It’s Giga Hertz Time
It’s hardly more than 48 hours ago that Intel regained the speed crown for the fastest microprocessor with Pentium 4 at 2 GHz. In return AMD dropped its prices so much that even the slowest Pentium 4 looks like a badly overpriced product. However, this price cut won’t be good enough if AMD wants to remain a big enough player in this processor game. Informed people know that Athlon is an excellent product that doesn’t need staggering clock speed numbers to compete with Pentium 4. The majority of buyers out there however, is buying clock speed. We’ve now reached a clock difference of 600 MHz between Intel’s 2 GHz Pentium 4 and AMD’s 1.4 GHz Athlon processor. It’s tough to convince unknowing customers that 1400 is the same as 2000. AMD needs to come up with something and it should be rather soon.
Northwood Chasing Willamette
Before we will get to the latest developments and ideas of AMD, we should start with the latest Intel stuff, simply because it is shorter, easier to understand and because it is IDF time, where Intel is ‘architecting the digital universe’ for all the poor bastards that think that the many billion year old analog universe just ain’t good enough.
Right now Intel is shipping samples of future Pentium 4 processors with the 0.13 micron core ‘Northwood’ to OEMs, at core clocks of 2 and 2.2 GHz. Those samples are of ‘A2’-stepping, but running just fine, even though the A2-microcode (that is always used to fix minor processor bugs and usually found in the motherboard’s BIOS) is only in beta stage. As we already know, the future ‘Northwood’ Pentium 4 will have 512 kB second level cache, which is one reason why its performance is superior to that of the current ‘Willamette’ Pentium 4. The 0.13 micron process ensures a lower power dissipation, so that Northwood will remain cooler than Willamette, at least at clock speeds of 2 and 2.2 GHz.
Northwood’s performance advance over Willamette is just enough to equalize the performance loss that you encounter when running Pentium 4 in a i845 motherboard with PC133 SDRAM instead of an i850 motherboard with PC800 RDRAM. Therefore the ‘Northwoods’ will be some 5-10% faster than current Pentium 4 processors at the same clock, which is due to more than just the doubled L2-cache size. Other architectural improvements that haven’t been disclosed yet are also responsible for the improved performance.
The new Pentium 4 is currently targeted for a release in December 2001, but since times are difficult Intel may release it earlier.
Hi, I’m Tualatin. Please Call Me Celeron
New Celeron samples with ‘Tualatin‘ core have found their way to the top tier OEMs already quite a while ago. Intel has chosen a funky way to abuse its ‘Tualatin’ core. The Tualatin-Celeron samples are equipped with 256 kB of second level cache and are running at 1.2 GHz, using the good old pathetic 100 MHz front side bus to ensure mediocre performance of this lovely little processor core. The release of those new Celeron processors is expected close to Microsoft’s WinXP launch, which won’t be long anymore, as WinXP just got Gold.
I guess that the 100 MHz FSB information is bad news for the ones of you that had hoped the new Celeron would be a full-blown Pentium III processor. However, our experiences with the Tualatin-256 at 1.13 GHz showed that overclocking from 100 MHz to 133 MHz FSB should not be too difficult. In this case the new Celeron would be a very attractive product.
AMD Reinvents The PR-Rating
Now we have finally reached the funkiest news in this article. I already spoke about it before, AMD needs to do something about the huge clock frequency gap between its Athlon processors and Pentium 4. Shortly before the news arrived at my desk I was discussing this very issue with my editor Frank Voelkel, who is still obsessed by the mission to teach the masses that clock speed ain’t saying nothing about performance. I told him that I gave up on it. Endless regurgitations of “don’t just buy clock speed! AMD’s Athlon is often faster than Pentium 4 even though it runs at a lower clock!” makes a man tired, especially someone who is struggling to get the word out for more than 5 years (with very few brakes). So I said “Frankly, the best thing AMD could do is bring back the good old PR-rating.” Frank didn’t like it. We all hated those PR-numbers used by AMD, Cyrix, IBM and Centaur a few years ago, because they were terribly misleading.
Still, what choice has AMD got? How much do you think a marketing campaign would cost that would make all those potential customers understand that Athlon at 1.5 GHz is faster than Pentium 4 at 2 GHz? I don’t even think it would be possible. When Intel’s engineers designed Pentium 4 for high clock frequencies they knew they had AMD by the balls. So AMD HAS TO fall back onto the good old PR-rating idea unless it wants to lose its balls.
We Call Them ‘Modelizers’ …
What do you think of, when you hear the word model? I guess the majority would associate the same as I and think of beautiful curves, nice hair, soft skin … The married ones of you might think of cars and different engine sizes. Well, AMD is gonna give us ‘models’, but it ain’t any sexy chicks and it ain’t anything for picking up sexy chicks either.
The next Athlon, which besides still remains nameless (It won’t be ‘Athlon4’, but it may be ‘AthlonXP’), will use the ‘Palomino’ core, as we all know. The ‘Palomino’ core is currently available as ‘AthlonMP‘ in SMP configurations and as ‘Mobile Athlon4‘ in notebooks. Its launch as normal desktop processor has been expected for weeks now, but we will have to wait a bit longer, since AMD found some issues with Palomino running at 133 MHz (266 MHz DDR) FSB on many motherboards, specifically on those with VIA north bridges. There won’t be any 100/200 MHz FSB Palomino-Athlons, so that AMD requires that all future Palomino-motherboards have their LCL-filters (for the INCLK bus signals) altered, which won’t make motherboard makers happy and which will take some time.
This time is required anyhow, because the new Athlon won’t be specified by its clock speed anymore. AMD is having very rigorous plans. It starts with the printing on the chip. Future Athlons will be specified by ‘MODEL’ numbers. For instance a Palomino-Athlon that runs at 1.4 GHz will be MODEL 1600, because AMD considers Palomino 1.4 GHz to be at least as fast as a Pentium 4 1.6 GHz. The printing on the chip will be ‘A1600 …..’, even though its a 1400 MHz part only. The same is valid for motherboards and their POST. A new BIOS Writer’s Guide prohibits the BIOS from ever displaying the true core frequency of Palomino! It has to display the MODEL number instead. Unfortunately, H. Oda’s WCPUID still reports the actual clock frequency and even WindowsXP is still reporting the clock rather than the MODEL number. It is unlikely that AMD can force Microsoft into changing WindowsXP code, now that it has just gone Gold.
I don’t want to comment much on this issue, because I can very well understand AMD’s difficult position. Something has to be done about the clock frequency issue, but I don’t feel that this return of the dreaded PR-rating in form of MODELS is going to be a very good idea.
OEMs are already voicing their concerns. They are afraid that customers who bought a MODEL 1600 could sue the PC-maker once they discovered that the processor inside is only running at 1400 MHz. It might not come to that, but the introduction of MODEL-numbers instead of real clock frequency will create a lot of confusion. I hope that AMD will reconsider. Maybe the last word hasn’t been spoken yet. I suggest AMD runs some kind of reader contest on their website, collecting suggestions for how it could tackle the clock frequency problem. This won’t cost much and it might bring us a better solution than the return of the dreaded PR-rating.