We've received a lot of response to our Intel roadmap article from yesterday. Some people still continue to say I am too critical with Intel, while many others called my projections for Intel as 'too optimistic'. This is again a good sign for me how controversial my articles seem to be seen and I am pleased with that. Unfortunately I have to disappoint those readers who consider myself as overly negative towards Intel, because new intelligence shows that my projections seem to have been too optimistic indeed.
Before I will spill out the newly found details, I'd like to publish an email I received yesterday. Please see it as some kind of small editorial. It is again critical with Intel, but Nils' thoughts are worth being considered well.
I read your article on Intel's latest roadmap and the P4. You seem to be very optimistic. I chuckled and occasionally laughed at some of Intel's ideas.
AMD is positioned with the EV6 bus as a processor-interconnect. This design is vastly superior to anything Intel has come up with. EV6 is in the same class as the proprietary designs used by specialty firms that actually make effective systems using arrays of Intel processors. And AMD is walking into the market with a proven, seasoned MP design that should easily outperform another that Intel can offer.
More importantly, AMD is pricing its MP technology at levels that almost anyone can afford, while Intel is aggressively opposing use of its lower cost processors in MP systems. Those Intel marketing people are crazy. So AMD is going to have zero inventory and pre-sell all the Athlons it can make while Intel relies of filling long pipelines and warehouses with parts to move the awesome output its fabs can produce.
Egghead still has surplus IBM Pentium computers at closeout prices - for $150 or less. This is how badly Intel's marketing strategies have affected inventories of older products. Four-year-old computers selling way below cost and failing to sell.
It really doesn't matter what Intel is talking about doing next year. The AMD Athlon/760 and Athlon/760MP combinations are going to win over the market and allow AMD to finance expansion of its fabs quickly. If AMD's capacity was 4 times as great, Intel might find itself as number 2 in sales next year.
Why is the Duron selling poorly? The market is already full of Celeron systems that don't sell well. Duron is far too close to Celeron in its target market image.
I am looking forward to buying a dual Athlon system based on 760MP glue - with an nVidia AGP video card that has good Linux drivers. By January or February of next year, the hardware, the new Linux core, KDE 2.0, and even better nVidia drivers will all be available. It is likely that GIMP and POVRAY will be improved also, giving me a great graphic design program and a fabulous ray-tracing program. Maybe a nice 20" monitor will be in the budget range also.
I would never count on Intel to match DDR SDRAM to the i830. Why not? Intel prefers to support only its high-end designs with premium memory subsystems. Consider the Xeon marketing policy as an example. Intel could have kept Slot 1 and improved its Pentium offerings by adding large L3 caches - but that is not a consumer market strategy. Intel wants to price its high end products in a way that makes them look like bargains compared to Sun servers while still making tons of money. So the i850 might use DDR SDRAM but the consumer market i830 will use much slower memory interface design. Offering DDR SDRAM with the i830 might kill sales of Intel's much higher priced 'server market' designs.
This is a sad situation. Intel and Microsoft desperately want streaming media supported adequately to appeal to the consumer market. Slower memory interface designs make it more difficult to deal with the enormous workload that now exists just for decompressing and moving streaming video to the display memory. When mpeg 4 arrives, most existing systems will not be powerful enough to deal with that technology, even for playback. And Intel will be hyping Pentium 4 systems with RDRAM as the solution while people who own AMD Athlon/760 systems with DDR SDRAM enjoy mpeg 4 playback.
Meanwhile, nobody wants to talk at all about Ellison's New Internet
Computer or about the 3 month old installation of those units at the DISD (Dallas Independent School District). Not one word. I believe that the new MSN Companion has more power than Ellison's design - but I can't prove it because the Ellison NIC processor has never been mentioned. Our public libraries are desperately seeking low cost solutions to satisfy their need for simple, low cost units that can be maintained remotely. When I am asked about solutions, I can only say 'wait until next year'.
Thanks for all your work. Do not forecast Intel's decisions based on your own knowledge of technologies. Intel is NOT making decisions on consumer needs or on providing good value to consumers. Intel is playing a game designed to only make maximum profit.
These are harsh words against Intel again, but instead of the people who call me 'anti-Intel biased' without supplying any reasons why Intel's actions should be seen in a better light, Nils Dahl was actually able to explain his point of view, which has indeed got something to it. You will understand that once you've read my new findings about Intel's plans with 'Almador', 'Brookdale', Northwood' and 'Willamette'. They are not looking as good as my write-up from yesterday. Does Intel care about its customer's wishes? You really wonder ...
Intel 830 'Almador' Chipset With Integrated Graphics And Without ICH3
Yesterday you've learned about this future platform for the upcoming Pentium III 'Tualatin'-core that is manufactured using a new copper-interconnect 0.13 micron process and supposed to be launched in Q2 or Q3 2001. I expressed my hopes that this next Pentium III chipset might support DDR-SDRAM, but now I've got new information that positions i830 or 'Almador' very differently, so that I am starting to have my doubts about the possible DDR-SDRAM support.
'Almador' will actually be a chipset that's replacing the current i815-chipset, which is unable to support the upcoming 'Tualatin'-Pentium III. Integrated into the 'Almador'-core will be a new 3D-graphics engine that is supposed to be 2-3 times 'faster' than the integrated 3D-deccelerator of i815. Even though 'Almador' will not support RDRAM as main memory, Intel seems to be using a trick to still fulfill the deal with Rambus . To speed up the 3D-graphics engine integrated into 'Almador', Intel will add an optional local graphics frame buffer module, called MRIMM. You can obviously figure that this MRIMM will be a special memory module using PC800 RDRAM. Thus 'Almador' won't be quite as 'RDRAM-free' as we hoped. The good news is that Almador will support an additional AGP-graphics card as well, but unlike i815 you'll be able to use both, the integrated as well as the added 3D-graphics engine simultaneously for some kind of dual-monitor support.
Although available by the time of 'Almador's' release, the upcoming new south bridge 'ICH3' will not initially be included with 'Almador'. ICH3 is coming with USB 2.0 support and some other nifty new features, and there is no technical reason why ICH3 wouldn't work with 'Almador'. However, Intel doesn't want to position 'Almador' as high as 'Brookdale', which will be released almost simultaneously as new Pentium 4 platform. 'Brookdale' will of course be equipped with 'ICH3'.
You can see that the outlook for 'Almador' has changed significantly. As chipset with integrated 3D-deccelerator and without ICH3 it looks much more like a low-end solution then it did initially. Therefore I am having my doubts that DDR-SDRAM will find its way into the design of 'Almador'. Obviously, Pentium III owners will have to count on VIA to provide a DDR-platform for their processor. Intel sees Pentium III as the 'second-class' product for next year, which is why it won't be teamed up with the same high-tech that Pentium 4 will get.
Pentium 4 'Willamette' Is Just A Short-Term Filler
We are only a month away from the launch of Pentium 4 right now. In the second half of November Intel will start shipping this new designed x86 32 bit processor, which will introduce the new Socket423 alongside with a new chipset, the i850 or 'Tehama'. The industry as well as customers are awaiting this new processor with mixed feelings, because the performance expectations aren't quite clear, systems with Pentium 4 will require a new platform and this platform will only support RDRAM, which doesn't exactly enjoy a particularly good reputation. I really don't want to say more about this, as we will be able to provide some real facts about Pentium 4 very soon.
In the Roadmap Article Part One from yesterday I presented 'Brookdale', a future chipset for Pentium 4, which won't be using RDRAM, but might support DDR-SDRAM instead. New information has now brought up a surprising new twist to 'Brookdale'. My sources say that 'Brookdale' will not be working with the Pentium 4 processors released in November. Instead, 'Brookdale' will be the platform for the next Pentium 4 code name 'Northwood' , that will be released in summer next year.
Basically, the upcoming Pentium 4 with the 'Willamette' core is only going to have a rather short life span and the same is even valid for Socket423, which is supposed to be phased out together with the 'Willamette'-Pentium 4 as well as i850 in mid 2001. The 'Northwood' Pentium 4 processor will be using a new Socket called 'mPGA478'. The 'Brookdale' chipset is not supposed to be even validated with 'Willamette', but only with the new 'Northwood' Pentium 4.
My information says that PGA423 (Socket423) will not be compatible with mPGA478 and so it might well be that whoever should go for a Pentium 4 Socket423 system with i850 chipset in the next months might end up in a dead-end road this time next year with no processor upgrade path left. Intel has told me now that there will be a transitional Pentium 4 version ('Willamette-N'?) that will run on both, the 'old' i850 chipset as well as the 'new' Brookdale chipset, with sounds similar to the 'Coppermine-T' solution in the Pentium III area. However, who wants to buy a Pentium 4 system in the next 6-8 months if there will be an advanced version of this processor in Summer 2001 with a completely different package and socket replacing the first Pentium 4 already? To remain fair I've got to remind you that AMD has played the same game with us as well. AMD came up with Thunderbird and SocketA in June 2000, only 10 months after the initial release of Athlon for SlotA. The situation was similar, as there was no reasonable processor upgrade path for owners of SlotA-platforms.