|
|
 |
|
|
|
 A few days ago, AMD and Intel published financial results of the last quarter ended December 31.
For AMD, sales have slightly increased to $1,374 billion. This represents 3% improvement compared to the previous quarter and 2% improvement compared to the same period last year. In fact, the average price per CPU is clearly dropping and AMD has sold 26% more units compared to last year and 76% in particular for laptops. Profits on this segment are reduced and fall from $272 million last year to $63 million.
 For ATI, sales are of $398 million and the manufacturer lost $13 million. $120 million come from activities tied to mobile products and consoles and the rest is imputable to GPU and chipsets. Even if AMD doesn't really communicate on the subject because of a lack of possible comparison, an interesting thing to know is that one year ago ATI's sales were of $600 million per quarter. For this quarter, AMD published loses of $527 million because of the acquisition of ATI ($550 million).
Intel, however, is felling really well with $9.7 billion sales for the quarter and $1.5 billion profits. Now, the thing is that if these results are good sequentially with respectively +11% and +8% compared to the third quarter, they represent -5% and -55% compared to the same period last year. Intel announced to have broken new records in terms of units sales for processors and flash memory. Chipsets sales are stable and motherboards are dropping. Compared to the fourth quarter 2005, Intel has recorded +9% sales in US but -1% in Japan, -5% in Asie-Pacific and -17% in EMEA (Europe, Middle East, Africa). |
| |
|
 A few months ago, manufacturers started releasing products based on preliminary versions of the norm 802.11n. Lately, Intel has upgraded the Centrino to MIMO with the WiFi Link 4965AGN, a mini network cards interfaced with PCI-E bus based on the 1.0 version of the norm (the 1.1 version has been published a few days ago).
When it is working with a compatible peripheral, according to Intel, it is able to reach up to 5 times the usual transfer rate (300 Mbits /s … theoretical) and have a range two times bigger than products of the previous generation. Of course, it is also compatible with the 802.11a/b/g norm. Intel also said that the autonomy would be improved compared to current « Draft N » products because of lower power consumption in stand by. The manufacturer announced a very optimistic 1 hour of autonomy.
Several OEM including Acer, Asus, Gateway and Toshiba announced that they would integrate the Next-Gen Wireless-N in new laptop computers. Intel also has announced the release of a program called "Connect With Centrino" in partnership with Asus, Belkin, Buffalo, D-Link and Netgear in partnership with Asus, Belkin, Buffalo, D-Link and Netgear, officially to ensure a compatibility and optimum performances between their WiFi equipments and access points. Some will say that the Santa Clara giant is only covering its back because announcing products based on a pre-version of a norm that will only be finalised in 2008 isn't without risks in terms of interoperability in MIMO modes. |
| |
|
The microprocessor giant, Intel, would be planning a come back on the GPU market. We are, of course, speaking of graphic cards since Intel is already the market leader for integrated GPUs. Rumours that started circulating a little while ago about a project called Larrabee are now confirmed. According to Beyond 3D, the project has now an official name: VCG for Visual Computing Group.
The objective of this group is clearly to compete with ATI and NVIDIA for the fabrication of future graphic technologies and Intel has indicated that it will work first on the development of graphic cards based on multi core architectures for high end configurations. On the longer term, Intel will cover all segments including the integration of GPUs in CPUs like AMD is planning to do with the Fusion project.
Why has Intel decided to make some investments to design high end graphic cards?
 As you know, Intel's integrated solutions are often accused of providing low performances. But in fact, from an architectural point of view, they aren't as bad as we could think. For example, the GMA900/950 have higher calculation power than the GeForce 6600 GT and the GMA X3000 is the first PC GPU to have a unified architecture (the GeForce 8800 is the second). The GMA X3000 was also the first GPU to integrate the support of DirectX 10. It isn't however supported in practice. Why? Drivers are the answer. Developing drivers is really complex and requires a large and costly team. So, Intel has chosen not develop DirectX 10 drivers for the GMA X3000 since the final release date made that it wasn't obligatory.
As DirectX 9 drivers are far from being finalised and efficient, Intel is currently focusing on them. This isn't of course enough yet and Intel's integrated solutions often suffer from mediocre drivers. Solving this problem will take a lot of time, require a bigger development team and cost a lot of money. This is also true for small tricks that avoid GPUs making useless calculations. Doing all this work for integrated chipsets sold only a few dollars more than standard chipsets is obviously not profitable. So, if Intel makes this investment, why not exploiting it with a much more complete product range?
Obviously, this is not the only reason. Letting AMD have a better platform because of higher end graphic solutions via the acquisition of ATI isn't acceptable for Intel. In terms of image, a future simplified message like: Intel entry level and AMD high end is probably scary for the Santa Clara giant. Today, this isn't a problem because NVIDIA and ATI's graphic cards can be added to any computer without leading to an "image issue". This is the reason why AMD kept the ATI brand (can you imagine a computer with the logo AMD and Intel side to side?).
Intel could also count on Nvidia for the longer term. The two companies are currently partners as Intel needs Nvidia's chipsets for very high end multi-GPU solutions. Indeed, on one hand including an AMD chipset isn't conceivable and on the other hand Nvidia artificially blocks the use of Intel's chipsets via the drivers. This is the reason why Intel needs Nvidia, but this is only a temporary facade as Intel and Nvidia are relatively arrogant. Intel is arrogant because it is the market leader and Nvidia is because of the ambition of its leaders. In other words, Intel doesn't like being dependent of the "small" Nvidia. As a result, having a great graphic technology could help Intel getting rid of this dependency and not having its chipsets excluded from some markets.
 Another reason for Intel is the maximisation of the fabs' utilisations which is currently dropping. Intel's historical strategy has always been the production overcapacity to avoid taking risks. However, because of the evolutions of fabrications technologies, the level of utilisation has dropped and/or production lines based on older fabrications process have become unused. Building GPUs would help Intel maximising the utilisation of the fabs. It will also be able to use an internal technology and not an external manufacturer like Nvidia and AMD do for GPUs, at least for now.
Clearly, Intel's potential interests are multiple and the current message is very clear for ATI and NVIDIA: "We are on our way…". It is hard to predict today when Intel will enter the market or if products coming from the Visual Computing Group will be competitive. During the last IDF Intel said :
 Roughly, it means that current GPUs will disappear and will be replaced in the future by massively multicore CPUs. At least, this is Intel's vision. Is the current project intended to develop such architecture and at very long term (5 – 10 years) much different API? It as hard to know the current plans of the manufacturer, but it seems for now that Intel is working on the shorter term… |
| |
|
|
Copyright © 1997- Hardware.fr SARL. All rights reserved.
Read our privacy guidelines.
|
|
