Tom's Computex News
Редакция THG,  7 июня 2001


Introduction

Computex Team
Tom's Hardware's Computex 2001 team - Patrick, Tom and Omid

Another year has passed and once more, the IT world has made its pilgrimage into the far orient, the land of the motherboards, crazy cab drivers and Happy Kitty. It's my fifth visit to Computex, which is held every year at the beginning of June in Taipei/Taiwan R.O.C. This year's Computex was clouded by big product releases of NVIDIA and AMD, which kept me from starting my Computex reports any earlier than today. This computer show that once was some kind of secret tip for OEMs, system integrators and geeks has now become the most important event of the IT industry, making Comdex look extremely lame and CeBIT seem like a boring Moloch of gigantic size, but very little appeal.

The good thing about the fact that I had to work on huge articles about nForce, AMD760MP and Athlon 1400 is that now, by the time that I can give you my side of the Computex story, the simple and weird stories about this event have already been told, giving me space to focus on the interesting stuff from behind the scenes that is only available to a small minority of show attendees and only to those ones who have been here numerous times before.

Before I will start to cover all the motherboard companies and their new products, I would like to give you some information that I haven't seen anywhere else yet and that might help to shed some light into two of the hottest topics of Computex 2001. Only the subscribers of our new newsletter were given this information already 24 hours ago, so you might want to consider joining the newsletter list as well.

Tidbits Of NVIDIA's nForce

nForce

On Monday NVIDIA joined the growing league of core logic providers for PCs when it launched its feature packed new nForce chipset, formerly known as 'Crush'. We dedicated two articles (Computex: Nvidia declares war on Intel, Computex: NVIDIA nForces Success In New Market) to this highly important event, which might as well start a new era in the PC business.

Repackaging?

After NVIDIA's forceful presentation of nForce on Monday in the rather badly chosen Grand Formosa Regent Hotel 30 minutes away from the show, the super-chipset has become the topic of hot discussions and nasty rumors all over the show floor. The first rumor originated from companies that are not exactly friends of NVIDIA's bold move into the integrated chipset business. It said that current nForce version would require a very time-consuming repackaging process to add more power and ground pins to the chips, due to signal noise issues. The smirking sources of this rumor explained that this would most certainly delay nForce's planned launch date in the July/August area by quite some time. We decided to get the truth right from the horse's mouth and spoke to NVIDIA's nForce product manager Michael Lim. He was able to refute the rumor, but admitted that the repackaging had indeed turned out to be necessary. The chips presented at Computex however, are already of the new design, so if there should be any delay, it won't be due to this issue.

Positioning?

Another doubtful person was wondering how nForce is supposed to be positioned. The high-end is supposed to be reserved for the clock rate leader Pentium 4, since the majority of decision makers wouldn't care about performance as much as about clock rate. However, customers of the upper midrange segment, which is reserved for powerful Athlon-systems, would not really be too keen on integrated graphics and rather go for powerful add-in cards. For the normal midrange he considers nForce420D as too expensive, which in his opinion is the same problem that nForce220 would have in the value segment, carrying a price tag twice as high as VIA's KM133. Let me add that the person who expressed the above concerns happens to be in the upper management of a chipset maker that has its own chipset solution with integrated graphics. However, those concerns seem to be justifiable, since I could not find any of NVIDIA's five motherboard partners that had a convincing solution to the problem.

Current Bugs?

It is confirmed that the current nForce spin still has a few issues, which involve ACPI and USB functionality. However, the next chipset revision is supposed to arrive by the end of June and its stepping 'B01' suggests at least semi-final stage.

Support Of Intel Processors?

Another impressive fact is that nForce in its current version is actually not only supporting AMD processors. In actual fact you can just as well run Intel's Pentium III with it and be happy, which might not be quite as surprising if you remember nForce's close family bands to the Microsoft's upcoming Xbox, which, as we know, will be powered by Intel's Pentium III processor. We are not supposed to know about this, because Intel turned down NVIDIA's request for a front side bus license for its illustrious Pentium-line. So far about the rumor about NVIDIA's pro-active decision to build a strategic partnership with AMD. It looks more as if the guys didn't have a choice. Time will tell if Intel made the right decision for a change. We have our doubts.

Outrageous Pricing?

Finally the first information about nForce's pricing have emerged as well, and it is slightly surprising, to say the least. Motherboard makers are far from happy and wondering how they are supposed to sell their shiny nForce products. It seems as if NVIDIA is asking for no less than $60-70 per DonForco (nForce420D) chip, which marks a new record for mainstream chipsets. The 'value'-version of nForce, the nForce220 with the castrated 64-bit memory bus, doesn't come exactly cheap either. So far, NVIDIA is asking for some $40-50 for the nForcillo. Taiwanese industry sources consider the price of nForcillo (the 220) boards to be in the range of high-end OEM boxes, while the price of motherboards with DonForco would make them retail-only products. With this information, it is not too difficult to calculate the price of upcoming nForce motherboards. Full ATX motherboards with nForce420D would come at a distribution price of about $150-160, while MicroATX boards with chip might manage to reach $130. These prices put current i850 Pentium 4 motherboards to shame! NVIDIA will have a lot of explaining to do if it wants to make OEMs or retail customers understand this astronomical price. This could easily cost the marketing dollars that would be saved if NVIDIA should come around and lower the price of nForce.

NVIDIA's direct comment was very nice and I've got to admit that they might have a point: "why wouldn't you pay a little more for what you get?" Well, I guess we first need to know exactly what we will get. Our upcoming testing article will hopefully give a very good idea.

Northwood and i845 Madness

Northwood
Intel's sweet little 'Northwood'.

If there is one thing ubiquitous at Computex 2001 then it is motherboards with Intel's upcoming i845 chipset and Pentium 4 processors with the upcoming 0.13µ 'Northwood' core. While the big announcements of the show were NVIDIA's nForce and AMD's Dual-AthlonMP or Athlon1400, the real ruler of the show is Intel, and that with products that haven't even been released yet.

i845 Boards
No other motherboard maker had as many i845 motherboards on display as MSI.

The officially presented i845 boards with Socket423 and the new Socket478 were all designed for PC133 SDRAM only, which is obviously not the kind of platform that would help Intel's next flagship processor reach new performance heights. Behind closed doors however, every motherboard maker was able to show its i845 motherboards with DDR-SDRAM support. i845 DDR-SDRAM solutions are supposed to run just as stable as its PC133 siblings, but Intel seems to have strategic reasons why those boards are meant to live a shadow existence right now. I don't want to get into any speculation as about why the chip giant is holding back on this fully viable product, since we have heard and read enough conspiracy theories already. Fact is however, that right now Intel advises to run DDR-memory at only 200 MHz in those boards, while we were assured that 266 MHz DDR is operating just as stable as well. It is expected that i845 will support PC2100 DDR SDRAM once i845 will be released.

Intel's current plan is to launch i845 for PC133 by the end of August, while the current release date of i845 for DDR-SDRAM remains the beginning of 2002. Most people I spoke to are expecting however that Intel will move i845 for DDR up as far as to the beginning of October this year, which would only be one month after its PC133 sibling.

All motherboard makers agreed that they see DDR-i845 as the next BX-chipset, because it provides high stability and good performance. Right now i850 seems to be a bit faster than DDR-i845, but we will see if that might not change down the road. If Intel should finally drop RDRAM and move DDR-i845 up to October 2001, there is no reason to slow the DDR-chipset down artificially. However, if RDRAM should indeed remain, Intel will position it at the high-end to justify its price premium. In this case it would be helpful if DDR-i845 was either not even there of if it was at least slower than i850.

Personally, I think that Intel won't have much of a choice. The current market share of Pentium 4 in the processor business is less than 10%. DDR-i845 would be able to change that. I am sure that Intel has got a lot more to lose than the good old Rambus-deal that has been speculated to be a reason why Intel has been holding back DDR-i845 so far.

КОНЕЦ СТАТЬИ


Координаты для связи с редакцией:

Общий адрес редакции: thg@thg.ru;
Размещение рекламы: Roman@thg.ru;
Другие координаты, в т.ч. адреса для отправки информации и пресс-релизов, приглашений на мероприятия и т.д. указаны на этой странице.


Все статьи: THG.ru

 

Rambler's Top100 Рейтинг@Mail.ru