Intel Roadmap News 10/2000 – Part One, Desktop Processors And Chipsets – THG.RU

Introduction

Getting hold and posting Intel’s secret roadmaps has a long tradition at Tom’s Hardware, but I admit that it has been rather quiet in this area for several months now. Obviously Intel is not feeling too well right now and so they made even more efforts to keep those future projections as secret as they should be in Intel’s opinion.

Finally we were able to get an actual version of this highly anticipated paper and were very surprised to see quite a few unexpected changes. After Kyle Bennett from Hard OCP posted a few roadmap slides last night I will follow up on this with the explanation and evaluation you are used to at Tom’s Hardware Guide. Please don’t forget to have a look at Kyle’s peek at a future Pentium 4 motherboard with i850 chipset as well to round things up.

Intel Desktop Processors

I am sure that you are all aware that Intel is currently in a bit of a nasty situation. After the retraction of the Pentium III 1.13 GHz at the end of August, in which Tom’s Hardware played a major role, Intel’s fastest processor is currently still the Pentium III at 1 GHz. At the same time AMD’s Athlon 1.1 GHz is available for more than a month already and the Athlon 1.2 GHz will be released very shortly. There’s no denying that Intel is currently quite a bit behind AMD, as Athlon processors are clearly the performance leaders in the x86 market right now.

Pentium 4

Obviously Intel’s hopes to regain the performance crown are all focused on the upcoming Pentium 4 processor. This completely new design, also known under the code name ‘Willamette’, will enable much higher clock speeds, starting at 1.4 and 1.5 GHz by the time of its release. Take those high frequencies with a grain of salt though, since it seems rather obvious right now that Pentium 4 will be slower than its predecessor Pentium III in a clock-for-clock comparison using common software. This is not necessarily a problem, since Pentium 4 can obviously afford to be a bit slower clock-for-clock, because it won’t be released in clock speeds of less than 1.4 GHz, which is already 40% higher than the fastest Pentium III processor that’s available right now.

A look at the roadmap makes it even more obvious that Pentium 4 will most definitely stay way ahead of Pentium III, since Intel is planning Pentium 4 processors at 1.7 GHz or more for Q1/2001 already and Q2/2001 is supposed to be the time when the first 2 GHz Pentium 4 will see the light of the world.

Intel’s aggressive Pentium 4 forecast seems to be a strong sign that the production of Pentium 4 is running very well. Intel wants to force its way back to being the performance as well as the market leader in the x86-market and AMD will have quite a tough time to compete with 2 GHz Pentium 4 processors in terms of performance.

The story might be very different looking at future Pentium 4 sales vs. future Athlon sales. While AMD, Ali, VIA and even Micron will soon supply Athlon-platforms that support the highly-anticipated DDR-SDRAM memory, Intel has based its Pentium 4 chipset i850 entirely on Rambus’ highly-unpopular RDRAM memory. Market analyzes show clearly that a huge number of OEMs and the vast majority of consumers are avoiding the overprized and overhyped RDRAM memory like the plague. Unless Intel should change its chipset plans very quickly there won’t be too much acceptance of Pentium 4 in the market. I will discuss the chipset issue in detail a bit further down in this article.

Pentium III

Will Pentium III soon be dead? How could it? It’s the most successful microprocessor and as the market won’t welcome Pentium 4 and its RDRAM-platform Intel will continue to make most of its sales with this very processor. The only problem that Intel is facing right now is the fact that it is impossible to get the current Coppermine-core to any higher speed than 1 GHz reliably. Although it may sound crazy, we all will have to face the fact that 1 GHz will soon be a rather pathetic processor speed and thus Intel is working hard on improving the current Coppermine-Pentium III to get it across the 1 GHz barrier.

Coppermine-T and Tualatin

The most logical step to improve Coppermine is obviously to shrink its die down to a 0.13-micron process. This is exactly what is going to happen, but it is by far not all. The next proper version of Pentium III will use a core with the code name ‘Tualatin‘ and it comes with the following specs:

• 1.13 / 1.26 GHz at launch date
• 512kB on-die L2 Cache
• Differential Clocking = a new clock spec
• VRM 8.5 = a new voltage spec
• AGTL (1.2VTT) replaces AGTL+ (1.5VTT) = a new bus spec
• AGP 4X 1.5 V
• CPU will look more like current SocketA processors with the four pads, but with an additional ring around the flip chip to make sure that it doesn’t get destroyed by the heat sink, called FC-PGA2 = FC-PGA with integrated Heat Spreader. FCPGA2-processor will be 3.5 mm high instead of FCPGA’s height of only 1.9 mm.

On the first look the doubled L2-cache size of Tualatin looks most interesting to the majority of you, but you can see that the modified clock supply, voltage and bus spec will require a new platform for this future Pentium III processor. Currently there’s only one chipset in Intel’s roadmap that is supposed to cater for Tualatin, the ‘Almador‘ or i830 chipset, which I will discuss in the chipset section.

Intel is fully aware of the fact that owners of current Pentium III platforms might want to upgrade to the new ‘Tualatin’-core as well, so it will throw in another Pentium III version with the so-called ‘Coppermine-T’ core, which will also be manufactured in 0.13-process. This core will probably only come with 256 kB L2-cache (it might be 512 kB though; we are not sure) and it will be able to run on ‘Almador’-platforms as well as on current Pentium III motherboards with Intel’s i810, i815, i820 and i840 or VIAs 694x chipsets.

Both processors are currently targeted for Q3/2001, which seems extremely late. Intel seems to think that the high-end community will indeed go for Pentium 4, as 1 GHz will remain the limit of Pentium III until Tualatin’s release. I personally expect that we will see Tualatin and Coppermine-T a lot earlier next year.

Celeron

We all know that Celeron has lost almost all of its attractiveness since the release of AMD’s Duron-processor. Celeron is still specified to run at mediocre 66 MHz FSB, which is the main reason why Intel’s low-cost solution lags so far behind AMD’s in terms of clock-for-clock performance. Finally Intel is under enough pressure to enable 100 MHz FSB for Celeron, but according to the roadmap we’ll have to wait until Q2/2001, when Intel will finally bless us with the Celeron 850. All overclockers know that this CPU has become reality a long time ago, because it’s what you get when you run a Celeron 566 at 100 MHz FSB.

I am expecting to see Intel introducing the Celeron 850/100 a lot earlier than Q2/2001. It only depends on how boldly AMD’s Duron is able to destroy Intel’s current Celeron sales.

Intel Desktop Chipsets

The i830 ‘Almador’ Pentium III Chipset

A lot of speculation has been going on around this chipset since TheRegister first mentioned it several months ago. Even in the current roadmap Intel doesn’t want to admit the obvious though, leaving a lot of room for even more theories.

Basically, Almador is the badly needed chipset to support the new advanced Pentium III processor with the ‘Tualatin’ core. It supplies the new clock, voltage and bus specs and will obviously be released together with ‘Tualatin’ within the next 9 months. The most important issue with ‘Almador’ however is the memory support. What we do know for certain is the fact that this future Pentium III platform will NOT support the dreaded RDRAM memory. The roadmap states that ‘Almador’ will run with PC133 SDRAM memory, but everybody at the right state of mind will realize that PC133 will be more than old news in Q3/2001 when AMD-platforms will benefit from the much faster DDR-SDRAM for almost a year already. Therefore it seems a pretty safe guess that ‘Almador’ will support DDR-SDRAM as well, which might not be admitted in Intel’s current roadmaps because Intel has still got this unholy deal with Rambus, the company that pretty much hates DDR-SDRAM memory.

All in all, it seems rather likely that Intel will release the i830 or ‘Almador’ chipset and the ‘Tualatin’-Pentium III quite ahead of Q3/2001, bundled with DDR-SDRAM support. This could be a very interesting product and it might save Intel’s neck, which otherwise takes the risk of only supplying uncompetitive old Pentium III systems and a Pentium 4 that only runs with the unpopular RDRAM memory.

Here are the specs of the i830 ‘Almador’ chipset:

You can see that i830 is quite advanced in comparison with i815. It does not come with an integrated graphics core, the memory support is not limited to the pathetic 512 MB of i815 and it supports ECC. This is a clear sign that Intel regards i830 as a full blown chipset, not a low-cost solution as with i815.

The i830 chipset will further more be accompanied by the upcoming ICH3 south bridge, which will support the new USB 2.0 spec and offer 6 USB-ports amongst other things.

The Intel 850 Pentium 4 Chipset

I won’t say much about this chipset, as it will be released together with Intel’s Pentium 4 processor in a bit more than a months time, the second half of November 2000. We all know that Intel will be trying to force RDRAM down its customer’s throat once more, as there won’t be an alternative Pentium 4 platform available for a while. I personally guess that if Intel won’t be successful with the upcoming Pentium 4 / i850 bundle because of Rambus, it will finally drop this widely disliked company, which will most likely be the (well-deserved?) end of Rambus Inc.

Here are the (mostly well-known) specs of i850:

The Intel ‘Brookdale’ Pentium 4 Chipset

While i850 doesn’t seem to be very exciting at all, the ‘Brookdale’ chipset seems to me as one of the very highlights of this roadmap. It’s another Socket 423 / Pentium 4 chipset, but it doesn’t have much in common with i850 at all. In fact, this chipset may be the one thing to make Intel’s Pentium 4 a successful product at last.

In the roadmap that I’ve got in front of me it looks as if Intel wants to release this chipset in September 2001. This seems very late and I will tell you why. ‘Brookdale’ does NOT support RDRAM!!! Again we have a situation similar to the ‘Almador’ chipset. Intel’s roadmap is speaking of ‘PC133’-support, which comes across extremely pathetic, since PC133 SDRAM will have been replaced almost entirely by DDR-SDRAM by September next year and thus PC133 is not a sufficient memory solution for a high-end Pentium 4 system. ‘Brookdale’ is supposed to be a high-end platform however, as it supports up to 3 GB of RAM and thus more memory than i850. It also comes with ECC-support and ICH3. Thus ‘Brookdale’ gives the impression as if it is an advancement over i850 and so I think it’s 100% justified to expect Intel equipping ‘Brookdale’ with DDR-SDRAM support. Bye-bye Rambus!

These are ‘Brookdale’s’ preliminary specs:

My 2 cents about ‘Brookdale’ are that I expect Intel to move up this product and include DDR-SDRAM support. A release of a DDR-enabled ‘Brookdale’ by Q1/Q2 2001 would ensure the success of Pentium 4 and could threaten AMD quite nastily. If Intel fails to see this chance then they really don’t deserve to be the top microprocessor maker anymore.

Summary

Intel’s new roadmap has given me hope that the Santa Clara based company will finally get back to where it used to be. The high clock speeds of Pentium 4 that we can expect in the first half of next year should be accompanied by a chipset that has the backing of the industry and the consumers. The i850 RDRAM-chipset will not achieve that and Intel’s managers can’t be so blind that they wouldn’t see how Rambus Inc. is endangering Intel’s already damaged image. Let’s hope that ‘Brookdale’ plus DDR-SDRAM memory will be available for Pentium 4 as early as possible next year.

After Intel’s unfortunate retraction of the Pentium III 1.13 GHz it had become obvious that ‘Coppermine’ isn’t good for clock speeds in excess of 1 GHz. However, Pentium III is still having a bright future ahead, particularly if Intel should not offer ‘Brookdale’ as alternative Pentium 4 platform. Therefore I hope that Intel will release ‘Tualatin’ and i830 ‘Almador’ soon next year as well.

If you look at it closely, the decision between success and failure of Intel’s future in the x86-market lies in the question if Intel will continue forcing customers into using RDRAM as well as continue to ignore DDR-SDRAM. There is no doubt that Pentium 4 will have a good chance of success if Intel can supply the high clock speeds and if it will offer a DDR-SDRAM platform to go with it. The same is valid for ‘Tualatin’. Let’s be honest, the first stumble stone that lead to Intel’s long list of recent failures was the Rambus-issue with ‘Camino’ in fall of last year. If I’d be superstitious I’d say that Rambus has brought Intel nothing but bad luck. I suggest it’s more than time to cancel this unlucky union.

Stay tuned for the next part of the roadmap, which will present you Intel’s highly interesting plans in the mobile area and its strange policy in the workstation and server segment.

Please follow-up by reading HOT! Update Of Intel Roadmap News!.