Reg readers take the Intel A train
Letters My observation about Intel's use of 'A' to differentiate between two P4 processors, both clocking at the same speed, drew a large and mostly disapproving, response. First, a summary of the lines of attack, and then we publish some letters.
Intel has been using A, E and M naming conventions for a long time - remember the Celeron 300A?. Yes I do, and I'd have headed off a lot of emails if I had mentioned this first time around. However, it was the 2GHz and the 2AGHz sysmark results published on Intel's web site which prompted this new (for me) train of thought the use of letters as a clock speed marketing device
What has Intel's naming conventions got to do with AMD's True Performance Initiative?
Let me make it clear that Intel is not plucking letters out of the alphabet in response to AMD marketing moves on clock speed. However, Intel maintains a "clock speed is everything" stance when talking about this issue to the consumer space. But two chips running at the same clock speed, one with a letter appended to indicate superior performance, muddies this absolutist stance. We're moving into PR ratings and this e use of letters plays into AMD's hand.
AMD's model numbers do not imply performance equivalence with Intel chips, but with earlier Athlons Oh yeah? This is a marketing issue, the model numbers certainly do imply performance equivalence with faster clocking Intel CPUs, even if the explicit comparison is made between earlier lines.
It puzzles me why you would write that the "A" denotation of the Northwood P4 has anything to do with AMD's PR rating, and try to make it out as some sort of "victory" for AMD. Intel has historically picked letters to differential different cores and specs with the same mhz ratings long before AMD devised their PR ratings.
Is the "E" of a coppermine P3 because of AMD's PR rating too? The P3 550, P3 600, and P3 600B all had 550E, 600E and 600EB counterparts. It wasn't because of any PR rating, it was to identify each core without having to guess "Did I get a Katmai P3 600 or a Coppermine P3 600?"
Here's a link to Intel's Pentium3 and Celeron s-Spec charts to show you that letters are nothing new to Intel:
http://support.intel.com/support/processors/sspec/icp.htm (note the use of the letter "A" to differentiate between two Celeron 300, 533, 1000, and 1100mhz processors).
I'm not writing this to say that AMD's PR ratings is a bad thing, in fact it does help the less educated consumers make a more intelligent choice, but I'm saying that you're reading too far into the addition of a single letter to the P4; a letter which has been used by Intel before, and will be used again.
Overall I do like The Register, so keep up the good work. I just have a feeling that the article I metioned was only put on the website due to a slow news day.
Coquitlam, BC, Canada
You suggest in this article that Intel is jumping on the AMD wagon in terms of the PR ratings that Athlon CPUs use. Specifically, you suggest that the "A" designator for the Northwood part "indicates speed."
I am sorry to tell you this, but you are making a deal out of a common Intel practice.
Intel Celeron 300: Covington core 0.35m, no 2nd level cache
Intel Celeron 300A: Mendocino core 0.25m, 128KB 2nd level cache
Intel Celeron 500: Mendocino core 0.25m, 128KB 2nd level cache
Intel Celeron 500A: Coppermine core 0.18m, 128KB 2nd level cache
in the same manner, there were:
Intel Pentium III 600E, 600B, 600EB etc. differentiating 0.25m with 0.18m (Coppermine) cores, as well as 100MHz or 133MHz FSB.
Maybe Intel will introduce a PR rating like AMD, but the Intel Pentium 4 2.0A is not a move into that direction. :-)
Your article on Intel’s part numbers is not altogether true. Maybe you should have done some research before you decided to post it. If you did, then you would have discovered that before AMD’s introduction of model numbers, both Intel and AMD used letters in their part numbers to distinguish core changes within a product family.
In fact, Intel was the very first to use something other then just clock speed for their part numbers. With the introduction of the Coppermine core into the Pentium III family, Intel added E and/or B to the part number, such as the 800EB MHz Pentium III. Also, when AMD introduced the Thunderbird core to the Athlon family, a B (t-bird w/100 MHz DDR FSB) or C (t-bird w/133 MHz DDR FSB) was used to distinguish them apart from the original core.
So, using an A to distinguish the Northwood core from the Willamette is nothing new. But, you must understand that having a letter or two after the clock frequency is entirely different then using a model number scheme. Don’t get wrong, I love Athlons, and I think that the model number scheme is the proper way to go in this era of computing.
In your next-to-last paragraph, you commented: "But since when did a letter indicate speed?"
I'm dating myself, but I got into electronics and IT during the discrete/RTL/DTL/TTL era, where a lot of work was still done with discrete transistors, small ICs, and other discrete components. LETTERS WERE EVERYWHERE! They still are.
On transistors, for instance, an "A" or "B" or "C" suffix indicates a higher-spec part: Speed, input capacitance, current gain families, performance over temperature, package type, and so forth. On some products, letters indicate reliability levels or burn-in levels or temperature "soaks." Letters were/are used to indicate variations and later versions of a product; just a little faster, just a little more stable, just using a new, improved version of the hardware "core" in manufacture... So Intel is only continuing a tradition of long standing.
The experienced hardware engineer or hacker knows the difference; knows that letters can be invaluable indicators to shades or degrees of performance between similarly-designated models. Software hackers and developers and SysAdmins usually don't need to know, although many of the ones that I know keep up with the intimate hardware details out of native curiosity and a CYA mentality, or by a desire to have the best/fastest/newest hardware.
The general public only needs to be concerned insofar as it affects the product they receive. What does it do differently, by how much, how do I measure it, and what is it worth to me? In this sense, at least, it does appear to be a marketing ploy by Intel to counter AMD's "speed rating" numbers.
For you and your readers: What is the difference between a 2N2222, 2N2222A, and 2N2222P?
Well, I agree that it does sort of do that incidentally, but wouldn't you say that it's not the primary reason? Between the Willamette (.18 micron process) and the Northwood (.13 micron) cores for the P4, how else would you differentiate between the parts? Not to mention the mobile part, the P4M which also has to be thrown into the mix.
So, calling the Northwood the P4A is really sort of a necessity without even considering the marchitecture aspect. Remember the cacheless Celerons (P-II Deschutes core, .25 micron), and then the Mendocino core starting with 300 MHz? The 300 MHz part with the 128 kB on-die L2 cache had to be called the 300A to avoid confusion with the Celeron 300 *without* L2 cache. The cacheless Celeron only went up to 300 MHz, so only the 300 MHz Mendocino required the "A" label.
The "A" label is really nothing new, as you can see with the Celeron. The P4s with clock speeds higher than 2.0 GHz (ie., for which the Willamette core do not reach) do NOT have the "A" label. So, the "A" just tells us which is the Willamette and which is the Northwood.
Besides, the way you see it, it all comes down to advertising. Most retailers wouldn't even print the "A" label when listing specs in their ads, even if it was a P4A 2.0 GHz. How come? The simple fact that most retailers don't pay enough attention to technical details on the products. They will simply list the clock speed and be done with it.
Are you convinced yet? I don't think that the "A" tells us anything about 's marketing game plan. At least they're not using ICOMP any more...
Herman P. ®