Original Link: https://www.anandtech.com/show/665



On the desktop side, Intel may be coming off one of their worst years in recent history. With AMD beating Intel to 1GHz, processor shortages yielding price hikes, and the forced recall of the Pentium III 1.13 GHz, it seemed that Intel was doomed. Or so it appeared.

Intel may have taken a beating on the desktop market side, but as far as mobile processors go, Intel remained the uncontested top dog. With AMD only offering K6 series processors for mobile use and Transmeta products only slowly finding their way into systems, almost every notebook, be it budget or performance level, was powered by Intel. Yes, the year for Intel's mobile products has been good, but the lack of competition may not last long.

As we mentioned in our Comdex Fall 2000 Summary, AMD appears to be jumping into the high performance mobile market, as could be seen with ALi's prototype mobile DDR platform. Perhaps phasing out the K6 plus series processors in favor of faster Duron and/or Thunderbird parts, it may not be long until we see AMD take a jab at Intel once again.

Just exactly how does Intel prevent AMD from taking even more market share in which Intel has a stronghold? Examining Intel's latest strategies for the future, we can what Intel has up their sleeves, and it does not look bad at all. With the introduction of new mobile processor parts, new mobile chipsets, and even mobile chips set out to go head-to-head with Transmeta's latest offerings, Intel has quite a bit planed for 2001.

As you read this roadmap, please note that 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.



An old dog: The Coppermine mobile Pentium III

From what we have seen, it will be quite some time before we can bid adieu to our old friend the Coppermine mobile Pentium III. The Coppermine core remains completely unchanged in the higher speed mobile Pentium IIIs, including the standard 256Kb of cache, the normal 100 MHz front side bus, and Intel's SpeedStep technology.

For those not familiar with SpeedStep, let's take a look behind Intel's method of increasing battery life. Essentially, SpeedStep technology increases battery life by decreasing power consumption when running on battery. Since power consumption is a function of capacitance times frequency times the square of the voltage, there are quite a few way to reduce power on the whole. Obviously the best way to reduce power consumption is to decrease the voltage used by a component, as this is the variable that is squared in the power equation. It was therefore no surprise that Intel choose to reduce processor power by decreasing the processor core voltage.

The downside to the reduction in core voltage is that a reduction in clock frequency is also necessary. This is analogous to overclocking a CPU: the higher the core voltage during an overclock, the higher the frequency that can be reached. By taking these two factors into account, one can see how Intel came up with the idea of SpeedStep technology, which dynamically underclocks the CPU and decreases the core voltage when running on battery. For example, a mobile Pentium III 850 MHz running on AC power gives the CPU the 1.6 volts necessary to run at 850 MHz. Once the cord is pulled and the computer starts running on battery, the 850 MHz CPU dynamically underclocks itself to 700 MHz; a speed which requires .25 volts less or 1.35 volts.

The dynamic clock changes occur by actually using the operating system to help alter the power state. When a change in power source is noted by the operating system, it sends the CPU into a standby mode for a very short period of time. It is in this time window that SpeedStep is able to modify the CPU multiplier as well as the core voltage. The system is then "woken-up" in the new power state and is ready to go.

Interestingly enough, Intel's SpeedStep technology is still no match for AMD's PowerNow solution. As we explained in our PowerNow review:

Instead of constantly running a CPU at its maximum frequency, PowerNow! allows the CPU to dynamically adjust its operating frequency according to the needs of the tasks that the CPU is running. For example, launching an application requires immediate attention of the processor, meaning that the CPU would jump up to a fairly high frequency in order to make sure that the application loaded quickly. However, upon entering the application, the system has to wait for user input, meaning that there's no reason for the CPU to sit there at full power waiting for the user to input commands. So, PowerNow! instructs the CPU to drop in clock speed. Then, as the user begins to use the program, the CPU clock speed varies according to the needs of the application and the tasks that are running.

As we should all know by now, lower clocked processors don't always require that the same amount of voltage be supplied to them as those with higher operating frequencies. It would thus only make sense that as the clock frequency of the processor is varying that its core voltage should vary as well, which is exactly what PowerNow! does.

Strangely left out was any mention of Intel's "Quick Start" technology, which we first covered at this Fall's IDF. The technology, which uses the "hurry-up and get idle" theory, was supposed to compete with AMD's PowerNow! technology, but with Intel not promoting it anywhere, we don't know what to make of it.

With all the work Intel has put into the mobile Coppermine parts, it is no surprise that In the performance segment the mobile Pentium III chip remains top dog all the way up until late second quarter 2001. In fact, the current 850 MHz Coppermine mobile Pentium III should remain king of the mountain through the first quarter of 2001. It is not until April of 2001 do we see the mobile chip that will have many drooling: the 1.0 GHz Coppermine mobile Pentium III. This chip will be released along with a 900 MHz part, however this appears to be the end of the line for the Coppermine parts. And with it's death, comes a new kid on the block.



New Tricks: The Tualatin core

As you may have read about in our Intel Desktop & Chipset Roadmap from this month, the aging Pentium III Coppermine core seems to finally be reaching its limits. With the recall of the 1.13 GHz Pentium IIIs, it became quite clear that the 0.18 micron core was just not small enough to push the Pentium III up to the speed necessary to compete in today's market. Intel's solution: the Tualatin, a processor that will also quickly find its way into mobile systems of the future.

The Tualatin core, which is a shrink of the existing Coppermine core from 0.18 micron to 0.13 micron, seems to make the most sense for the mobile platform. A shrink in the manufacturing process may be important on the desktop side in order to allow for higher clock speed processors, but in the mobile market, a die shrink does much more than that.

First off, the reduction to a 0.13 micron process means that for each clock speed less heat is generated. Heat is a major concern when designing notebooks, as processors with a high heat output can not be sufficiently cooled while in the space constraints of a notebook. As you can imagine, cooling a processor in a notebook is quite a task and one that is not easily dealt with.

Although heat is a major concern, the cause of this heat is an even greater one. The reason that extensive heat is produced in higher speed processors is that the power needed to drive this chips is large. The resistance this power encounters inside the chip is what generates the heat. When the manufacturing process is shrunk, not only is die size and heat decrease but the resistance encountered within the chip is also decreased. For this reason, the Tualatin's 0.13 micron process will not only produce less heat than it's older 0.18 micron Coppermine brother but will also require less power. And less power means longer battery life and higher processor speeds.

As result of this decreased power consumption and increased clock speed, the Tualatin core is key to Intel's mobile success. Intel's plans currently call for the introduction of the Tualatin processor with SpeedStep technology in the second half of 2001, most likely some time in the third quarter, which means it should be announced nearly the same time as the desktop Tualatin-256 part. The chip will arrive in the mobile platform at two speed grades which have been loosely defined as greater than 1.0 GHz and greater than or equal to 1.2 GHz. Since there will already be a 1.0 GHz Pentium III part, it only makes sense that the Tualatin is clocked slightly above this in its introduction. We suspect that the lower end Tualatin may clock in at 1.13 GHz with a 1.2 GHz brother on the higher speed front.

The most interesting aspect of the Tualatin mobile part is its included cache size. You may have noticed that the Tualatin mobile chip is named just that, unlike the desktop's Tualatin-256 part. The reason for this is that it appears that Intel has stuck to its original Tualatin plans in the mobile front, as the chip contains 512K of cache. This is quite an odd decision on Intel's part, as it means a few things.

First off, the increase in cache size (over the desktop part) means that the mobile Tualatin part will actually use more power than the desktop part. Since the included cache increases die size and complexity of the chip, the mobile Tualatin parts will consume more power than the 256K desktop parts. The second thing that this means is that the mobile Tualatins will actually be more powerful than the desktop Tualatin-256. The increased cache size means that the mobile Tualatin part should perform faster than the desktop part in a clock for clock comparison. This is quite interesting, considering in the past we have become accustomed to seeing lower performing chips in mobile systems.



The new chipset: Almador-M

Just as in the desktop's Tualatin-256 case, the mobile Tualatin will require a new chipset. This chipset comes in the form of a mobile version of the desktop chipset that Intel canned: the Almador-M. Slated to arrive in the third quarter of 2001, along with the Tualatin, the Almador-M chipset is the direction all Intel based notebooks will take.

The Almador-M chipset comes with quite a few features that make it more powerful than the currently available 440 BX mobile chipset. First off, the Almador-M will finally allow for notebooks to handle both a 133 MHz front side bus and 133 MHz SDRAM support. This will finally allow for front side bus speed increases in notebooks which will begin to bring these systems up to speed. As we have shown time and time again, increased front side bus speed usually equates to increased performance.

In addition to an increased front side bus speed, the Almador-M chipset will include integrated graphics, most likely based off of Intel's current i752 chipset. As we have seen in the past, the Intel integrated graphics chips perform sub par when compared to other solutions. The same will hold true in the mobile market. With the announcement of NVIDIA's GeForce2 Go and rumors of ATI also working on a Radeon based mobile part, there is no way that a i752 based video solution will prove attractive to 3D notebook buyers. It is therefore odd that while the Almador-M has AGP port support for an external (nonintegrated) AGP 2x and 4x graphics chips yet Intel will only release the Almador-M with the integrated graphics solution. This may end up hurting system builders in the long run, as the integrated video will increase chipset price which will only be inflated more by the use of a third party video controller. The new platform will also have a new I/O controller hub, named ICH3-M. This new controller will add support to USB 2.0, but other than that not too much else.


Mobile Pentium III Module (lower square is the Pentium III, upper square is the North Bridge) - click to enlarge



An old friend: the mobile Celeron

The trustworthy Coppermine128 based mobile Celeron processor appears to remain Intel's method of attack on the value mobile market. With speeds easily ramping from the current 700 MHz speed limit all the way up to 900 MHz in the fourth quarter of 2001.

Intel plans to adapt the Celeron to the Almador chipset upon the Almador-M's launch in the third quarter of 2001 by releasing a new breed of the Celeron: the Celeron-T. Unlike the Pentium III that will actually be based on the Tualatin core, the Celeron-T will remain a 0.18 Coppermine128 based core. The Celeron-T is simply the current Coppermine based Celeron that is able to work with the new Almador-M chipset, making it an upper level value solution. The Celeron-T will arrive clocked at 866 MHz, and will slowly move to 933 MHz by the fourth quarter of 2001.

Since the mobile Celeron line has still not hit the speed limit that the mobile Pentium III is about ready to hit, it is easy for Intel to continue production of this chip while leaving the core unchanged. It seems that Intel feels secure with the Coppermine core hitting 1.0 GHz, so it is unlikely that the Celeron solution will move to the Tualatin's new process until after the Celeron is ready to hit 1.0 GHz, which will not be until 2002.



It's all about power

In addition to the full size notebook plans that Intel has, 2001 should bring some interesting solutions to the mini notebook and sub notebook arenas.

Intel is targeting their low voltage processors in the mini notebook market. Intel plans to have a low voltage Pentium III 700 in quarter one 2001, followed by a low voltage Celeron 600 in quarter two and finally a low voltage Pentium III 750 in fourth quarter. Intel is pushing the 440MX chipset for use in these sub notebooks, suggesting that the chipset can easily packed in a small space. It makes sense that Intel is pushing the 440MX into the sub notebook market, as it appears that 440MX based full size notebooks will be quick to disappear, replaced with 440ZX, 815EM, and Almador-M chipsets. This way, Intel can continue to sell existing 440MX chipsets without taking the time to design a new one.

Intel also plans to make some waves in the sub notebook market. Intel has made their Pentium III 500 and 300 run at 1.1 volts and less than 1.1 volts respectively. Once again using the 440MX platform in these systems, Intel is targeting this processors in fanless systems that require less than 7 watts total power consumption. A 600 MHz 1.1 volt processor is expected to come in the third quarter. of 2001 in a Pentium III package. The Celeron is not left out of the sub notebook field, with a 500 MHz 1.1 volt Celeron coming in the first quarter. of 2001. This will be followed by a 600 MHz 1.1 volt Celeron processor in third quarter.

It seems that Transmeta will not be the only company making chips for sub notebooks and mini notebooks. We suspect that the "low voltage" line is more targeted at the Crusoe chip, and here Intel does not seem able to compete to the Crusoe's minimal power consumption when at idle. Intel may be able to steal some of the Crusoe's market in the sub notebook arena, however, where Intel's chips use only 1.1 volts.



Conclusion

Not too long ago, one could judge the performance of a desktop system by seeing how far behind notebook machines were. It used to be that when notebook processor speeds began to surpass a user's desktop speed, then the desktop was getting old. No longer is this the case.

With Intel out to maintain their stronghold on the mobile market, the processor performance difference between notebooks and desktops seems to be shrinking at a breakneck pace. The mobile Pentium III running at 1.0 GHz will hit the market in April of 2001, at a time where many of our system are running at half the clock speed or less. Intel will continue to increase the performance of notebooks with the release of the Tualatin processor with 512K of cache, a part that the desktop industry will only see included with 256K of cache. Most likely a result of not wanting the Tualatin based Pentium III to compete with the Pentium 4, it still does not make up for the fact that the mobile Pentium III chips will soon be more powerful than the desktop version.

The plans may look strong for Intel in 2001, but the competition may quickly become fierce. With AMD leaning in the mobile direction and Transmeta producing faster and faster speed Crusoe processors, Intel has to meet the consumer's expectations. After loosing all the ground they did in the desktop market by not executing properly, Intel must know how crucial the coming year is for mobile parts. Luckily for the consumer, AMD and Transmeta know the same thing, meaning we could have quite an interesting year head of us in the mobile arena.

Log in

Don't have an account? Sign up now