMC.
In spite of the fact that Computex was in the previous month, July appeared devilishly saturated all the same. Not with announcements, but with presentations. The majority of platforms announced this month was designed on the basis of very popular i815E/EP or should we say i815EP stepping B, namely a board with Tualatin support: Asus TUSL2-C, SOYO SY-TISU, Soltek SL-EP2 +, ESC P6IEMT, three boards from FIC.
And if to take into account that Intel will be compelled to reduce of the price on i815EP this quarter: VIA has reduced the price for PM133 to $11-13, and someone sly had already began to manufacture i815EP, selling it for $21, so Intel will need to reduce the price for last integrated chipset in any case, thus making it even more popular.
There are more and more solutions that appear along with the process of Pentium 4 price reduction, both on i850 and i845. The companies "polish" first wave products, proceeding them to four-layer design and adding new features: ABIT TH7II-RAID (ATA-RAID, USB 2.0, the sales will start in December!), Asus P4T-F, Epox EP-4T2A. However, there were more solutions on the base of i845 for Socket-478 and, naturally, with SDRAM support. It's interesting that second echelon manufacturers have mainly showed themselves here: Jetway, ZIDA, Transcend, Soltek, AOpen. It is also necessary to note that 80% of these boards are delivered with built-in ATA-RAID controllers.
Boards under Pentium 4 gain popularity very quickly: for example, Legend Holdings, the largest PC manufacturer in China, has increased Gigabyte's request for such boards in July from 14 up to 30 thousand units per month. In august Asustek started to deliver 200 thousand P4 boards per month to Dell - for the first time the Taiwan motherboard manufacturer has received a request from Dell. ESC has reduced the average price of P4 boards to a level below $100, and the next year, when there will be integrated chipsets for Pentium 4 including ones from Intel, the popularity of this platform will raise even more. In the fourth quarter of the present year the share of Pentium 4 boards from some large manufacturers will make up the half of the total amount delivered by them.
However, VIA's business isn't bad al well. At least, its KT266 has quite a popularity: almost everybody presented boards on this chipset in July - from FIC, DFI and SOYO to ASUS, Fujitsu-Siemens and AOpen.
And it's more interesting about Apollo Pro266, almost all July boards on this basis were dual-processor: Supermicro P3TDDE on Apollo Pro266T, CUV266-D from Asus, J-953AS (266T) from Jetway, SDVIC from Rioworks, NEX 6620A from Nexcom. 440BX dual-processor platform for inexpensive servers had been the most popular before, but since Intel could not offer anything similar, VIA has occupied this niche.
In the beginning of August VIA should start deliveries of P4X266 which will cost manufacturers 25-35 percents cheaper than i845 SDR-version ($25-30)! Probably, it will allow VIA to keep the July tendency of gaining profit: the thing is that until July its income was dropping.
It's interesting that the demand for SiS production has appeared - it seems they buy everything this summer. First, SiS itself has announced mass releases of rather successful SiS 735 in August, and second, motherboard manufacturers have already presented many models on the basis of this chipset and its variant with memory controller with only PC133 SDRAM support: Gigabyte GA-7SDX on SiS 733, PC Chips M830LR (SiS 735), Chaintech 7SID (SiS 735), ECS K7SEM (SiS 735), Jetway J-835AS (SiS 735).
Taking into account that chipset has everything integrated, from graphics core to 100 MBit/s Ethernet, and its quite good speed parameters, and also the fact that the board from PC Chips, for example, costs about $70 and Durons are inexpensive now, SiS 735 may be named a hit of the season. And soon there will be another "tasty" thing - SiS645, a chipset for Pentium 4 with PC2700 DDR support.
All other developers behaved modestly enough. One board from ALi - Asus TUA266 on Aladdin Pro 5, though in July Ali has shown the best result for last 33 months: the volume of sales has made 12.3 million dollars. It's not a surprise, however, as the overwhelming share of the sales volume belongs to system-on-a-chip solutions and DVD drive controllers.
AMD also has nothing to be proud with: MSI, that was going to release K7D Master on the base of AMD760MP + VIA 686B in August, has refused this idea by the end of month after getting to know about problems of this sheaf, and has switched to the development of boards on the base of AMD760MP + AMD766. And in July Thunder K7 from Tyan has received Microsoft WHQL certification, having opened a way to the workstation market for Athlon MP: Microsoft warrants that there will not be any problems with it under Win NT and Win2000.
By the way, in July Tyan has released Tiger MP for "simple mortals" - the same Thunder K7 actually, but with a simple ATX power unit and without any advanced features - SCSI, LAN, VGA. In common it's a rather cheap variant.
NVIDIA, being realistic with its nForce, believes that this product will bring up to ten percents from the whole income next year. It's certainly about chipsets for Athlon, as Intel does not hurry to give the license for Pentium 4.
This month appeared an exact copy of June: the prices were stopping and dropping repeatedly. The general situation, nevertheless, was pretty clear: slow, but steady price fall. If on the 1st of July 128 MBit PC133 chip costed $1.85 - 2.0 on Taiwan, then already in the middle of the month PC133 has overcome $2 mark, and by the end of the month its price was changing within $1.62-1.72 limits, having decreased by 13 percents for one month. The cost of 64 MBit chip made $0.80-0.90 in the end of the month.
Taking into account the "back to school" season, that traditionally starts in September, it is necessary to say: our forecasts were true unfortunately, and the season of preparation for a new year of studies has not helped manufacturers. The next hope is Windows XP, but the future of this OS seems gloomy enough: there are certain doubts that Microsoft will be able to convince a significant amount of users to refuse from Windows 98, ME and also still fresh Windows 2000 for the benefit of Windows XP. It especially concerns corporate clients, especially in October, that is so improper for this purpose.
It is possible to assume that the expectations of computer industry about this OS will be shattered by serious disappointments in the fall. The hope for the probable rise of demand should be placed on Christmas only, in our opinion. And then one mustn't be too optimistic: rates of software requirements growth are reduced, transition on XP will be smooth enough, and capacities grow. In 2003 when the majority of 300-mm factories engages in work, it will be a shock, and today's events, as a result of which, for example, Hyundai was on the verge of bankruptcy, will seem an easy shake-up.
Some people doubt that all promised factories will start working. Not all companies will have enough money to realize this project in the planned terms: all four grands (Samsung Electronics, Micron Technology, Hynix Semiconductor and Infineon) have already informed about pushing the start of their 300-mm of factories about a half-year further (the end of the second half of 2002 and further).
Meanwhile even brands should think about the improvement of already existing manufactures (all of them plan to proceed to 0.15 micron technical process by the end of the year, that will allow to increase DRAM chip manufacture by 30 percents), rather than about the construction of new factories. Though Samsung and Micron carry out even more aggressive policy: these companies plan to proceed to 0.12-0.13 micron technology by the end of the present year. It is possible to assume, that by the end of the five-year plan just three companies will stay alive and capable of supporting new technologies: Samsung, Micron and Infineon.
Hynix, probably, may be charged off as it's not capable of spending much money on technical reequipment, and that will inevitably delete it from the leader list in some years. To decrease losses (having billion debts) company had to close the DRAM factory in Oregon for a half-year. It will also be reequipped for the latest technical process available to Hynix (0.13 microns), but other companies do the same almost without manufacture stops.
In the field of memory technologies DDR SDRAM is, undoubtedly, the leader and we should "blame" Crucial (Micron) for it. It announces DDRevoluton, aimed on progress of this memory type, then just reduces DDR price: $45 for 256 Mbyte PC2100 DDR module, and the price of a usual 256 Mbyte PC133 from the same company is $43.2.
In the beginning of July Transcend had time already to release 512 Mbyte PC2100 modules to the market, and as the company has announced, for $200 cheaper because of sharp decrease of DDR prices. The next year, that is already at hand, will bring us the sales of PC2700 333 MHz DDR modules, the only remaining before DDR II.
And situation with Rambus is rather strange. On the one hand everything is all right: the company will present the "Yellowstone" on Rambus Developer Forum 2001 in September - the technology allowing RDRAM chips to achieve bandwidth of 6.4 GByte/s, thus being actually completely incompatible with present solutions. And Kingston, inspired with Pentium 4 popularity in the third quarter, has promised to increase the manufacture of RIMM modules by 9-10 times! Hynix has indifferently announced that Rambus has successfully tested it's 256 Mbyte PC800 module.
On the other hand Rambus income drops: according to the results of the third financial quarter, the profit has made only 3.7 million dollars. Intel released its i845, that showed quite good productivity in preliminary tests, and has simultaneously decided to stop deliveries of RIMM modules with Pentium 4. Taiwan motherboard manufacturers promise to increase the share of Pentium 4 boards up to 70-100 percents of all deliveries by the end of the year. And some court words: the italian judge has denied Rambus's requirement to stop SDRAM manufacture at Micron italian factory until the court decide whether Micron breaks Rambus patents. Seems like agony.
July was boring and predictable. As usually, NVIDIA was the first, announced nothing, but having whole lots of cards released on its chips. Basically they are new variations of MX-200 and MX-400 from large and not so large companies and also some GeForce3 solutions from little companies. The undoubted hit of the season is Unika G7900 - an MX-400 card with 128 (!) Mbyte of 6 ns memory.
Even ATI looked more interesting this month. Certainly there were not any new chips; a new RADEON VE MAC EDITION card was released, being the same VE but with other drivers. But the company announced another new technology for a forthcoming Radeon2 - SmartShader. As usually the problem is about the complete incompatibility with NVIDIA cards, though DX 8.1 supports PixelShader 1.4, that is the base of SmartShader technology, so ATI will have to try hard to convince not so enthusiastic developers to use it.
And, at last, right at the end of July ATI has completely redeemed Appian's HydraVision technology for creating multimonitor cards. Actually there's just a trademark left from Appian, under which it's going to sell ATI cards.
And about cards - their amount grows more. C.P.Technology, a specially created PowerColor division for the releases of card on "non-NVIDIA chips", is going to release 8 cards on ATI chips by the end of the year, five of which - on all Radeons. The anonymous chinese companies have already begun to produce similar cards (on the same Radeon VE), making them real competitors for GeForce2 MX-400. So ATI may really increase the amount of clients, and then talk more business with game developers.
All aforesaid is actual for KYROII and STM as well. Today there is more than five cards on this chip from different manufacturers, they cost about $100-150 for 64 Mbytes variants. According to hearings, ST4800 will appear soon, on which Hercules plans to release Prophet 4800 card with 225 MHz chip clock rate!
Well, basically, that's all. It is just necessary to mention the professional Wildcat II 5000 accelerator from 3Dlabs, released in the middle of the month, it's the light variant of Wildcat II 5110 designed for AGP-slot and having 80 Mbytes of memory onboard.
This area was really saturated this month. First thing to say about CD - RW is that there is plenty of models with 24/10/40X speed formula. Second and third echelon manufacturers like WAITEC, CenDyne, QPS or LaCie were traditionally first to present them, and only then such grands as Plextor or Hewlett Packard have released their models.
The amount of shown models (16/10/40, 20/10/40) slightly surpassed the quantity of High-End solutions. And everybody showed himself here - from tiny companies like Freecom or I-O Data to giants like Philips, Mitsumi, Yamaha, LG, TDK and Plextor. Models are, naturally, for any taste - internal, external, ATAPI, SCSI, USB 1.1 and 2.0, FireWire, PCMCIA - anything you like, and rather cheap.
Certainly, there were CD - RW/DVD-ROM combo-drives as well. The sales of Ricoh's MP9200A (20/10/40/12) have begun at last; Samsung has released ultra slim SN-408B (8/8/8/24) especially for notebooks; Yano has presented external LBDV-2010FA (20/10/40/12) with IEEE 1394 interface.
Though why use CD-RW if you can write directly to DVD-disks? There were many of such offers in July, and the "pure" DVD-RAM drive has been shown only by LG (as it's the first such drive of this company), the others offered something more twisted like DVD-R/RW drive on Pioneer chassis (Logitec), DVD-RAM/R (Matsushita) or just Boa DVD-RW drive with IEEE 1394 interface (EZQuest). Apparently, DVD-RAM is becoming pushed aside by alternate formats of DVD-disk rewriting.
And some words about formats in general and about alternative ones, in particular. The specification of DDCD version 1.0 including Double Density CD Read Only, Double Density CD Recordable and Double Density CD ReWritable formats has been released at last. New DDCD standard will be licensed by Royal Philips Electronics and Sony together. In spite of the fact that the specification was released in the beginning of July, only Sony has presented DDCD drive model, and at once CD-RW by the way.
While ones increase volume, others try to raise speed. Oak in the second half of July has announced a new chip, capable of 32X compact disc recording. But for the future. Today clients are promised already existing 24X recording, but for the first time with constant angular speed.
Leaving optical drives, it is necessary to mention this society column event: Lite-ON and JVC have united the manufactures of similar drives after LG and Hitachi, and JVC has become "the senior brother" with 51 percent of shares of the new enterprise. These companies are going to build new factories in China and to completely pass to own element base (when created).
Now about HDDs. This month was rich with external models and models with flash memory, that often doesn't mean the same: 80 Mbyte WD FireWire (a usual Caviar with IEEE 1394 interface) or Logitec Ultra 160 SCSI Cheetah are not flash models anyway. Not speaking about internal 100 GByte HDDs from the same WD (7,200 rpm Caviar) and Maxtor (5,400 rpm 4W100H6) or even 10 GByte Conner, surprisingly emerged from mysterious bowels of China. Yes, Conner promised to return to this business a year ago and to start releasing cheap HDDs, but so silent and without fanfares?.. By the way, if it's really Conner, the company may be congratulated as the promise is kept: the price of 10 GByte HDD has made just $63.
But coming back to flash memory HDDs: there were solutions for any taste - from tiny 1.1 GByte CompactStor and Ritek in Compact Flash Type II form-factor to monstrous E-Disks from BitMicro with the highest 75 GByte capacity for today and UltraWide SCSI interface (of course in standard 3,5" form-factors). It's also necessary to mark a Toshiba's solution - tiny 5 GBytes PC Card Type 2 HDD which has appeared for retail sales already in the end of June, and also SmartDisk FireLite Super Slim Drive (the same 5 GByte HDD, but with IEEE 1394 interface) both easily slip in your shirt pocket.
[ To be continued...]
Write a comment below. No registration needed!