Intel confirmed Friday the cancellation of the Tejas and Jayhawk processors, but did not unveil any code-names or other details concerning its microprocessors due to launch in 2005.
Intel’s Roadmap Vague
The information about future projects is typically held under-wraps and is not widely discussed publicly. However, some portions of details about the upcoming microprocessors are frequently unveiled at variety of events like Intel Developer Forum to prepare the industry and the general public for the future of technology. Additionally, quite a lot of information is leaked by sources close to the companies, which is sometimes done intentionally by the firms themselves so that to learn the reaction of the market. While the code-names for the Pentium 4 derivatives, such as Prescott, Northwood, Tejas, Cedar Mill, etc, as well as some of their technical specifications have been known years before the products were meant to arrive, today Intel does not unveil any details about the microprocessors that will come to replace the existing products in desktop and 2P server platforms.
The absence of the information about future products may be a sign of an ongoing dispute within Intel about the next-generation technology. The latter may mean that the roadmap of the company is not totally formed up in the way it may be presented to or discussed with partners. Intel does not talk about micro-architecture and whether the dual-core chips scheduled for 2005 introduction have been taped out or not.
Intel tends to bring new technologies into the market in the first half or the year. Generally, the company needs to finalize product specifications and possibly tape the chip out about 10 to 12 months prior to the commercial product launch, several sources said. In case there is no information about the next-year’s chips right now, this may represent some shifts of introduction towards the second half of 2005. According to the current plans, the current Pentium 4 “Prescott” will cease to evolve in the Q1 2005. The Tejas and Jayhawk were designated to arrive in the Q2 and the Q3 2005 respectively.
There is a clear confirmation that the dual-core processor for MP servers internally called Tulsa is shelved, but Intel stressed Friday that the next-gen chip for multiprocessor servers code-named Potomac is still in plans, despite of earlier assumptions.
The Santa Clara, California-based chipmaker will hardly produce the new chips using 65nm process technology, as it is scheduled to begin the ramp only in the very late 2005 or even early 2006. Therefore, we may be almost sure that the timetables for the new fabrication process and dual-core chips’ introduction are not aligned.
Investors Happy with the Move?
Despite of indefinite roadmaps, Intel’s shares rose $0.49 to $26.47 on Nasdaq after the news about the cancellation of Pentium 4 successor code-named Tejas and the next incarnation of server chip Xeon code-named Jayhawk.
This may indicate that investors felt a relief for the company’s future, as Intel has had various problems with delivering of both of its NetBurst-based 90nm chips Prescott (Pentium 4) and Nocona (Xeon). On the other hand, the rise may be connected with Monday’s introduction of the Dothan processor – an improved version of Intel Pentium M chip that is the base of the company’s extremely successful Centrino platform.
NetBurst or P6?
Intel’s spokesperson said that “2005 will be a great year for Intel” because dual-core chips “are great for customers”. Intel is emphasizing that the idea to come up with dual-core chips in 2005 goes from the discovery that the company is able to deliver such products earlier than it thought originally. What the company also stresses is that the new dual-core chips will have the same platform requirements as cancelled Tejas – this is important for mainboard makers and system integrators, who will not have to change their existing designs and spend additional R&D money.
The real questions about Intel’s vision of the future come when we remember about the technologies the Tejas was expected to bring according to various sources: Vanderpool – a technology that splits personal computer into several virtual parts that work independently and use the same resources of the PC, LaGrande – a universal security feature that protects confidential information on the PC. While there might be more capabilities to enter with the Tejas, the Vanderpool, LaGrande and 64-bit capability that is available on Prescott were the most important to come to life with the Tejas, some unofficial sources noted.
While quite some sources believe that Intel will go Pentium M route after the abandon of Tejas, it is not clear whether Pentium M-based future products really had all the mentioned capabilities built-in from the ground-up or Intel will have to redesign the cores. Furthermore, there were NetBurst-based dual-core microprocessors in plans that really should have Vanderpool, LaGrande and EM64T enabled. Today it is not clear whether Intel accelerated dual-core NetBurst or dual-core P6 plans, and that is the most significant intrigue here because of the lack for official information.
What We Do Know Just Now
Before Intel finally officially unveils its visions of the future, let us summarize once again what we do know from unofficial sources about the next-generation desktop microprocessors from Intel:
- Compatible with the Tejas on the platform level;
- Uses Quad Pumped Bus;
- Possibly comes-in LGA775 packaging;
- Platforms (e.g. Lakeport) to feature dual-channel DDR2 667MHz memory and sport PCI Express interconnections;
- Platforms to support FB-DIMM.
On the platform-side everything is pretty clear – in case microprocessors use similar bus, it makes no problems to switch from architecture to architecture provided that there are no special design tweaks in the platform core-logic to bolster performance on processors with one particular micro-architecture. But what about the insides of Intel’s future central processing units that will be the base for next-year’s 2P servers and 1P desktops?
Comments currently: 7
Discussion started: 05/10/04 02:29:47 AM
Latest comment: 05/11/04 07:34:54 AM
There's always room for Intel to eventually bounce back with a competitive product. But as everyone's pointing out (http://www.theinquirer.net/?article=15795) this won't happen anytime soon. Intel is in a world of hurt now with Prescott on tap for at least 3 more speed increases (3.6, 3.8, 4.0 - can you say hot, hot, and even more hot!) after the 3.4 GHz. IMO Intel should be completely avoided for the rest of 2004 unless something miraculous changes with their problems. There'll never be 64bits enabled in any Prescott lest it fry itself and leave the nice black pattern on the motherboard.
Anenome, don't forget that upgrading may be harder than you think with reports that removing the processor even once from an LGA775 could break it due to its brittle design... which Intel is blaming motherboard makers for. I'm sure they're feeling pretty jerked around by now for first being forced to adopt and invest in all of Intel's new standards, and then learning about the uncertain future of LGA775 without Tejas.
Cancelling Tejas shows there will be little chance for resolving the thermal problems in Prescott since Tejas would have added dual core to the same Prescott architecture. Prior to the 90nm son-of-a-PentiumM coming out, the materials will not be Strained Silicon or SOI and so the problems will remain until I'd guess late 2005 or possibly 2H 2006 (delays along with the inevitable new technology bugs.)
My whole point is that Intel won't be ready by December anyways, so just wait for the FX-55 (or 4000+.) Even if AMD does report any problems with 90nm I can still buy a new PC with PCI-Express, faster Hyper Transport, and the 939pin socket version of the FX-53 as a much better band-aid. Then once any 90nm problems get resolved I can upgrade to an FX-57 or whatever that will eventually be dual core. I'm having a hard time finding any problems with AMD right now. I've always been a big fan of Intel, but lately it just looks like poor decisions and arrogance, followed by the neccessity to copy AMD over and over again. DDR2 is not important anytime soon, especially with performance increasing beyond DDR500 thanks to OCZ and others. I'm anxiously awaiting the benchmarks and over-clockability between the FX-55 and whatever else is available at the time. I just hope it comes out as soon as September, and that shortly after a PCI-Express ATI All-In-Wonder will be available too. I've been following this stuff every day since December and can only wait so long. My P3 1Ghz is in dire need of replacement since it's at the bottom of most 3d shooter support lists.
05/10/04 10:48:42 PM]
Have many Intel machines - have always worried about the "little misc things" that go wrong. Even now years later people write of the issues that the Nforce2 series still has, not totally disastrous issues save for the peripherals going away, but annoying things nonetheless. Intel, in side by side 24/7 boxes still gets the nod for rock solid stability after months and months of running. There just seems to be more tolerance, more testing done to make sure the Intel stuff runs, so its a bit more than the cpu I'm weighing in here.
But, how can one not look at the FX and not like it? It's the Porsche of the cpu world. Fast, 64 bit, and its not like this next machine won't have games on it 75% of the time. But with dual cores I'm thinking more heat problems. FX doesn't generate tons of heat but it does generate a fair amount, and dual core that chip and things could get toasty. AMD has not had some easier time with 90nm it seems. Then there is the cache design issues of the two chips and the shared data problems. Dual cores are simple in theory, but more complex to get the threads to work well. Given how fast the FX is, well that might not be any concern really, but we'll see on that. I think in the end, the 90nm AMD chips are going to be late, maybe very late, dual core will be very expensive and right now they've only promised it for Opteron, not the regular desktop. I guess the last issue is cost. I'm not fond of single channel options, and I've had them before, so the regular Athlon64 series isn't something I consider an option in my range of possibles. With Intel I can choose a moderate cost cpu, and then just replace it with something faster, but similar priced later, when a faster chip has fallen from "new" status. For example a 3.2 @ $250 or so, then to use the example a 4.0 at $400 when its fallen from its 617 intro. The pricing for FX is always the same, so it hurts now and it hurts later (but it IS FAST - so well that's kind of ok), but it plays into upgrading. The socket 775 thing though - uggh, those are going to have serious issues.
So FX I'll worry about stability, and I mean the rock solid absolutely perfect stability @ a moderate high clock, cost and the upgrade path.
Intel I worry about the upgrade path, the cost of that damn DDR2, the lifespan of the socket 775, even the lifespan of the DDR2 if they are just going to move to buffered. Its sad but given the idiotic cost of DDR2 the pricing of these two options is closer than makes any sense.
So in a pure performance for dollar the FX wins hands down and then some. If stability, or that nagging worry that components added later might be incompatible or there might be hidden issues later with Nforce3, then the Intel solution looks "ok", hardly good. Intel's silence on what we'll be able to do later with any platform bought now makes "ok" down to "kinda ok".
In the end I guess the FX is becoming the clearer and clearer option, but I foresee the 90nm version, with SSE3 and the improved 500mhz memory controller not coming maybe till near the end of the year, maybe 2005.
If Intel picked up the ball, things could change. If the FX-55 came, things could change. Or maybe I'll just listen to the constant advice of nearly everyone and wait for the FX-55. Thanks for constantly battering me with the advice, I'll listen sooner or later :)
05/11/04 03:37:00 AM]
Add your Comment
Enter your username and e-mail address. Password will be sent to you.