Pentium III: an immature mike?
At the end of 1.992, tired of that his competition (AMD and Cyrix especially) should copy not only his designs of microprocessors but the same trade names, Intel decided that the name of the successor of 486 would not be 586, but Pentium...: as if it did not have the Latin enough in spite of being a dead language, above it is a publicists' grassland yankees!
In that epoch it seemed impossible that that curious name was perpetuated until the XXI.th century. but while some things change, like for example the meritorious step of AMD of company "asimiladora" of ideas foreign to powerful technological innovator, others remain. The "classic" Pentium came first, then the renovation Pentium MMX, the powerful one and innovator Pentium Pro, the successful Pentium II... and we are already before Pentium III also known as Katmai. Will he be a worth member of the scale Pentium? Let's see it...
A glance on the outside
At first sight, a Pentium III (from now on "P3") looks alike very much to a hybrid of Pentium II and Celeron. Ahead it has the typical form of black cartridge to tune in to the Slot1 that already had the Pentium II... but for another side it is naked, like the Celeron.
Intel names this format S.E.C.C.2, to separate it from the format S.E.C.C. of Pentium II and of the S.E.P.P of the Celeron (they will not have got tired very much at the time of looking for names, not). The target looked, on having eliminated one of the faces of plastic, is to increase the refrigeration of the chips, so much of the mike in itself as of the chips of cache memory L2, since thus the heat spendthrift supports straight on them.
Otherwise, little on that to comment. The new format is a good idea, although it is not anything that it thrills too much, but the mike in itself does not have anything outstanding physically, looks alike very much to the most recent Pentium II.
Is Pentium III innovative? And if it it is, why? Let's compare him with his immediate predecessor, Pentium II:
Well, it seems that there are no a lot of differences: truth? Since not, they are not. During enough time, we wait for many that the P3 should come to the market with 64 KB of L1, or a 133 MHz bus, or with the L2 working at the same speed of the mike (as in the Celeron and the AMD K6-3)... vain hopes. Technologically, the current one P3 is completely identical to a 350 MHz Pentium II or more except for the new instructions SSE.
Pentium III = Pentium II MMX-2?
As we were saying, the evolutionary jump that has ended in the P3 has been the incorporation of 70 new instructions called officially SSE, Streaming SIMD Extensions (extensions SIMD of flow), although for the long time we knew them as KNI (Katmai New Instructions, new instructions of the Katmai, the technical name of the P3) and many people prefer to call them, more commercially, MMX-2.
Probably the most suitable name is not the official, but preferred of the publicists: MMX-2. The original instructions MMX mean Multimedia eXtensions, a logical name if we bear in mind that they were created to increase the yield in the multimedia applications (those that combine image, sound and/or video).
The problem of the above mentioned instructions MMX (that there incorporate all the mikes from already outstanding figures Pentium MMX and AMD K6) was that they could not be used together with the FPU, the mathematical unit of floating comma of the mike, of enormous importance in applications like games or CAD. This was doing that many programmers were not optimizing the programs for MMX, since "only MMX" or "only FPU" was not an agreeable election.
The P3 solves this problem of two ways:
To understand the process that these instructions continue to accelerate the calculations we can concentrate on the word SIMD: Single Instruction, Multiple Byline; the only instruction, multiple information. These instructions allow to realize the only complex operation with several information instead of realizing several simpler operations, being able to do up to 4 operations in floating comma for every clock cycle.
Also, some of these 70 new instructions optimize the yield in multimedia paragraphs like the reproduction of video MPEG-2 or the speech recognition, while others accelerate the access to the memory.
His problems? Clearly, that so that yield increase exists, the applications must be optimized for the new instructions. Namely that in applications not optimized (99,99 % of the current ones), a Pentium II and a Pentium III at the same clock speed give a few identical results.