If you get into a computer shop and watch a little their managers and shop-assistants you will often hear them saying: "Well, if you don't have enough money now I recommend you to take this one. In a little while you can upgrade it". This is really a bad guy who says it, or let's say he's cunning. Why? Cause an upgrade does not exist.

I'm really tired of dull talks about an excellent scalability of platforms, about replacement of a processor with a more powerful one in a half a year and so on. Let's turn to the recent times and platforms that we successfully buried at that time. So, everything started with... let it be Slot 1.

Slot 1. This was a new high-efficient platform for a new Intel's processor Pentium II. Surely, there was a new chipset as well - i440LX. OK, for such a purchase we sold all Socket 7 stuff and bought a new mainboard and a new processor. Some unlucky owners of Socket 7 boards with SIMM had to buy a new memory as well. Excellent! But just a bit later the old chipset gave a way to the i440BX. We left the Slot 1 just in case, though it wasn't useful anymore as the 100 MHz FSB wasn't supported by the i440LX chipset, and all new processors (which we wanted to take for an upgrade) were going to be designed only for a 100 MHz bus. So, we had to change a board. Which entailed replacement of the memory with PC100 as it became popular by that time.

After that Intel tried to make us buy i820 + RDRAM. It was a grandiose marketing step as with such a platform a user should have forgotten about all processors and chipsets but for Intel's ones. Besides, a user had to pay an arm and a leg for the motherboard and RDRAM. But that idea failed. As you know, nature adhors a vacuum, and Intel's place was taken by VIA which just increased a SDRAM frequency by 33 MHz, and we got a PC133 platform.

By that time the Slot 1 passed on, though all users were engaged in changing their "insides" and didn't notice it. By the way, it was the only time when the industry "went the wrong way" by producing Slot 1 -> Socket 370 adapters which were a great benefit for owners of old boards. Such "mistakes" never took place again, and Slot A -> Socket A adapters remained a legend.

AMD, such a friend of all "unfortunates" was, in fact, a follower of the Intel's strategy. Slot A -> Socket A, FSB 200 MHz -> 266 MHz.... Besides, AMD had also a DDR SDRAM! First, a user got a board on the VIA KT133 (200 MHz FSB), then a new processor working at FSB's 266 MHz, then a new board for this CPU on the KT133A. But that was not an end at all, as the potential of newer Athlons could be realized only with new DDR memory! Moreover, the DDR had some variations: first it was PC1600 DDR, then came the PC2100, and soon we will see a PC2700 (DDR333)…

Having released a Pentium 4 Intel quickly replaced one socket with another: Socket 423 with Socket 478. But I don't see why they did it as a 0.18 micron Willamette is produced in both versions and there are boards on the new i845 both for Socket 423 and for Socket 478. So, if all of them are compatible why do they develop a new socket? Sure, it might be an improved version, but why then they didn't make such chipset from the very beginning? It is like a secret of the today's boards on the i845. Everybody knows that in a month the i845 will get a PC2100 DDR support. I also understand that a throughput of the PC133 is not enough for the Pentium 4. Who needs the today's i845?

As it turned out, they fool us, right? Well, not always. It is true that the Pentium III with a 133 MHz bus works faster than the Pentium III 100 MHz FSB. And it is true that the Athlon with a 266 MHz FSB is more efficient than the Athlon 200 MHz FSB. And AMD processors do need DDR memory working at least at 1200 MHz. But such a desirable and long-awaited upgrade is not possible. Well, in fact, it is, but the outcome will be zero because a processor released a year later needs an absolutely different platform. They deceive when promise an excellent scalability forever. This concerns not only a processor: just try a powerful GeForce3 first together with a Pentium III 600 MHz and then with a Athlon XP 1500+ or Pentium 4 1.7 GHz and then compare.

On the other hand, it's no good when a platform remains the same, and neither FSB nor the memory change. For example, Intel Pentium III-S and Celeron on the Tualatin. Socket 370? But what if to install this processor into my old board? Our Socket 370 is a bit different, it needs a board with the same socket but on the i815 B-Step. You got a board on an i815EP just a few days ago? I'm really sorry... Be careful, don't mount a new processor on this board as you may burn it.

But it is not the worst thing. Suppose, a processor maker plays a fair game. For example, when AMD released the Athlon Thunderbird for Socket A it didn't change anything, at least, the components were back compatible. But did it help them? Nope. Just try to install an Athlon 1300 into an old mainboard designed for Athlon 700. You will fail because the power consumption is now much higher. A weak VRM chip doesn't suit here, though the manual says "Athlon 1000 MHz and more". It turned out, that that "more" is limited by 200 MHz.

There are plenty of other bright examples. It is clear that manufacturers of computer components do not allow an effective upgrade. It is unprofitable. First because it is necessary to maintain sales at a high level. Secondly, to make an upgrade possible it is necessary to develop components scrupulously and carefully and it is desirable that a manufacturer know his "tomorrow". And who, among IT giants, knows its future plans in detail? Nobody.

So, just forget about an upgrade. An upgrade as a way to change cardinally a performance of your system without replacing at least half of the components does not exist. Moreover, two upgrades coming one after another cost 1.5 dearer than a new computer with the same characteristics. C'est la vie.

Write a comment below. No registration needed!