Processors

The main event of this month is certainly autumn Intel Developers Forum. The main questions were: how are the things with Itanium and with Pentium 4? Both questions are answered without the large furore but nevertheless... The motto of this IDF was "Powering the Net, Connecting the Net", so the processors were not the leaders of the programme (let us recollect the name which Intel has given to the architecture Pentium 4 - NetBurst).

From the real things only one processor - 1 GHz Pentium III Xeon was announced. As Intel claimed it was the fastest processor for servers and workstations. Then the same can be declared and about 1.13 GHz Pentium III - the new Xeon actually is slightly modified Coppermine in the cartridge for Slot-2.

What is more the architecture XScale is declared that is more known under the title StrongARM2. First RISC core of such level was developed by Intel singly but was not taken from DEC. The basic concept is reflected in the title - extremely scalable processor for different applications: from cellular telephones up to the Internet - consoles depending on required power consumption having the clock rate from 200 MHz up to 1 GHz, and the first variant exiting at the end of the year will have the clock rate just from the middle of this range - 600 MHz.

(By the way about DEC. Just in August the group of its former employees who have organised the company Alchemy Semiconductor presented its variation on the same theme - 32-bit MIPS processor under the title Au1000 - with the clock rate 200-500 MHz.)

But the fundamental moments are still not released Pentium 4 and Itanium as it was mentioned before. Itanium as it was promised would have the clock rate 800 MHz but Intel could demonstrate maximum 733 MHz. Pentium 4 would have the clock rate 1.4 GHz, but Intel could already demonstrate 2 GHz chip with air cooling.

Though we expect some problems with temperature for Pentium 4 - according to the specification Intel 42-million transistor chip with the supposed square 217 sq. mm will have heat dispersal 52 Wt and will require half a kilogram heat sink, the additional ventilator in the case and so on. The heat sink on the northern chipset bridge will be required as well.

So let us return to Itanium. 800 MHz was not watched neither on LinuxWorld nor on IDF where the prototypes of servers on this processor were demonstrated, only 500-733 MHz was. According to the plans the processor should be released this summer, though the date of commercial systems release was transferred at the beginning of the next year. It may be possible to correct something by that time-on IDF Intel refused to demonstrate SPEC results for Itanium, having declared that the results of this generally accepted standard were not so important.

Thus AMD is given a start - the discontinuity between the release IA64 in the market and declared on August, 10 x86-64 - 64-bit architecture from AMD, its answer to IA64 - the evolutionary development of x86. As well as Intel in the progression of x86-64 AMD is going to rely on Linux network in many aspects, and this operating system will allow to provide the transition to IA64 and x86-64 very easily. It is good that in the case with x86-64 it will be also possible to fulfil 32-bit applications in the same session.

If we started talking about 64-bit architectures, it is necessary to finish this theme. Sun in August at last released the samples of long hoped for UltraSPARC III, and in September the processor must appear in public. And because of the delay it will appear rather perfect: 750 MHz are expected instead of 600 MHz.

In October on Microprocessor Forum Alpha 21264E may be represented as the improved variant of the today's core optimised under 0.18 mcm technology and having 1.75 MBytes L2 cache on the chip operating on 2/3 from the clock rate. The result is in the simplified technology and large clock rate (1.2 - 1.4 GHz). Though it is not known yet whether this processor will be released at all.

And now we shall return to ordinarier direction and even move to the opposite side - from server processors to mobile ones. In August Sony declared about the intention to change Intel and at once with two of its competitors. For the first time in the notebooks VAIO bar the processors not from other companies will be used. The notebook on the basis K6-2 + was declared, and the model on Crusoe is planned. And VAIO C1 became the first officially announced large manufacturer of notebooks on the chip from Transmeta.

By the way the neighbourhood AMD and Transmeta is rather symbolic-this month the companies have declared about the intention to complete the technological union. It is supposed that Transmeta is interested in the new system bus AMD - Lightning Data Transfer, and AMD would like to come at Transmeta works on processors energy consumption drop. The variant of joint development of the processor which will protect both companies positions in the market < $400 PC - Samuel2 competitor from VIA and Timna from Intel.

Transmeta nevertheless has released the new processor which most likely will be used by the manufacturers of notebooks in their models: TM5600. 700 MHz chip is completely the same by the architecture with the predecessor TM5400, but as against it the size of the cache on the chip has grown almost twice: from 348 KBytes for TM5400 up to 656 KBytes for TM5600. However 0.18 mcm processor consumes 10 percents less power at operation, though the productivity just opposite has appreciably grown.

Toshiba spoilt the impression declared that it is not going to use Crusoe in its notebooks (and it is one of the Transmeta investors!) - Pentium III perfectly copes with this role. It is an interesting and reasonable declaration. Really the effect from the processor energy consumption drop is not so significant here as it can be understood from the declarations of Transmeta: in comparison with the screen illumination the saving some watts is not so significant, and the productivity and the name - a little bit weaker.

For example VIA doesn't have any name in processors at all and the released Cyrix III hardly promoted its construction. Let's find out what Samuel2 which is going to be declared by the company in September can do for this purpose. 0.15 mcm, small square, energy consumption and price, up to 1 GHz to middle of the next year-sound not bad. However taking into account the production volume about 50 thousand processors per a month it is interesting extremely theoretically.

In August there was a lot of information about Intel plans. Pentium 4 release will probably be a little delayed - not the end of the third quarter but the middle or the end of the fourth quarter, that is it will happen probably in November. It is supposed that the processor itself was not the main problem but chipset was, that is the last year's history with Coppermine/i820 is repeated.

It is important to wait for the release of Northwood, 0.13 mcm version Pentium 4. By that time the architecture will be completely debugged, and the price as the result of new technology will fall. And the clock rate will increase up to 2 GHz. It will be nearly at the end of the third quarter in 2001 - in one year.

And before that time Tualatin-0.13 mcm Pentium III planned on 200 MHz system bus, with cache L2 512 KBytes, having the clock rate 1.2-1.4 GHz being released in the middle of the next year will be probably the most optimal product in the processors Intel 0.13 bar. Coppermine-LV being released in the first quarter of 2001 and Celeron conversing on 100 MHz system bus in second half of same year will be rather quite good key blocks of Intel position. Though Intel classes the situation more optimistically and considers that by the middle of 2001 the part of Pentium 4 among its processors for High-End PC will be about 50 percents.

But these are only plans, but the life is going on. After the appearance of the new processors the old ones are being retired: in August Intel got rid of Pentium III 500, 533 and 550 MHz and Pentium III Xeon 600 and 667 MHz. However the situation with the new processors is not so simple. The deficiency of the number of models still takes place, for example 700 MHz Xeon - " 700 MHz Cascades is recently in the large deficiency ". What to say about 1.13 GHz Pentium III which was got rid of at the end of August? Intel made the core Coppermine be at the end of its resources, it can stand such rates. And Coppermine-LV will approach only a few months later.

But AMD felt fine at this time. It has combined the first Athlon birthday with the price drop to it and to Duron having dropped the prices on 800 Athlon up to $215, and on 700 MHz Duron - up to $79. It has released 1.1 GHz Athlon which was as usual sold by the Japanese a couple of weeks before the official announcement $775 each at the official price $853 (nonsense!).

The company is complete optimistic about the future. Fab35, 300 mms factory construction costing 3 billion dollars is planned to come into operation in 2004. To tell you the truth AMD does not hope to cope with the construction itself and is looking for the partner. Probably this time it will have to address to the help of Motorola which the company continues to cooperate with.

However we shall see how the things with the new processors will be, they may cope themselves. In 2001 the delivery of 0.13 microns Sledgehammer should start, soon after that 64-bit processor for PC-Clowhammer will appear and we are waiting for 5 new processors from AMD even sooner: this autumn or winter High-End Mustang, Mid-End Palomaro and Low-End Morgan will be released. There are 5 of them because the last two ones will appear both in usual and in mobile variants.

Motherboards

The most beautiful news of the month: Iwill has declared the first motherboard with support DDR SDRAM - KA266-R on the basis ALi MAGiK 1. Naturally the leading parameters are caused by the chipset: Socket A, 100 and 133 (200 and 266) MHz bus EV6, support up to 3 GBytes DDR SDRAM with the same clock rate, UltraATA/100 RAID (Iwill has done its best). However I have to state regretfully that the chipset release was marked only by the declared board. Everybody is waiting for VIA chipsets.

Meanwhile it doesn't hurry up. SiS has declared the only chipset this month - very long ago promised SiS730S, superintegrated one chip decision for Socket-A processors.: PC133 SDRAM, UltraATA/100, AGP 4X, Ethernet - the most powerful for today integrated chipset for Athlon. The list of companies got interested in it does not surprise: Abit, Acer, Asus, Biostar, Chaintech, Gigabyte, LuckyStar, MSI, SOYO …

It is necessary to clear up the problems with production, and everything will be all right. However Sis treats this question seriously: the construction of two more planned for operation with 300 mm slices will start in October in addition to already constructed. The construction according to the plans should be completed already at the end of the next year and should give SiS the volume 20 thousand 300 mms slices with 0.15 micron chips. The production will begin half a year later - the equipment installation should be completed at the end of March 2000 and in June 2002 the test should take place.

But let us return to chipsets and motherboards. The outstanding 440BX for two and a half years doesn't give up. Abit - EP-BX7+100 and BE6-II 2.0 have declared their new models on its basis. ATA/100 RAID was added to all the properties of the new generation, so only the absence of support AGP 4X and standard 66 MHz for AGP at the system bus rate 133 MHz remain.

The market of boards on Apollo Pro133A has become more active this month by the dual-processor models: Epox, Soltek. The availability of ATA/100 RAID controller which is absent in Apollo Pro133A is also the standard possibility. And DCS at the end of the month variegates its bar by the release of the uniprocessor 3M941 on its basis.

In general the month was very quiet: the majority of new chipset was released in June - July, and then the boards on it were declared, so that now those who fell behind are catching up - by one or two models on the chipset. The same situation is and with KT133: Abit has declared KT7-RAID where the essence of the occurred with KT7 variations is clear from the title, and Tyan has exhibited its Trinity KT especially pressing on its stability.

By the way a few words about stability. Abit has declared that it slightly complicated matters with its system SoftMenu III, therefore the part of its motherboards released before July 7 were hardware incompatible with cards on GeForce2 GTS. Besides the epopee with motherboards on KX133 is going on: Biostar has recognised that its boards on this chipset don't support Slot-A Thunderbird and called on its users to use the proper motherboards with the proper processors.

But to our surprise there are no problems with the integrated chipsets. By the way the boards on them continue to catch up. Iwill has declared its WO2-R,WO2 on i815E and WO on i815, Soltek - its SL-65ME +. These are absolutely standard models, but Iwill on WO2-R has already habitual ATA/100 RAID. And the complete parity with VIA was saved because Soltek and Transcend have declared the cards on PM133.

So it is calm and silent. The next round will begin only since the fourth quarter when VIA will represent its chipsets:KM133, KL133, Pro266, PM266, KX266 and Intel - its server i850/i860. Then in the first half of year it will be just opposite. VIA will release its server PX266 and integrated KM266, and Intel - chipsets of the average level: Tehama-E, Brookdale, Camino3/ICH3, Almador/ICH3.

New chipsets are already announced by VIA but in parts. In particular in September the next slice of mosaic was open: HDIT (High-Bandwidth Differential Interconnect) Architecture, specialised bus for the link between northern and southern bridges. 4.2 GBytes per second for the system bus and the memory and 2.1 GBytes per second for AGP and other slots should be quite enough for the nearest future. Taking into account that for any trivialities integrated in the southern bridge like two-channel UltraATA/100, the built in network controller etc., the appearance of one more additional capacity is meant.

One more important moment is VIA and 3Com declaration about their technological cooperating: 3Com will give VIA its network controller core for building in its future chipsets. The first results of the joint creativity should appear in the spring of 2001 and first will include Ethernet interface, and then the appearance of such possibilities as ADSL or cable modems support is supposed because both technologies use all the same Ethernet.

As for the graphics core for chipsets VIA the problems here are also being solved. S3 and VIA have declared about re-structuring of their agreement on graphics business S3 sale. The companies maintain that now all the claims of Taiwan government disabled the deal should be taken out.

Also it is necessary to analyse the news on server chipsets. To our surprise there were a lot of them this month due to IDF. Itanium - GX460: support up to 4 processors, up to 4 PCI controllers, separate I/O controller, PC100 SDRAM. Foster - i870 is one of the most interesting chipsets. Not depending on processor, suiting both for Foster and McKinley by the special chip that is the dealer between the processor and the concrete model of the processor. So the manufacturers of server boards have very few problems: for the models release on the new market it is only necessary to change one chip.

In 2002 the next server chipset Intel - Plumas is being prepared for release: the support up to two processors, consists of three chips - the northern bridge and divided into two parts (I/O and PCI/PCI-X) the southern bridge. It will be the first chipset Intel supporting the bus InfiniBand.

But it will be not very soon. But Caspian (EV6 chipset for Alpha) from Samsung should be in large-scale production in the first quarter of 2001. The motherboards on it: UP1500, UP1700, UP2700 should appear at one time. All three will support DDR SDRAM (Samsung is the largest manufacturer of RDRAM. And what to do? The clients demand support DDR SDRAM), but the differences start further. If the lowest model supports 700 MHz processors, AGP 2X, UltraATA/66 and 32 bits PCI, the two remained ones are already much more serious decisions: 1 GHz and higher, up to 8 GBytes 200 MHz DDR SDRAM, SCSI and Ethernet, six slots 66 MHz 64 bits PCI.

Memory

It was an interesting month though rich not only with events. DDR SDRAM continued to strengthen its positions. After the release of the first chipset with support this memory type Kingston declared about the beginning of deliveries 200 and 266 MHz modules DDR SDRAM and about the determination to increase their production in the fourth quarter - it is time.

One more similar moment is the official support DDR SDRAM Intel. If it is still not clear with Brookdale, server i870 and Plumas will support DDR SDRAM unambiguously. Add Caspian, chipsets VIA, chipsets ServerWorks … Moreover and the first appeared of systems on DDR SDRAM demonstrating the increase in productivity in comparison with PC133 SDRAM on 5-15 of percents.

At the beginning of August this PC133 SDRAM price at first stopped growing and by the end of month slightly fell. Though Mosel Vitelic for example promises for the fourth quarter the deficiency of 64 Mbits chips and informs that their contract prices for September - October have exceeded $9.

The Korean giants are still increasing the production of 128 Mbits chips DRAM. Samsung since September should lift their production on 50 percent: from 10 up to 15 million chips per one month, though it will still be less than 64 Mbits chips production - 22-23 million per one month. Hyundai copes a little bit worse - now it produces only 6 million such chips per a month, so that its production will be led to 15 million only by the end of year. But for 64 Mbits the revolutions exceed Samsung twice-the company produces 44 million such chips per a month and is going to reduce this digit by the end of the year up to 41 million.

Actually this is all the news. And now a few words about the most interesting fact - the great activity of the lawyers connected as usual with Rambus. In middle of the month the company declared about the next claim this time concerning one of the largest manufacturers SDRAM and RDRAM - Infineon. The essence is all the same - to force to pay fee for SDRAM. Besides the debatableness of the quarrel with the company having the quite good patent briefcase and capable to fight for its own hand there is a question - why to annoy the industry knowing about the preparing claim?

At the end of the month two claims from the manufacturers DRAM ¹2 and ¹3 were declared one after another, Hyundai and Micron accuse Rambus according to the antiexclusive enactment, having required the cancellation of their patents on SDRAM. By the way Hyundai directly has declared that it has made it because Rambus intended to submit to court and on it, so that the company had to answer first. Certainly such claims sometimes last for years, but the situation all the same is very hard to be solved. Hyundai and Infineon have nothing to lose, Rambus probably as well.

Video

The month was hot enough and as usual generally thank to nVidia. At first GeForce2 MX has worked well. There is Spectra F11 from Canopus here which is not distinguished by adherence to the standard design copying and Innovision Multimedia which first declared PCI cards on this chip and the similar Prolink declaration followed this.

By the way we can already see the continuer capable to cause the new group of the cards GF2 MX Pro announcement. Nothing is known about the clock rate, though it is possible to assume its insignificant increase, and SDR memory will be changed to DDR. GF2 MX Pro will become the leader on productivity in comparison with close by the price GF256. As for its price - this is the second question.

It is supposed that on October, 30 GF256 will be retired, so GF2 MX and GF2 MX Pro will become low products in bar GeForce.

In the second half of the month NVidia released GeForce2 GTS Ultra. Overclocked up to 250 and 460 MHz the chip and the memory gave the outstanding productivity 1 Gpixels per second and 2 Gpixels per second and the outstanding price in $499. But nVidia can allow it - Voodoo5 6000 capable to make it a real competition is not released yet in spite of all promisings 3dfx. Simultaneously with the declaration of the new decision from nVidia Creative, Hercules, Leadtek and ELSA declared their products on its base simultaneously with the declaration of the decision from nVidia.

By the way about ELSA. In August the talk that the company stops the production of graphics cards appeared. The cards for the company will be produced by nVidia. The categorical disproof was received immediately. In the market of the professional decisions for graphics servers: nVidia will work with the cards design and OEM, ELSA uses the reference-design and does not work with OEM. The market of non-professional cards: ELSA uses the reference-design from nVidia and also continues to sale cards and to support the users.

But we shall return to GF2 GTS Ultra. It was already said about the complexity of the situation. It is not clear now whether to consider it the autumn product or to wait for NV20. They say that today - GF2 GTS Ultra and in spring - 0.15 microns NV20. It is matched both with other facts (the plan for TSMC change to 0.15 micron technology) and with common logic - why should nVidia hurry up? All the same it seems that nobody is preparing something revolutionary (where are Matrox, 3dfx and ATI?)

And in the meantime, in waiting NV20, nVidia has decided to try itself on for a long time a field of actions, having revenged 3dfx for that history with multitexturing. They has submitted to court on 3dfx, accusing that in breakdown 5 patents, and demanding the prohibition of sales of cards of bars Voodoo3, Voodoo4, and Voodoo5. The keyword in this case, is obvious Voodoo5. And key digit, on all visibility, - 6000.

Though Voodoo5 6000 which month continues to hang in air, 3dfx nevertheless absolutely for nought of time does not lose, dilating bar Voodoo4/5 - was released PCI variant Voodoo5 5500, in September it is expected PCI of variant Voodoo4 4500. It is a pity: and August the situation with trilinear by a filtering for these cards was explained: Voodoo4/Voodoo5 officially do not sustain simultaneously it and multitexturing.

So that we are waiting for Voodoo5 6000 and then for Rampage and the products of purchased 3dfx Gigapixel. They make a rather good impression: they say GP-1 makes GF2 GTS DDR consisting of five times less quantity of transistors (15 against 3 millions) and by the tile architecture providing acceptable productivity at usage PC100 SDRAM as the video memory. GP-3 and GP-4 are being developed. Quicker! Or nothing will remain from 3dfx in a couple of such quarters. Or nVidia earning much money for this market (the income in this quarter is 170 million dollars) will just purchase it.

As for the rest of the companies everything is calm and quiet. ATI after the release of cards on Radeon256 declared ALL-IN-WONDER on its basis. All is habitual but the possibility of audio lead through S/P DIF output is something new as well as DVI output.

In August its neighbour kept silence. They say that the next cards Matrox use very fast but rare enough FCRAM and DDR FCRAM as memory.

By the way both canadian companies as well as nVidia have declared about the support of again declared on this IDF AGP 8X. Frankly speaking nobody from the marketing department could explain to the buyers in what way AGP 4X is better AGP2X. The situation with AGP 4X is repeated - the autumn chips should support the new specification, though there are no chipsets with its support. Moreover both ATI with Matrox and also 3dfx this time according to our information negotiated with the Taiwan manufacturers of videocards about cooperating. The attempts to destroy this monopoly of nVidia were undertaken by them and earlier but in vain. We shall see how this round will end.

But S3 is out. In August the company has officially declared closing the business on graphics cards production. As for chips - that department is constantly being sold VIA.

Disk drives

The main news of the month is the representation on IDF Seagate, APT and Vitesse of the first hard disk supporting 1.5 Gbits per second Serial ATA. Naturally there are no motherboards with support SerialATA in the market today is not present, that is why the help APT and Vitesse is necessary. The creation PCI controller of this interface was required. The development of the final version of the specification should be completed this autumn, and taking into account the well worked question with backward compatibility its appearance in the market should not take a lot of time.

As for more traditional hard disks the situation here was interesting enough. This month the new models were not declared. This month it was declared about the beginning of delivery declared before hard disks. For Maxtor they were the bars DiamondMax 80 and DiamondMax VL 40, and for Western Digital - 7,200 rev/min Caviar on the basis of same 20 GBytes of disks. 7,200 rev/min hard disks continue to capture the market, and in the first quarter of 2001 as it is expected the symbolic board will be overcome - the demand for 7,200 rev/min models will exceed the demand for 5,400 rev/min hard disks.

Quantum in August delivered nothing, it declared about the next development. More exactly just about its completion - about the completion of the specification IEEE-1934 hard disks for home electronics engineering development. Let's keep it without any comments - Sony with WD for already a long time work in this direction, and not long ago Sony declared the first digital video-tape recorder.

With drives CD - RW the business was much more cheerful. At first from the card CD - RW one more manufacturer disappeared - Hi-Val has declared about the bankruptcy (already the second company in this sector declaring about the bankruptcy lately - recently Smart&Friendly has done it). Secondly Elantec Semiconductor, one of the large manufacturers of controllers for a CD - RW, has declared about the creation of the chip giving the recording/overwriting possibilities with the speed 20X. And now compare all this with releasing in September 1.3 GBytes specification DDCD … Despite of the supporters DVD-RAM loud declarations, the CD - RW by no means is going to die.

Let's return to the realities of life. There was enough CD - RW drives this month from all their largest manufacturers: here we can see HP with 5 new models (IDI, SCSI, external, in-house) with the combination 12/8/32 and Sony with external Spressa (SCSI-2, PCMCIA, USB) with combinations accordingly from 12/8/32 up to 4/4/6. Here we can see Yamaha with CRW8824FXZ with FireWire interface and the speed 8/8/24 and QPS c 12/10/32 FireWire Que! and at last AOpen with ATAPI 12/10/32 CRW-1232. So you can choose whatever you like. Somebody can prefer CD - RW/DVD-ROM combine MP9120A from Ricoh - 12/10/32 + 8X.

Also it is necessary to mention only one renaming. In August Clik! turned in PocketZip. Îh, those marketing departments …Why didn't Iomega like the old name?

Write a comment below. No registration needed!