Intel is bringing one of their worst years in terms of execution and product launches to an end.  While the year started off pretty strong with the Pentium III and its Coppermine core boasting a scalability advantage over the Athlon without any on-die L2 cache, things quickly took a turn for the worse for Intel. 

Processor shortages throughout the first half of the year led to increased prices and gave AMD the opportunity to gain some of the high ground with their Athlon.  Combined with solid motherboard support for the processor, the Athlon began to eat away at some of the pie that was previously reserved for the Pentium III. 

We really started to see signs that we weren’t dealing with the same Intel we had been used to when AMD managed to not only beat them to releasing a 1GHz processor but also to making those 1GHz processors widely available.  All of the sudden Intel was struggling to compete on a clock for clock basis, not necessarily in performance but in terms of actually achieving the clock speeds that AMD was pushing for.  We should all know by now that clock speed sells when it comes to the retail market.  Since most uninformed buyers don’t have a list of benchmarks to make a buying decision with they concentrate mainly on CPU clock speed to define whether something is “faster” or “slower” than another option. 

The latest shock came in the complete recall of the 1.13GHz Pentium III processors which was almost single handedly inspired by Dr. Thomas Pabst of Tom’s Hardware.  Intel was apparently trying to pull a fast one on the market by sacrificing stability for clock speed, a practice that had previously been completely unheard of from Intel.  Remember the days of complaining about AMD’s yields and praising Intel’s quality control?  How times have changed…

Now we’ve all given Intel a hard time because of their recent failures, but the real question to ask is how do they plan to recover?  AMD has been taking advantage of the position they have been put in, and Intel is finally waking up and realizing that they’re not competing with a “low-cost” competitor anymore.  Gone are the days when a Celeron processor can compete with AMD’s entire processor line. 

Let’s take a look at how Intel is positioning themselves to make a comeback and then we’ll analyze what their chances are at taking back some of the ground they’ve lost. 

None of the information contained in this article is provided by Intel and the following roadmap may not hold true.  Let’s just call it a set of “informed” guesses at what we think Intel will be doing in the next year.

The Celeron gets back on the bus

The announcement was made not too long ago that Intel would be scrapping their Timna processor due to a lack of interest from OEMs.  The Timna CPU, as we first mentioned in our coverage from the Spring 2000 Intel Developer Forum, was supposed to be the first CPU from Intel that used a technology they called “Smart Integration.”  Smart Integration was basically a simple term which indicated that the Timna CPU would feature an integrated memory and graphics controller, thus making it perfect for sub-$600 PCs and possibly even notebooks. 

Unfortunately the memory controller that was to be integrated into the Timna CPU was the same RDRAM controller that had caused the failure of the i820 and i840 chipsets.  Intel then planned on using their MTH to provide SDRAM support for Timna platforms, but with the MTH bug and the recall of all MTH boards those plans fell apart immediately. 

Throughout next year the Value Processor from Intel will continue to be the Celeron based on the Coppermine128 core.  By “Value Processor” we mean that it will be found in pretty much all retail sub-$1000 systems. 

We’ve continued to mention that the Duron, even at its current clock speeds (700 – 800MHz) will have no problem competing with even a 1GHz Celeron, so how does Intel hope to make the Celeron competitive in 2001? 

When the Coppermine128 core was first released we asked Intel why they weren’t moving the Celeron line to a faster (100MHz) FSB which would definitely help increase its performance.  Their reasoning was basically that the Celeron would move to the 100MHz when it was absolutely necessary. 

The release of the Duron processor and its subsequent trampling of the Celeron in terms of performance has met Intel’s criteria for making the 100MHz FSB absolutely necessary.  The result is that in the first quarter of 2001 we will see the introduction of the last 66MHz FSB Celeron CPU and the first 100MHz FSB Celeron processor. 

Having just recently released the Celeron 733, expect to see a 766MHz part (66MHz FSB) and an 800MHz (100MHz FSB) Celeron in the first quarter of next year.

The Celeron will hit 850MHz before the end of the first half of 2001 and we should see speeds of 900MHz and higher before the years end.  At the end of 2001 the move to all 100MHz FSB Celerons should be complete. 

The mainstream Pentium III

With the upcoming release of the Pentium 4, the Pentium III will drop down a notch and will now be Intel’s mainstream desktop CPU. 

At the start of this article we mentioned the 1.13GHz Pentium III recall, well, according to our sources, the 1.13GHz Pentium III won’t hit the streets until sometime in the second quarter of 2001.  This means that the fastest Pentium III will continue to be the 1GHz processor for quite some time now. 

The recalled Pentium III 1.13GHz won't be reappearing anytime soon

The reasons behind this are two fold.  For starters, Intel would definitely not want another repeat of the 1.13GHz recall.  Even though the recall was not on a large scale (less than 200 CPUs were actually out in customers’ hands), it truly impacted our readership as well as other enthusiasts since it was the first time in quite a while that an Intel desktop CPU was pulled back. 

The second and quite possibly more realistic reason is that Intel doesn’t want the Pentium III competing with the Pentium 4 any more than it already is doing.  As you’ll soon find out, a 1GHz Pentium III on an i815 platform can give even the 1.5GHz Pentium 4 a serious run for its money.  A 1.13GHz Pentium III would only shorten the performance gap that exists between the Pentium III and Pentium 4 and even tilt the standings greatly in favor of the Pentium III. 

Interestingly enough, the slowest Pentium III on Intel’s 2001 roadmap is the 800MHz part which Intel is targeting at the $1,000 - $1,200 system price range.  Keep in mind that this is just slightly above the $850 - $999 price range which is supposed to be occupied by the Celeron 766/733 in Q1-2001.  By the second quarter in 2001 the slowest Pentium III on the roadmap is the 933MHz part, and by the third quarter of next year everything but the 1.13GHz and the 1GHz Pentium IIIs disappear from the roadmap. 

What’s happening to the rest of the Pentium III line?  Intel is unfortunately in a bit of a predicament here.  The Pentium III, especially the 650 – 750MHz range that isn’t present on their 2001 roadmap, would end up outperforming even the 850MHz Celerons that are due in the middle of next year.  But, clock speed sells, so Intel is going to have to ditch the lower clock speed Pentium IIIs as to keep them from competing with the Celerons. 

By the end of January you can expect all of the Pentium IIIs below 800MHz to be priced between $100 and $200, with very little ($10/$20) or no variation between the prices of different speed grades.  If you’ll notice, this is pretty much the price range that all of those CPUs fall into currently, illustrating that between now and next year you shouldn’t expect to see too far of a drop in price for these sub 800MHz Pentium IIIs.  If you’re looking to buy one of these CPUs, don’t expect to see a huge price drop anytime soon. 

Speaking of price, we can expect the 1GHz Pentium IIIs to drop another $100 between now and the start of next year.  No word on how much the 1.13GHz Pentium IIIs will cost upon their release.

The new Pentium III

The current Pentium III processor is based on the 0.18-micron Coppermine core which was introduced in October of last year.  Unfortunately Intel seems to have reached the limits of what they can do with the process as was evident by the trouble they had with pushing the Pentium III up to 1.13GHz. 

With the Pentium 4 not immediately able to become a mainstream solution, for Intel to release a faster Pentium III does make some sense.  The only way to do this however would be to somehow improve the yields on the 0.18-micron process, or give the Pentium III a needed die shrink.  The latter is the course Intel will take and in Q3-2001 Intel will release a 1.26GHz Pentium III based on a new 0.13-micron core.  This new core is known as Tualatin-256 and is a modified version of the Coppermine core that will use Intel’s new 0.13-micron process.  There will be no performance enhancing features of the Tualatin-256 core other than the ability to hit higher clock speeds.  As the name indicates, the Tualatin-256 will actually use an on-die 256KB L2 cache like its Coppermine predecessor. 

The decision to outfit the Tualatin core with 256KB of L2 cache came recently, as the Tualatin was originally supposed to be a 512KB part.  In fact we will see a 512KB Tualatin core however it will be a mobile part only, the desktop part being differentiated from it by the –256 designation in its name. 

The Tualatin-256 Pentium III will still be a Socket-370 part however, as you’ll see in our analysis of Intel’s chipset future, the Tualatin-256 core won’t work on current i8xx boards. 

As far as we know, there is no version of the Tualatin-256 core that is planned to be released with support for currently available Socket-370 platforms. 

Making the Pentium 4 a success

Intel is in a bit of a bind.  They can’t rely on the Pentium III to compete with the Athlon any longer as the processor is currently stuck at a 1GHz clock speed, unfortunately the Pentium 4’s RDRAM dependency, at least for now, may keep it out of the hands of the users that want more than a Pentium III but aren’t ready to make the plunge for a 1.4/1.5GHz Pentium 4, an i850 board and two sticks of RDRAM. 

Intel is going to be working very hard to ramp up production and acceptance of the Pentium 4.  The hope is almost lost to have the 1.4/1.5GHz Pentium 4s enter the market as mainstream solutions, so by the end of this year the Pentium 4 processor will have been able to make a small entrance as a high priced member of the performance market segment in systems priced above $2,000.  It is next year that things get interesting for the Pentium 4.

By the end of the first quarter of 2001 Intel expects the 1.5GHz Pentium 4 to completely take over the performance market segment, and thus paves the way for the 1.4GHz Pentium 4 to begin to be seen in the more expensive mainstream systems ($1,500 - $2,000).  In order to aid the Pentium 4’s transition to a mainstream processor, Intel will be releasing a 1.3GHz Pentium 4 in Q1-2001. 

However the 1.3GHz Pentium 4 surprisingly disappears from Intel’s roadmap by the end of the first half of 2001, at the same time that the 1.13GHz Pentium III is introduced.  There is a great possibility that the performance offered by the 1.3GHz Pentium 4 will be too low to compete with that of the 1.13GHz Pentium III thus causing Intel to ditch the 1.3GHz Pentium 4 after the 1.13GHz Pentium III hits the streets. 

Also before the end of the first half of 2001 we can expect to see the Pentium 4 hit 1.7GHz, forcing the 1.5GHz part down into the mainstream market segment as well. 

By the third quarter of next year Intel hopes to have a 2GHz Pentium 4 out and ready to take over at the top of the performance market segment.  At 2GHz the Pentium 4 could be quite a contender, especially if, by this time, it has a DDR SDRAM platform to reside on.  As you’ll see from AMD’s roadmap which we are going to be publishing shortly, the Athlon is set to hit 1.7GHz at around the same time the Pentium 4 hits 2GHz.  With a DDR platform, the Pentium 4 could begin to compete in the performance market segment. 

Also in the third quarter Intel will be forcing the 1.4GHz Pentium 4 down another level, this time to be found in $1,200 - $1,500 systems.  Remember that this is around the same time that the 1.26GHz Tualatin-256 Pentium III will be released which could end up giving the 1.4GHz Pentium 4 quite a bit of competition.  At this point, the slowest Pentium III listed on Intel’s roadmap is the 1GHz processor with everything else dropped in favor of the higher clock speed Celerons. 

By the end of 2001 2GHz will be the resting place for the Pentium 4 using the Willamette core, however we will see the introduction of a new Pentium 4 core before the year’s end.  In preparation for this, around Q3-2001 the Pentium 4 will be available in two flavors, a Socket-423 version (the initial Pentium 4 socket) and Socket-478 (mPGA478) version.

The Socket-423 Pentium 4 will be the first to debut

The reason for supporting the Socket-478 interface is because the next Pentium 4 core, codename Northwood, will be a Socket-478 only processor.  Not much is known about Northwood other than that it will debut before the end of next year at 2GHz or above and will most likely be a 0.13-micron part.  At this point, the Pentium III will be almost gone from Intel’s roadmap with the 1GHz, 1.13GHz and 1.26GHz (possibly higher) parts, all residing in the $1,000 - $1,200 mainstream market segment. 

Chipsets

We have some good news and some bad news for you all.  The bad news is that after a lifespan of well over two years, the BX chipset will finally die off after this year.  The good news is that it will be taking the horrendously unsuccessful i820 chipset along with it.  And from now until the end of the first half of 2001, Intel’s chipset roadmap is pretty simple. 

The Pentium 4 will continue to use the i850 chipset as its only option from Intel throughout the first half of 2001.  The Pentium III will use the i815, 815E and 815EP.  We already know that the only difference between the i815 and the i815E is that the latter uses ICH2 which adds support for 4 USB ports and Ultra ATA/100, but not many know what the i815EP is. 

Intel's ICH2 taken from our first i815E board

The i815EP is going to take over as the mainstream chipset from Intel and it’s pretty much what the i815 should have been in the first place.  It is basically the i815E without any integrated video, which helps reduce its cost.  In 1000 unit quantities the i815EP chipset can be expected to be $1 - $3 cheaper than the i815E.  And when you’re talking about shipping hundreds of thousands of boards, that is a pretty decent savings.  Unfortunately the i815EP still has the same limitations as the regular i815E chipset, mainly the 512MB memory limit and the 2 x CAS2 PC133 SDRAM module limit. 

Also in the second quarter of 2001, all i815 chipsets (E/EP included) will feature the new B-stepping of the MCH which adds Tualatin-256 support.  This unfortunately means that, as we mentioned before, current i815(E) owners won’t be able to use the upcoming Tualatin-256 based Pentium IIIs.

On the low end, the i810E will continue to dominate since it is cheaper than even the regular i815 and for an entry level system, add-in graphics cards don’t matter all too much.  The i810E will also get an upgrade, basically replacing its ICH1 with a more up-to-date ICH2 making the i810E2 chipset.  This will mean that most of Intel’s chipset line will be using ICH2 throughout 2001, according to Intel the shift towards 100% ICH2 usage is the main goal here. 

A notable absence from Intel's Desktop Chipset Roadmap is the Almador chipset, which you may or may not have heard of as the i830 chipset.  Intel has decided to scrap the chipset in hopes of keeping their roadmap a bit more simple, but more realistically it seems as if Intel is preventing the Pentium III from stealing too much glory from the Pentium 4. 

The Almador (i830) chipset would have given the Pentium III a non-crippled environment to work in, offering essentially the benefits of the i815EP chipset without the 512MB SDRAM limitation.  This would make it a solid offering for workstations and even single processor servers, two areas Intel’s chipsets have been lacking in as of late.  However fears of it stealing too much thunder from the Pentium 4 were most likely contributors to the canning of that chipset. 

It looks like you’ll have to turn to VIA for a faster Pentium III solution with their Apollo Pro 266 (DDR). 

On the Pentium 4 side, you may have heard the name Tehama-E (Tehama being the codename for the i850).  The Tehama-E was supposed to be the Socket-478 variant of the i850 chipset (i850E) however in order to keep things sane on the chipset end of the roadmap Intel has gotten rid of that name.  Instead the i850 chipset will be used on both Socket-423 and Socket-478 boards, much like the BX chipset was used on both Slot-1 and Socket-370 motherboards.

The i850 chipset has another hurdle to overcome.  While the i815E and i815EP can be found in the $30 - $40 range (per unit in 1000 unit packages), the i850 is currently priced at $75.  Even the recently released AMD 760 chipset is only $39 in the same 1000 unit quantities.  It will be towards the end of the first quarter that the i850 will drop in price to close to $50, which is closer to the price of the i840 chipset.  Don’t expect to be able to find i850 boards that are much cheaper than what most i840 boards go for in today’s market (expect a minimum of $200). 

The last chipset we have to discuss from Intel is the Brookdale chipset.  Brookdale will be the first Pentium 4 chipset from Intel with native SDRAM support, unfortunately, according to Intel’s current roadmap, it will come in the third quarter of 2001 and it will only boast PC133 SDRAM support.  According to Intel’s current roadmap, it will be over a year until Brookdale gets DDR SDRAM.  If Intel sticks to this plan you shouldn’t expect DDR SDRAM support from an Intel chipset for the Pentium 4 until the first quarter of 2002.

Final Words

To most of you it should seem pretty simple what Intel has to do to make their product line a success.  The Pentium 4, at higher clock speeds (since the Pentium III is architecturally limited to lower clock speeds than what the P4 can achieve), with a DDR SDRAM platform from Intel could be just what the doctor ordered. 

You don’t have to see the benchmarks to know that RDRAM will be holding the Pentium 4 back.  Let’s take a look at a quick chart that compares the performance of a Pentium III 1GHz on an i815 with PC133 SDRAM to the same CPU on an i840 with dual channel PC800 RDRAM.

The above chart, taken from our AMD 760 Review makes it very clear that even the dual channel RDRAM memory controller found on the i840 (the same memory controller found on the i850 by the way) can’t even begin to compete with the PC133 support of the i815 chipset. 

Intel refuses to comment on the possibility that VIA is working on a Pentium 4 chipset.  Usually when a company denies that something is happening there is reason to believe that they are telling the truth, but when a company refuses to comment on an issue, chances are that something is going on.  It wouldn’t be surprising at all if VIA has a Pentium 4 chipset in the works. 

For VIA, the opportunity would be tremendous.  It would allow VIA to step up to a point where they would become pretty much the sole provider of chipsets for Socket-A platforms and Socket-423 platforms, not to mention a potentially successful Socket-370 line with their upcoming Apollo Pro 266 chipset. 

Chipset availability and memory support, or lack thereof, could make or break the Pentium 4 and unfortunately for Intel, things don’t seem too pleasant for the Pentium 4’s launch which is now just two days away.