Processors

Here the month began from some problems both for AMD and for Intel. However AMD owes for its problems in July to Gateway whose problems with own motherboards have forced the company to refuse for some time usage 1 GHz Socket-A Athlon in the systems and for some time questioned reliability Thunderbird and KT133. Then the situation was clarified but it was a bit unpleasant.

However for Intel July was even more interesting from this point of view - company simply has cancelled output of two server processors. Just now Pentium III Xeon 700 MHz with 1 and 2 Mbytes of the cache L2 on the chip was produced as the time to produce 800 and 866 MHz versions of the similar processor has already come according to the plans. And then the clients began to object having required to give them time to test for the beginning systems on the basis 700 MHz variant and only after that to move somewhere further. Thus only 900 MHz processor that will appear in 2001 approximately at the same time with x86 server processor of the next generation-Foster will become the next perfect PIII Xeon under the plans Intel.

In general the month was quiet enough, from most significant events it is necessary to mark only planned fall price on processors Intel - on July 19 prices on model Pentium III above the average level from 700 up to 933 MHz have considerably fallen from $214 up to $669 as Pentium III 600-667 has remained the same minimum level - $193.

The following fall price on Pentium III is expected in September, by the way then rather serious on plotting scales the similar action is expected and from AMD - it is supposed for example that 1 GHz Athlon then will fall up to $495 and its 800 MHz colleague up to $190. Probably September fall price Intel will be the adequate answer.

And from real events it is necessary to mention only output of the new virtual processor Intel - 1.13 GHz Pentium III which limited deliveries began from the beginning of August. Thus Intel the second time became the leader and it is possible to expect that Athlon AMD the second time caught it at the end of August with its 1.1 GHz will produce then more processors on this clock rate. Here it is necessary to take into account the fact that Fab30 grows and by the end of the year will have a half of rated power under the forecast of AMD.

And what about Pentium III? On 1.13 GHz by the way the life of the veteran is not ended. In 2001 0.13 microns Tualtin with 512 Kilobytes of the cache L2 and clock rate from 1.2 GHz calculated on frequency of the system bus up to 200 MHz should be produced together with Pentium 4 (both Wllamette-423 and Wilamette-479). Presumably a maximum can be in about 1.4 GHz that will allow to close a breaking on a clock rate between Pentium III and Pentium 4 and to provide the large attractiveness i850.

By the way Israelean Fab18 is already seriously engaged in produce Pentium 4 and the clock rate of chips is rather far from stated 1.2-1.4 GHz and is 1.7 GHz. It seems that Intel tries to not give even possibility for matching operating on the same clock rate Pentium III and Pentium 4.There won't be normal Pentium III Xeon, Foster will appear only next year, mass availability of systems on Itanium is expected only in the first half of 2001… In such conditions a store even on a clock rate will really be helpful.

Sun in this case looks extremely advantageously because it has assured that due to unexpected progress UltraSPARC III which output is planned this quarter will start not on 600 MHz as it was supposed earlier but at once on 900 MHz. That is especially interesting, taking into account delays of the competitors with their power processors - Intel c them Itanium and Samsung c 1 GHz Alpha.

However delays Intel and Samsung mean nothing in comparison with the project under the title of Elbrus - 2000.It is still searching the investors and partners for production being improved during this search: the last information from the developers describes it in different variants from 0.18 micron of the chip with the cache L2 2 Mbytes, square 360sq.mm and clock rate 1.2 GHz till 0.10 micron SOI of the chip with the twice large cache square 310 sq. mm and clock rate 3 GHz. It is necessary to find the manufacturer.

Motherboards

This month is interesting and replete enough. Even 440BX was marked, the last cards on it seemed to be produced within the last months. However it was not enough for the manufacturers of motherboards to have the support UltraATA/100 and the models like Abit BX-133 RAID and SOYO SY-6BA+100 have appeared: with official support frequency the system bus 133 MHz and UltraATA/100 RAID by the controller onboard.

It is difficult to expect that 440BX will live up to SerialATA so its history probably is finished and we can consider the card like named above the top of its evolution. By the way some words about its evolution. In July Intel left the first model of the motherboard on this chipset - SE440BX. Really today Seattle of a sample 1998 does not maintain any criticism at matching with the last models of cards on same chipset.

Apollo Pro 133/133A is becoming a bit old. Everyone who wanted has already presented the models on it and now all advertisements in this area are about separate products like cheap Abit VL6 for system integrators or dual-processor Gigabyte GA-6VXD7 as an exotic decision for inexpensive servers and workstations. On the other hand at all its disadvantages in cases when except double CPU configuration the support of frequency the system bus 133 MHz is necessary and 440BX is not used Apollo Pro 133A suddenly becomes a rather good variant.

KX133 is in error and it seems to become a chipset with a record short time of life fast having given the place to KT133 and new DDR SDRAM ñhipsets. In the beginning of July still there were findings - out which motherboards on this chipset work with Thunderbird and then it was clarified about problems with A3D sound, with ECC SDRAM. About problems with graphics it was known earlier… In common the impression about the last child VIA is not too good.

There are no such problems with KT133 and ñhipset is newer so that models on it were declared much more actively though the fundamental spike was in June. Nevertheless someone has lagged behind the fundamental stream like DFI, Abit or Transcend, presented their cards on this chipset and someone has already presented new variants of the cards on it like Soltek declared SL-75KV-X (variant SL-75KV with voice diagnostics) or Asus with A7V-VM (rather integrated variant - microATX with Ethernet 10/100 onboard).

VIA is feeling rather optimistically concerning to its chipsets and has decided to get their fifth manufacturer - Hyundai, which presumably should receive the order for production 0.35 microns ñhipsets already in the end of August - the beginning of September. In general the business for VIA is recently excellent - for the last year volume sales of the company increased more than three times and soon it would be possible that the large part of sold ñhipsets under processors Intel can infringe the rights Intel.

There are some problems with Rambus… It was easier to reconcile with VIA than to continue trials in courts - at the beginning of July the companies have declared about the agreement on the fundamental questions and distributing frame - licensing. Under the terms of the agreement Intel receives from VIA the sum at once and some percent from sold P6 chipsets. In addition Intel dilates the available agreement on licensing from 1998 (that one because of which breakdown the companies were judged) so that it can reflect realities of today.

Intel stable continues to support DDR SDRAM from the upper side supporting ServerWorks with its server chipsets technology and from lower side - forgiving VIA, ALi and SiS with their from chipsets of the average level. It seems that money received as a result of the bargain with Rambus doesn't not pay back losses, which are carried by the company that goes against market demands.

One more confirmation is i815 which was met by the market with the larger enthusiasm than produced at the same time with it i820E. Naturally it is not absolutely correct to compare these ñhipsets directly though… Intel itself has compared them on the site where in 9 tests from 11 the system on the basis i815 has shown the larger productivity. So that it is no wonderful that on i820E it is not declared any card (!!!) till now that seems to be a record and the cards on i815 continued to go by continuous stream: SOYO SY-7ISA and SY-7ISM, Transcend TS-ASL3, Epox - EP-3S1A, Biostar M6TS for the second month. AÎpen was in general marked at once by 3 models from High-End AX3S Pro up to cheap MX3S.

Remaining integrated chipsets do not have such mass support though they are not forgotten. SOYO presented in July Socket-A a card SY-K7VMM on the basis KM133 with the integrated kernel Savage4, Soltek presented the model on the basis of its analogue under the bus GTL+ - SL-65MV on PM133, Asus declared the motherboard CUA on the basis Aladdin TNT2.

But the essential of month is nevertheless chipsets with support DDR SDRAM which at last after the year of waiting began to come to the market. VIA was not the first company presented them, it was a modest and silent ALi, presented on July 26 at once 4 such chipsets - under EV6 and GTL+ in classical and mobile variants. All 4 of them support PC1600 and PC2100 DDR SDRAM with maximum volume up to 3 (!) Gbyte, naturally AGP 4X, UltraATA/100 and other expected possibilities.

The price pleases, the variants ñhipsets for desktop PC cost accordingly $ 31 and $ 33 for ALiMAGiK 1 and Aladdin Pro 5 that in combination with the prices on DDR SDRAM gives the possibility of obtaining systems operating this type of memory and comparable with usual SDRAM systems. ALi promises mass production only in the fourth quarter, then analogs from VIA, SiS and AMD should approach so the declaration a card on AMD-760 is expected already even since September.

Moreover it is already possible to not doubt that DDR SDRAM chipsets for Pentium 4 which originally was planned by Intel as RDRAM-only chipset will exist. At first on Platform 2000 the Taiwan triple has declared about the determination to release similar ñhipsets and VIA has declared that it will release such chipset without dependence from the receiving a permission from Intel.

Then Intel officially has declared that next year its own chipset for Pentium 4 with support PC133 SDRAM - Brookdale will appear and couple of days later it was clarified from the Taiwan manufacturers of motherboards that Brookdale should support not only PC133 SDRAM but also DDR SDRAM. And then Intel again officially has confirmed that it considers a question about creation its own ñhipset for Pentium 4 with support DDR SDRAM.

It seems that the history approaches the logic ending. It was time when Intel considered DDR SDRAM as memory for 440BX. If there was no signed agreement with Rambus the history would be different. Especially taking into account that the development Advanced DRAM Technology has already started so that is impossible to call the future RDRAM memory in any case.

Memory

Basically nothing too fatal and defining was happening during this month - in fundamental the subject links of the previous months were finished. The manufacturers of memory continued conversion their production on 0.18-0.175 microns, on Taiwan the majority of them are finishing this process and under the statement Powerchip conversion from 0.25 micron on 0.18 micron will allow them to reduce the cost price of production chips DRAM by 40 percents that in combination with a rise in prices on memory will allow to provide rather good profit.

And the prices for memory really confidently continued to grow and more than appreciably: only for July for example on the data ICIS-LOR average prices of 128 Mbytes the modules PC133 SDRAM have grown world-wide more than on 10 percent. Micron at the end of July has declared about raising contract prices of 64 Mbit chips SDRAM above 8 dollars (some months ago prices were at a level 6 dollars) and as well as other manufacturers assumes that in future rise in prices will proceed.

In the field of technologies all there is only development of already existing tendencies. In this schedule for example representation Kingmax 150 MHz SDRAM - rather specific product which usefulness to someone except the fans of extreme dispersal on the first view is hardly obvious-is characteristic enough.

One more exotic sort PC133 High-Speed SDRAM (HSDRAM) demonstrated on Platform 2000 by Enhanced Memory Systems in structure of the prototype an inexpensive workstation. Under its applications the productivity of a similar system was for example in OfficeBench on 28-49 percent is higher than for similar systems configured with PC100 SDRAM and PC800 RDRAM. The case with the tests on a site Intel was already the second case for July when it was proved that SDRAM has the large productivity in the majority of real situations and in the tests too

By the way on the data of poll opinion conducted by Semiconductor Business News on Semicon West Trade among the manufacturers of testers RAM the demand on equipment for testing DDR SDRAM considerably exceeds demand for testers RDRAM. On the data Teradyne for example 80 percent of their high-speed testers bar Aeries are got under DDR SDRAM. Practically nobody hurries up to place the additional orders on testers for RDRAM and manufacturers of the similar equipment overestimated perspectives of the market have to look for other markets where the testers 1 GHz can be necessary-for example SRAM.

Something is happening between Rambus and Intel by the items of information arriving all month. The agreement between the companies from 1996 was reconsidered: according to it for obtaining 4 millions shares Rambus at the price of $2.5 it was required from Intel during two next one after another quarters to provide a part of chipsets with support one and two-channel RDRAM in volume all ñhipsets not less than 20 percent. In the new variant it is necessary to sustain such part during any two quarters.

Then in middle of the month it was rumoured about the antitrust claim to Rambus preparing by the largest manufacturers of memory in August with the purpose to consider void the patents of the company in connection with SDRAM and moreover Intel was supposed to take part in the consultations on this question. And there is nothing to say about the claim made by Intel at the end of the month that it considers the question of creation DDR SDRAM ñhipset for Pentium 4.

Rambus in the meantime continued to operate in its old way collecting the manufacturers DRAM of the lower quality level and at the end of July it was declared about acquisition of the license on SDRAM Oki Electric and that Mitsubishi may be the next. Probably in future we will find out about for what reason the Japanese manufacturers DRAM one by one surrender at Rambus but now it is possible only to guess about reasons of such territorial dependence.

However Rambus has shown itself in July and in technological sphere approximately at the same level as Kingmax. It has dispersed up to 150 MHz SDRAM and Rambus - RDRAM up to 1,066 MHz having magnified thus transmission capacity on third up to transmission capacity 266 MHz DDR SDRAM - 2.1 Gbyte/s. Very soon Samsung and Hyundai have declared about determination to present such chips saying nothing about their price.

Video

Traditionally this section has a lot of news. During July a lot of leaders from this section of the market were noticed.

nVidia remained the leader again. The month began with continuation of the declaration cards on GeForce2MX - Leadtek WinFast, Gainward CARDEXpert, Asus AGP-V7100, Abit Siluro, Prolink PixelView… They are similar to each other implementations of standard design from nVidia. More - less pleasant exception is Hercules declared in June its card. It uses combination 175/183 MHz in 3D Prophet II MX instead of standard combination 175 MHz chip and 166 MHz memory receiving a rather appreciable increase in productivity in comparison with cards on the basis of standard design.

However we know this standard design. As it was clarified in July any card on GeForce256 can be converted into a card on Quadro by resoldering of several resistors and it won't be only a simple variation of ID the chip - the system really in some professional applications will demonstrate magnifying of productivity in some times.

Now it is time to think what it is necessary to resolder on cards with GeForce2 GTS - at the end of July the new variant of the professional chip from nVidia-Quadro 2 was produced and it marked a new way in policy of the company: nVidia for the first time officially began to product cards on own chips. (If not to take into account partnership with large OEM in connection 64 Mbytes of cards on GeForce256). However the company will not compete with its own clients: only ELSA has expressed interest in Quadro1 and the cards on Quadro 2 besides nVidia will be produced again only by it.

As for 64 Mbytes variants of cards… It is interesting that the leading partner nVidia on GeForce2 GTS Creative Labs has given to understand that it is not going to release on the retail market 64 Mbytes variant 3D Blaster Annihilator 2 - only for large OEM. It is not absolutely clearly what such difference in marketing policy from the competitors who have declared such models is caused by.

However 64 Mbytes is dangerous enough. ATI has produced 64 Mbytes Rage Fury MAXX and so in result it had to sell it actually at cost price- $84 after reset $50. MAXX doesn't work in double chip mode for Windows 2000.

The truth is that its time is over. It should "keep a defence" on the rate of lead the scene up to the approach Radeon 256 and it was produced in July and that is over. On July 17 three cards on the new chip ATI: 32 Mbytes card on SDR and DDR memory and 64 Mbytes card with DDR cost from $199 up to $399 were officially declared. The two last ones are delivered at once but the first one-the cheapest only since September. A few days later on Mac World it was declared and about the release RADEON Mac Edition exiting in September too.

3dfx with Voodoo3 on the platform Apple felt enough confidently and Voodoo5 should only strengthen its position. Though Voodoo5 is not all right: the time of yield Voodoo5 6000 is constantly delayed. Though on the other hand what market do these cards have in comparison with mainstream Voodoo5 5500?

Cards on next chips won't appear soon but the work is being done. It is rumoured that when the shareholders asked 3dfx at the last assembly to demonstrate results of purchase Gigapixel, the company could already demonstrate the first alpha - version of the chip operating with the twice greater productivity than GeForce2 GTS DDR and at switched on full screen anti-aliasing.

The operations are carried on and on G800 and we hope it will be something really worth taking into account that Matrox missed a cycle not having produced G600 this spring and just produced hardly improved variants G400 - G450 that was a cheap decision for OEM and system integrators but not a threat for GeForce or Voodoo5. Moreover and the deliveries will begin only at the end of August.

We can expect a lot of events in autumn. Rampage, G800, NV20, Sidewinder… By the preliminary messages the autumn graphics chip will have productivity more than 1500 Mtexels, the full screen anti-aliasing, hardware T&L, with a clock rate from 200 MHz, attempts to use different technologies of build-down a load on the bus of memory - tile architecture, etc.

Disk drives

The month was quiet enough that however is usual for this section. MacWorld is the only thing to speak about. Fujutsu has declared new bar of magneto-optics - DynaMO SF consisting of two SCSI models made for 640 and 1300 Mbytes disks. Imation has presented 20 Gbytes disk drive Travan with FireWire interface. EZQuest and LaCie have presented the models FireWire a CDRW of drives. LaCie began delivery PocketDrives - bar 6, 18, and 32 Gbytes of hard disks with an overall dimensions 8,9 õ 14,6 cm. and less than half-kg weight.

Naturally MacWorld was not the only thing to speak about. The manufacturers of CDRW continued to declare models with most cool for today combinations of rates: TDK and Sanyo have declared for August 12/10/32 drives (such synchronism however is not surprising taking into account that a basis of both models is the chassis from Sanyo). Sony has a bit lagged behind having declared two new drives of bar Spressa with rate 12/8/32x. Yamaha began delivery CRW8824SZ - first Ultra SCSI of the drive with rate of overwriting above 4x.

In general CD - RW remains on spike of popularity and is not going to leave: the sectioning of Philips occupied in the production of CD - R/CD-RW drives has declared about the schedules to double by the end of 2001 volume production of these disk drives up to 20 millions units grounded on the forecasts of magnifying demand on them. It is expected that only this year more than 35 million similar drivers and 3 billions of CDR/CDRW disks for them will be delivered.

The size even 800 Mbytes of disks becomes occasionally small but the operations and in this area are carried on. The same couple Sony and Philips in July officially declared the schedules on development of the data sheet 1.3 Gbytes compact discs having a preliminary title Double Density CDROM/-R/-RW and promised that in September its development will be completed.

Taking into account that to provide maximum mild and painless junction to operation with more capacious disks the data sheet maximal for the manufacturers of drivers the data sheet relies on existing technologies (the same wave length of a laser beam - 780 nm is used) and also considering that the largest manufacturers of controllers for CDRW drives already have hurried to declare products with support DDCD the new technology has a quite good perspective in the market. We wait for the announcements of new drives on autumn Comdex.

As for hard disks there was nothing outstanding but some progress took place. The situation can be described so in spite of the fact that Maxtor has delivered a record on volume of hard disks - its bar DiamondMax 80 has maximum size 80 Gbytes. Usage 20 Gbytes disks (it is not first here) in combination with usage up to 4 such disks in the hard disk is a usual evolution. By the way besides DiamondMax 80 Maxtor has presented two bars of hard disks - DiamondMax Plus 45 (it is already 7,200 rpm but maxim three disks 15 Gbytes each) and DiamondMax VL 40 - cheap model on the basis 20 Gbyte disks. Naturally all three bars - UltraATA/100. All three models should be accessible this month.

Quantum was not interested in quantity having declared the competitor only DiamondMax Plus 45, but here it left Maxtor behind by usage the same 20 Gbytes disks - in result Fireball Plus AS at the same 7,200 rpm has a maximum volume in 60 Gbytes. In addition technology of build-down noise - technology Quiet Drive Technology (QDT) in combination with liquid bearings gives in a mode of search noise in 32 dB.

Also it is necessary to mention only about a situation with interfaces of data transmission: the developers SerialATA have declared that the creation of the new data sheet comes nearer to the end and this quarter there should be its final variant with transmission capacity 1.5 Gbit/s and further 2x, 4x, etc. Unfortunately we hardly shall see support of this data sheet in ñhipsets of this year: the support UltraATA/100 is already incorporated in all important processors of this year and hardly someone will have time to include also SerialATA there.

Write a comment below. No registration needed!