Chipset Update
VIA Chipsets
The big news in chipsets this month is the KX133, however this time it is not very good news. It was revealed during the past week that the KX133 chipset is not compatible with the Thunderbird processors. There has been a fair amount of speculation about the reasons why, and even some negative comments towards both VIA and AMD, since those who have just purchased motherboards based upon this chipset may already be at the end of the line for processor upgrades.
A VIA spokesperson would not confirm nor deny that this was a problem, however he did emphasize that the KX133 was specifically designed for the K-75, and that at the time it was developed the 1GHz+ processors were still thought to be over a year away. An AMD spokesperson would not provide any specifics about the reason for the incompatibility, but said it was ‘not surprising’ that a chipset designed a year ago for the K-75 would have some compatibility issues with the Thunderbird.
After doing a bit of digging, it appears that the problem is actually the same issue that may prevent the development of cost effective Slot A to Socket A converters (Slockets). It seems that some of the necessary timings for higher frequency Athlons (the address bus hold time has been specifically mentioned) were not known nor anticipated by VIA in their development of this chipset, so the KX133 timings are just outside of what the Thunderbird requires. Trace length is always an important consideration in regards to timings, so it is entirely possible that some good motherboard designs will actually work with Slot A Thunderbirds. For most people, however, this isn’t really much of a consolation, since AMD does not plan on putting any Slot A Athlons in the channel, as they are intended only for OEMs.
The focus is now going to be on the KZ133 chipset, which, as reported last month, has been ‘tuned’ for the Socket A processors. The VIA spokesperson indicated that while both the KX133 and KZ133 are electrically compatible, the pin locations are slightly different so that trace lengths can be kept as short as possible for each type of processor connector. Though some have criticized AMD (and Intel) for going back to a socketed connector for the CPU, both companies have said that higher frequencies require shorter traces, which a socket allows for. Of course, with on-die caches now standard, there is really no reason to mount the processor on a card anymore, so cost is a big consideration too.
It is well known that KZ133 will support the 133MHz bus speed, as well as all of the other features of the KX133, however it also appears that it will include ATA/100 support. It also seems likely that some of the critical timings will be addressed before the official release (no pun intended), if they have not already done so. One upside to this is that a Socket A motherboard will support both the Duron and the Thunderbird, allowing for a relatively inexpensive upgrade today, and a potentially excellent upgrade path in the future.
The KX133 issue cannot be good news for VIA. After delaying the introduction of this chipset for almost six months, and ramping very slowly, it would seem that the KX133 is now destined to go the way of the Dodo (that would be the i820, for that need the punch line explained). Fortunately for VIA, the Apollo Pro133 chipset is still their number one seller by a very large margin, but one has to wonder how many KX133 chipsets in production or in inventory will be wasted, and how many motherboard manufacturers will be pissed off because they will have relatively worthless KX133 boards in inventory.
Intel Chipsets
There is not much news from Intel regarding their chipsets, mostly because their chipset roadmap seems to be in complete disarray. As reported last month, the i820 chipset is all but dead and the i840 is too expensive for the mass market. The i810 chipset is still selling, but sales are ‘steady’ rather than rising. For this reason, the i440BX chipset is still the most popular Intel chipset, but is rapidly losing ground to the VIA Apollo Pro because it lacks the advanced features many users have come to expect today (4x AGP, UDMA/66, PC133 memory support, etc.).
The last roadmap shows the i815 and i815E chipsets being officially launched in June, however manufacturers have said that they are only supposed to provide motherboards for OEMs, with channel availability in September. Industry insiders say that most OEMs and motherboard manufacturers are not happy with the current situation with Intel chipsets, and are anxious to see the issues resolved.
Recent news reports seem to indicate that Intel’s patience is getting very short with regard to DRDRAM pricing. Unfortunately, I’ve heard that Intel signed a long-term agreement with Rambus, so getting out of it may involve a hefty fee that Intel might still be deciding whether it is worth it or not. Obviously, the pain is not so great yet to convince Intel to just amputate the limb, but one must wonder if it will eventually come to that.
Other Chipsets
In an interesting development, the AMD 750 chipset, which was originally criticized for not having the advanced features necessary for ‘popular acceptance’ such as AGP 4x, is now the premier Slot A chipset because existing motherboards based upon it are compatible with the upcoming Thunderbird. Whether this gains AMD some respect as a chipset company or not remains to be seen, but one has to be amused at the thought of millions of SD-11 owners being able to heckle the ‘foolish’ people who waited to purchase a ‘more advanced’ KX133 based motherboard. Am I displaying a sick sense of humor here?
At WinHEC (week of April 24) AMD demonstrated a system based upon their upcoming 760 chipset, which supports DDR SDRAM. A few manufacturers have indicated that they have layouts already, and are in the process of testing and debugging. This means that every chipset manufacturer except Intel will have a DDR enabled chipset by early Q4. The anticipated date for actual product shipments is late Q3 or early Q4, with volume most likely in mid-Q4. Even Micron will reportedly enter the chipset arena by manufacturing their DDR enabled Samurai chipset, along with a variation that supports the Athlon.
Be the first to discuss this article!