Intel to ship 45-nanometer chips in 2007 - Page 2

Oh shoot! I forgot about the keyboard! What's different about it? It's still QWERTY right?

The primary thing for Intel is to maintain its manufacturing lead. If they continue to do that, Apple will be more than fine using Intel parts. Seeing as the changes are to the server roadmap parts that use the downhill side of a fab's life cycle, I don't think this will hurt Intel's fab advantage.

I also agree with Programmer about the memory controller. It isn't as big a deal as you make, or AMD or others make. What is truly hurting Intel is that Prescott completely blew up its power budget and resulted in Intel putting a ceiling on its clock rate. If it shipped at 4 to 4.6 GHz clock rates (which it can), nobody will be lustfully wondering when Intel will use an integrated memory controller because it would be competing very well against K8 without it, and everyone will be thinking how cute it is with AMD making gains in the 4+ socket server market. It's the only place that the on-chip memory controller offers a true advantage to AMD.

Prescott is a reminder to Apple that anyone and everyone can fail though. The advantage with Intel, though, is that it has the ultimate safety net with its fab power.


It's a pretty serious assumption that AMD server systems will be cheaper or higher performing. in 18 months time.

So, you just add 64-bit to the list of qualifiers and suddenly AMD is king of the mobile world?

Umm, no. Core Duo is going to destroy this AMD Turion "X2" chip, 64-bit and all. It's a 90 nm chip competing against a 65 nm chip! Intel will sell 20x more Yonah chips than AMD sells dual-core Turions. And I'm being generous I think.

I'm wondering how many 35 Watt TDP "Taylor" parts AMD will be able to ship a month. 1000? If they can ship volume quantities of a 90 nm 2 GHz 35 TDP part, I'll be very impressed, but I'm really doubting it and it's just a boutique release waiting on the 65 nm fab.

Just like on the desktop, right?


Which will soon migrate to 65nm with help from IBM, who do happen to have some expertise in this area, or so I hear.



Since when is quantity = quality?


I don't pretend to have any idea how many chips they will ship, but it's not all in the quantity. Intel ships more P4s than AMD and it's Athlon64, but ask any knowledgable person which one is better and you might get an answer that starts with A and end with D.

Intel is strong in the notebook arena, but it's not the only serious contender for the market.

Um, no? Why would it be just like the desktop? The market is totally different. Intel has a competitive product. Has a successor product that looks to be more competitive. And has a manufacturing advantage.


Define soon. I'm hearing Q4 06 pre-Christmas, ie, December 06.


Quantity in this case is an observation of yield. AMD's 90 nm fab can produce 35 Watt TDP dual-core parts by cherry picking the best power optimized parts, the parts that operate at low voltage and higher clock rate. But this cherry picking means that there will be very few of them, say less than 10% of the chips on the wafer. The higher the clock rate, the fewer there will be.

This has been done plenty of times before by both companies, but it's a marketing exercise and not a serious attempt at filling market needs. Ie, a boutique product. We will see what clock rate it will debut at. If they want to be competitive, they'll need to compete with a 2.33 Yonah. If they can only ship 2 GHz, there's not much point in shipping it. If it is only 1.8 GHz, AMD won't ship it.


You are using the term serious pretty loosely.

It's mostly QWERTY. But it has a smaller space bar, few extra keys for Japanese-language input, and some characters have been moved for some unknown reason (e.g. the quotation mark and the @ has been moved to another locations on the keyboard.)

And AMD has a competitive product. Has a successor product that looks to be more competitive. It doesn't have the manufacturing capacity Intel has, but that speaks nothing of the actual quality of the chips that come out of AMD fabs.



1H of 2006 is what I hear.



AMD seems to have partnered with IBM, who do happen to have the capacity to produce more than a few low-voltage high-clock rate chips.


2.2 to 2.6 Ghz.




And you're understimating the company that just captured 20% of the market in both desktop and notebooks.

GC The notebook market is a little different than the desktop market, don't you think? The chip with a nice GHz that gives the most battery life will dominate (with emphasis on battery life). Throw GHz out because they are all close enough and I believe Intel will continue to have a chip that performs better in the battery life arena than AMD and will continue to outpace AMD enough to maintain a solid marketshare in the notebook market.

True, but some people have been saying that Intel Dual-Core desktop chips are going to "destroy" AMD X2 dual-core chips, and... that didn't materialize. Actually, it was a far cry from being equal to, let alone destroy.

Anyway, AMD Turion "Taylor" is not out yet, so there's no direct comparison performance and watt wise. I guess we'll have to wait and see what it delivers.

noticed high end Server processor still manufacured in 130nm and 90 nm...

thought that smaller the size it is easier to bump the L2 Cache memory, it seems server CPUs are the last one to move to the new fabrication process like 90nm, 65nm then 45 nm?

reason being?

Because servers do best with tested and trusty parts, on a stable platform and manufacturing technique.

A 2 GHz Yonah is about equal to a 2 GHz Athlon 64 X2 in performance, but at half the wattage. 65 nm manufacturing allows Intel to manufacture Yonah at higher clock rates, at lower watts than AMD can with Taylor at 90 nm. Ie, Intel will be able to ship 2.16/2.33 GHz Yonah chips in Q2 06 at about 35 W TDP while AMD will probably ship a 1.8 GHz part at about 35 W TDP. Ie, it's better quality.

No, I don't really see AMD competing for laptops until they move to 65 nm.


We will see. If AMD is truly able to, we would hear much more of it then we are. I'd be afraid that they can't ship in 2006.


IBM doesn't produce any low-voltage high clock rate chips. They produce a lot of high wattage high clock chips (Xenon, Cell, 970), or even higher wattage lower clocked chips (Power5+). The so-called power-optimized 970fx chips are really just lower wattage really low clocked chips.

The best they have appear to be a 1.8 GHz 970fx at 37 Watts max and a 1.6 GHz 970fx at 21 Watts max. Somewhere between 1.6 to 2 GHz, the 970 architecture really begins to eat up power.


I'd be very curious at what wattage. I'd estimate a 2.2 GHz to be about 45 W TDP at 90 nm.


They moved from up from 16% to 19% in 2005. Impressive in the face of the 800 lb gorilla. But no, I don't think I'm underestimating them. The power will be in the fabs, smaller companies can't afford them by themselves anymore, and Intel has the money to move forward the quickest with other companies lagging further behind.

If Merom bombs like Prescott did, they Intel is in trouble. If it is successful, AMD will lose marketshare again. No one seems to be saying that Merom is not living up to expectations.

AMD existence is very important for PC world to keep the monopoly away and create new good products ... especially now IBM now no longer doing the PC CPU business...

As far as Apple is concerned, you can pretty much forget AMD processors. Intel is the Mafia of technology companies, they're even worse than Microsoft. Once you partner with Intel, you don't make deals with their competitors. Only a few big companies (like HP) have been able to do it.

Intel's long history of brutal, thuggish business tactics is one of the things that sickens me about Apple's x86 transition. It's a matter of record, Intel doesn't deny a bit of it - they say it's completely legal.

You watch - Apple will never use AMD.

Full ACK!

A candid sentiment, to be sure. I find reading books about technology companies to be interesting. But, I have not seen one on intel. Have you got some reference to a book on intel? I will have to do some looking around on intel history articles.

By then, won't we have switched from calling them nanometers to picometers, resulting in 10pm chips?