Celeron 366 and 400: the accessible speed
Did anybody believe seriously that Intel was going to remain of crossed arms hoping that AMD should on taking "his" market from him? In spite of everything what could criticize him like monopoly of fact that is in microprocessors and chipsets, nobody can doubt that Intel has the potency and the skill to change an adverse situation with a simple but effective blow of hand... and has done it again.
Brief (?) history of the Celeron
For that they do not realize still very well what the Celeron is that thing about, we are going to sum up quickly (or not so rapid) the evolution of this chip, insulted in his origins and that now seems to arise like the panacea as for quality / price.
About the spring of 1.998, Intel realized that the segment of market of processors of average - low level (what the Englishmen name the entry-level, literally the level of entry or of game) was being cornered by the mikes K6 of AMD, more powerful than the "cheap" mikes of Intel, the Pentium MMX. Also, the arrival of the K6-2 with his progress at bus level (100 MHz instead of 66) and new instructions for 3D (them 3DNow!), all this at a price much more accessible than the slowest Pentium II, they were imposing an immediate answer.
Since Intel is not partial of lowering the prices without much ado, he decided to re-use the strategy of making mikes of cut away services that it had already used in 386SX and them 486SX. On April 15 it presented the new one "Pentium II-SX", which was named Celeron; it was consisting of a nucleus identical to that of Pentium II but without 512 Kb of secondary cache memory or L2, in addition to presenting the naked circuit badge, without the black plastic casing, to low the price of costs.
The criticism was not late in coming from all the fronts. The new mike had a yield very low in any type of applications that were making use of the cache memory L2, as there are all the ofimáticas, Internet, treatment of images... In fact, the yield saw so penalized that the model to 266 MHz was becoming only 233 MHz easily overcome by his ancestor the Pentium MMX.
Some games were the only application field, thanks to his very good mathematical unit, completely identical to that of Pentium II; in this respect, to say that the very widespread rumor of that the Celeron are defective Pentium II lacks absolutely foundation. Simply there are models with little or no cache memory L2, as it is appreciated in the following photos:
Top photo: mike Pentium II in all his interiority, without the black plastic casing. It takes the cache memory distributed in several chips (the black rectangles) to the sides of the nucleus (the central metallic piece).
Low photo: Celeron Mendocino; he lacks casing. It takes 128 Kb of cache memory L2 integrated to the nucleus, hence his big size.
It found it hard enough to Intel to admit his error, what undoubtedly it helped to that the K6-2 were obtaining a strong implantation, almost over the expectations of the same AMD. But enclosed Intel can recognize an error when the numbers make it clear, and finally on August 24 the Celeron announces with cache memory, Celeron "A" or Celeron Mendocino, in fact the technical name of the nucleus of the chip (based on the Deschutes that most of Pentium II take to 333 MHz or more, together with 128 Kb of cache memory L2 integrated).
But the fourth part of 512 Kb is 128 Kb only, if the mathematics do not trump me, so these mikes will be slower enough than Pentium II: truth? Since not, for nothing. The big trick in favor of the Mendocino is that his cache memory works at the same speed of the mike, instead of to the half as in Pentium II. This is something of a vital importance, since it means that even the slowest Mendocino has a cache memory that works 75 MHz more rapid than that of the most rapid Pentium II, as it is appreciated in the following picture:
This was leaving us with a few quite interesting mikes as for quality / price, much adapted for works of office computerization and games, with only two problems: his relatively low maximum speed (only 300 and 333 MHz models existed, while Pentium II comes up to 450) and a cost still slightly higher than a solution based on the K6-2. Well, since both problems have evaporated overnight...
New mikes: ancient sockets?
Very much it has been commented about the introduction of the Slot1 like method of connection of Pentium II and the Celeron. For the one that does not know the topic, the problem is in that the above mentioned connector and the bus P6 on which it is based are protected by strong patents property of Intel, who refuses to license them completely to any of his competitors.
This has led AMD to extending the useful life of the socket Socket7 providing him with a 100 MHz speed, what has allowed him to keep on offering an economic platform for the family K6 while it prepares the SlotA, its own version of a connector equivalent to the Slot1 for the future K7.
But although the Slot1 has expired perfectly with the intention of Intel of creating a virtual monopoly as for connector for the mike and chipset, it presents a problem: it is excessively expensive to be able to compete with serviceability on the market of average - low level, where a few thousands of pesetas can decide a buy. So he has thrown reverse gear and has decided to create a new socket for his mikes Celeron, the called one Socket370 or PPGA370.
The above mentioned socket has the size and the form of the Socket7, but it turns out to be incompatible with him since it uses a type of different bus and also he adds many other pines, up to a number of 370 (they understand where from the original comes nombrecito: truth?). This class of socket lows the price of the manufacture opposite to the system of badge of printed circuit that they have come using the Celeron and the Pentium II, being perfect for Celeron Mendocino who do not need extra space to lodge the cache memory chips on having integrated it.
It must be born in mind that it is a question only of a change of physical interface, the yield remains absolutely constant since there is used the same logic (the bus P6) and works at the same speed (for the time being 66 MHz, in the future perhaps up to 100 MHz).