Chipsets for AMD processors
SiS
Despite that motherboards based on chipsets by NVIDIA and VIA are widely being sold, I haven't yet come across "live" products based on SiS chipsets. But SiS is a pioneer at promotion (at least, in the announcement) of chipsets with the PCI Express interface for Athlon 64 processors, and SiS756 chipset was officially announced as early as in March 2004! It was made following the classical dual-bridge scheme and by functionality it is very close to VIA K8T890.
 SiS756 chipset
The north bridge is a regular PCI Express bridge that has only the PCI-E õ16 graphic interface linked to the south bridge through own SiS MuTIOL protocol which provides a bandwidth up to 1 GB/s. The south bridge SiS965 is much more functional than VIA VT8237R: it offers two additional ports of the PCI-E õ1 bus as well as four SerialATA/150 ports, although the queue of NCQ commands is not supported. Full support for SerialATA/300 will be implemented in the new south bridge SiS966 which should appear closer to the end of the year. Besides, the LAN controller SiS965 supports Gigabit Ethernet.
It is easy to see that SiS756 offers only those features which are indeed in demand today and at minimum cost. The available four SATA/150 ports will work for a long time, especially in view of the fact that almost all modern DVD drives offer the parallel interface. The lack of support for SATA/300 also causes no special problems, at least today when hard disks with such interface have not yet gained essential occurrence.
ATI
ATI also tried competing against the leading manufacturers of chipsets for AMD64 and after a not quite impressive debut at creating chipsets for Intel processors released two chipsets - integrated Xpress 200 (codenamed RS480) equipped with the graphic core on the base of Radeon X300, and its discrete version Xpress 200Ð (code name RX480). Their major specifications are similar to those for VIA K8T890: Xpress 200 offers the classical architecture with two bridges, the north bridges are equipped with the 1 GHz Hyper Transport bus, one PCI Express õ16 and four PCI Express xl ports. The official south bridge of the chipset is the IXP400 chip, however, it is also possible to use M1567 chips by ULi (Taiwan) since they both are linked to the north bridge over the PCI Express bus.
 ATI Radeon Xpress 200 chipsetThe base modification Xpress 200 is equipped with the integrated graphic core on the base of Radeon X300 (with support for DirectX 9) running at 350 MHz, so it requires active cooling. The core offers two pixel pipelines with one texture unit in each, as well as four vertex units, which is not bad at all for integrated graphics. At the same time, it is also possible to use both shared system memory UMA (Unified Memory Architecture) and the dedicated frame buffer (up to 128 MB), as well as combine both these techniques (the HyperMemory technology).
Evidently, in the latter case the performance of video system will somehow go up versus the second and quite noticeably versus the first (at the same time, the end price of the motherboard will also go up), however, it is doubtful that even if the HyperMemory memory mode is used the performance of integrated video will reach the level of discrete video processor ATI Radeon X300, let alone more "serious" video cards. So using Xpress 200 for games is out of the question. The south bridge IXP400 is almost a complete replica of VIA VT8237 bridge, with the number of supported SerialATA/150 ports increased to four (versus two in VT8237). Like VIA, ATI assumes that with the emergence of high-speed PCI Express bus there is no longer a need for the integration of Gigabit LAN MAC-controller into the chip since one full-featured external chip will cost cheaper and will run not at all slower than integrated.
ULi
Among the multitude of varied solutions available today on the market of PCI Express chipsets for AMD processors, one of the most original is the M1695 chipset which has recently been released by ULi (former ALi, Taiwan). ULi's approach to building the chipset is quite successful: its both bridges are linked with a 8-bit bi-directional HyperTransport bus, which gives more freedom in choosing south bridges. Since the architecture of HyperTransport supports serial connection of several devices, nothing prevents using it, together with the south bridge M1567 (or instead of it), e.g. the AMD-8132 chip with support for PCI-X (can be used in High-End workstations and servers), or another bridge with the graphic interface PCI-E x16 (do they mean to say they will beat NVIDIA nForce 4 SLI 16x?).
 ULi M1695 chipsetWhen connecting the north bridge M1695 to the nominal M1567 on the motherboard, simultaneously three graphic interfaces can be implemented - PCI Express x16, AGP 8õ, and PCI (the latter is purely theoretical, because today few recall the existence of PCI video cards). At the same time, in the north bridge there is only the graphic interface PCI-E x16 (plus 4 additional PCI-E x1 channels) which can be configured as two PCI-E x8 channels, i.e. the chipset supports the SLI mode for two graphic cards.
The south bridge M1567 contains customary interfaces - two UltraATA/133, two SATA/150, eight USB 2.0 ports, audio AC97 2.3, 6 slots for the PCI bus, and 100-Mbit/s LAN, as well as the full-featured AGP 8X interface.
In view of the high flexibility of M1695 architecture that allows building well scalable systems of any level (from entry up to High-End), the product by ULi has it all to become a good alternative for users of Athlon 64, and, at least theoretically, press NVIDIA on this market. But to do that, the Taiwanese company will have to do a mere "trifle" - having given up the former image of manufacturer of solely "paper" chipsets, bring its product to a stage when it can be launched to a mass production.
Conclusion
Finishing the review in which we examined all the currently available PCI Express chipsets for AMD64 and their nearest prospects, let's try looking a bit further: what joy will the next year bring to us? It will "please" us with another total replacement of the AMD platform. And that is related to the planned transition to the new 940-pin Socket M2 which will start in the second quarter of 2006. Despite that the number of pins in M2 and Socket 940 coincides, they can't be compatible to one another. Compatibility to Socket 939 is therefore out of the question. The cause of such sharp move of AMD, according to the statements of the company representatives, is the transition to new DDR2-667 memory. Although, frankly, I am having great doubts in the validity of such statements - all that very much resembles another attempt of AMD to improve its financial standing at the expense of overly credulous customers.
Join the conference with all your questions, remarks and suggestions.
 |
Top Stories: |
 |
 |
|
MoBo:
| |
|
|
VGA Card:
|
|
|
|
CPU & Memory:
|
|
