Intel's tricks!

Author: Razvan MIHAIU
razvan_rem@rem_mihaiu.name (please remove '_rem' and 'rem_')
From: www.mihaiu.name
Date: 07/03/2001



Welcome to Intel's tricks; I am sure that this is an intriguing title for most of you. Intel is a semiconductor manufacturer with a very good reputation to the general public and this reputation is well deserved for the most because Intel was a leadership force in the semiconductor industry for more that 3 decades. Historically Intel started the microprocessor business with their famous Intel 4004 - the first real microprocessor in the world. Intel managed to stay ahead of the competition for this entire period - with short downfalls, but nothing of real significance. What makes this company so successfull? On one side is the hard working Intel enginner. On the other side... well this is what we will analyze in this article - the 'dirty games' played by Intel to its competitors.

In order to gain advantage over their competitors Intel has used production & marketing tricks; I will start with a production trick - probably the oldest trick that was evident for the general public - at least to the well informed one.

486FPU

Intel's 386 line of processors had a tremendous success starting from 1985 and up to 1989. When the processor line was close to it's EOF (end of life) another company (AMD) had produced a 386 that was significantly faster than Intel's 386 offering at the time; Intel's 386 top operating speed was of 33Mhz while the fastest 386 produced by AMD was working at no less than 40Mhz. At that time 7 Mhz was nothing short of impressive. In order to stop AMD from gaining market share Intel quickly pushed 486 into arena. The new 486Dx with it's improoved core was much faster than any microprocessor from the previous generation. However soon Intel faced a new problem: the price. The 486 was not able to compete against the old 386 on a price basis and because of this AMD was still gaining significant market share in the low-end sector. Designing a new processor for the low-end sector could take up to 6 months in development & production. Since AMD's 386 was already on the market Intel was forced to use a more drastic approach: take a standard 486Dx disable the FPU and sell it as a low cost 486 - the 486Sx. The 486Sx was practically a CPU without integrated FPU - like in the previous generation. The odd thing was that the FPU was actually inside the processor but it was deactivated; the FPU was manufactured then disabled. The official 486 FPU - that was soled as a separate chip - was a full 486Dx that in fact disabled the 'main' Sx variant when it was inserted. Thus if you had a 486Sx and with a 486 FPU all you had in fact was two 486Dx processors.

Some of you might say that in this situation you could take a 486FPU and use it as a stand alone processor; theoretically this was possible but practically it was not because Intel took every possible caution to ensure that the 486FPU will be activated only when it was in a FPU socket and when a 'main' processor exists, even if the latter was completely unnecessary.

With this strategy Intel has prooved that for a whole line of processors the real cost of a single unit is aproximatively the same regardless of their Mhz speed and/or additional features like cache or integrated FPU. Yet, the cost of a processor from the same generation varies dramatically with clock speed - and possibly other factors. An example often found in practice: if the low-end version of the processor is costing 150$ the high end version could easily cost 1000$. In fact Intel has prooved that it can sell top of the line processors at the price of the low-end line and still have profits. The world was shocked! Intel is making huge amounts of money because there is no direct link between the processor's manufacturing cost and the price at which it is sold.

Additional info:

A private person from Japan sued Intel for such practices. Intel was accused of selling the same product under different names.

Pentium MMX

The marketing campaign behind Pentium MMX was - at least to say - misleading. Intel claimed that the MMX extension set (MMX = Multimedia Extensions) could significantly increase the processor's throuput in most applications. Intel urged its customers to switch from Pentium to Pentium MMX using this argument. The reality was quite different: applications that where not compiled with MMX instructions would not gain anything from MMX; what is even worst is yet to come: the software industry is driven by inertia, as a result applications that would use MMX would not appear on the market sooner that two years after the introduction of Pentium MMX.

In order to sweeten the deal for the customer Intel did indeed something that would increase the processor's performance in most applications: it increased the L1 cache from 16Ko to 32Ko; however this bigger cache couldn't push the proccessor's performance in the class that was promised. For this reason Intel did not capitalized on it at all.

Pentium III - a new generation?

Pentium III has the following advantages over his older brother Pentium II:

  1. a new instruction set called SSE; (in fact the successor of MMX)
  2. some suplimentary buffers that increase the overall performance;

The facts would suggest another name for Pentium III - Pentium II SSE, much like Pentium with MMX was called Pentium MMX. Intel deliberately created over night a new generation of processors. The real reason behind this name is that the main Intel competitor namely AMD was about to release its 7th generation of processors under the code-name Athlon. Pentium II was only a 6th generation and a Pentium II SSE would still be just that - an improoved 6th generation. The truth is that Pentium III is just an enhanced 6th generation.

Celeron - a pathetic bus: 66Mhz

Upon its creation Pentium II was a real performer: nothing from the competition could match its performance. Pentium II helped Intel gain even more market share. The competition's stronghold remained in the low-end systems because Pentium II was expensive. The first attempt to penetrate this market was made with Celeron. This first Celeron was just a full Pentium II without cache. As you might have guessed it's performance was not great because it was too dependent on the memory subsystem. Both AMD and Cyrix managed to produce processors that could do very well against this version of Celeron.

Strangely, the average user perception of Celeron was adequate and the sales were not successful. Determined by this defeat Intel released on the market an extremely powerful processor called Celeron Mendoccino; Intel managed to beat the competition in performance but at the same time it manage to beat itself.

Take a look at the following table:

  Cache size: Cache speed: Observations:
Pentium II 512 Ko half speed Built in an external cartdrige
Celeron 0 Ko - no L2 cache at all!
Mendocinno 128 Ko full speed L2 cache is on-chip


Mendocinno was cheap and in some applications outperformed the Pentium II; for most of the other applications the gained performance was not worth the price premium. As a consequence Mendocinno became a fierce competitor to Pentium II. In order to limit Mendocino's performance Intel imposed a very strict limit on its bus speed: 66Mhz. At the time of this writing this limit is only partially suspended: the fastest Celeron now enjoy an official bus of 100Mhz. Pentium(s) II or III are using a bus frequency of 133Mhz for quite a while. Pentium IV has a quad-pumped bus working at 100Mhz being equal to an effective bus of 400Mhz; despite this Celeron is still handicapated. Now you understand how powerful this solution was at the time of the launch, in fact so powerful that threaten the high line of processors from Intel; this is why Celeron is stucked with a slow bus: just to make it unattractive; Celeron is very well capable of higher bus speeds.

Intel i815

Intel i815 chipset is the succesor of the famous BX chipset. Despite its age the BX chipset remains one of the best chipsets ever made for SDRAM memory. When i815 was launched the BX was more than 2 years old. Reviews have been made with the BX running at an officially unsuported 133 Mhz FSB against i815. In the following table you can notice how the two configurations stand up one against the other:



  i815/Solano i815/Solano2 Intel 440BX
std FSB 66/100/133 66/100/133 66/100
max FSB 66/100/133+ 66/100/133+ 66/100/133+
AGP-clock at 133 MHz FSB
66 MHz 66 MHz 89 MHz
PCI-clock at 133 MHz FSB 33 MHz 33 MHz 33 MHz
Memory Clock 66 - 133+ sync/async 66 - 133+ sync/async 66 - 133 sync
AGP Transfer Rate AGP4x = 1 GB/s AGP4x = 1 GB/s 0.667 GB/s
ATA Spec ATA 33/66 ATA 33/66/100 ATA 33
No. of USB ports 2 4 2
Integrated Graphics Yes Yes No
Maximal Memory Support 512 MB PC133 512 MB PC133 1 GB PC133

Only the AGP is overclocked at 133 Mhz on the BX chipset; even with overclocking its AGP does not reach the bandwith of i815's AGP. It is clear that the i815 chipset has a theoreticall advantage over its older cibbling BX, yet it does not manage to perform better. It was very strange that a chipset released 2 years latter was not capable of doing better against the BX. Some may blame the lack of resources or ?competences? in developing i815. Many have suggested that the thruth is to be found somewhere else: politics. At the time Intel was strongly pushing a new memory technology - RDRAM that was not developed by JEDEC and which was patented by Rambus Inc. This type of memory was not royalty free; companies would have to pay royalties. At the end of the day the consumer would pay the bill.

Intel was having strong ties with Rambus plus a significant part of their stock. Intel would have only to gain if this type of memory would became the mainstream. IMO the i815 chipset was a victim of this policy. The BX chipset was already making a very hard time for RDRAM because its performance was impressive. An even better i815 would make RDRAM unjustifiable at all.

Note:

Of course, nobody was able to proove one way or the other.

NetBurst

NetBurst - a new technology for a new processor. Just for the brain dead.

With the introduction of Pentium 4 Intel has invented the term 'Netburst'; this words should actually caracterise Pentium's strengths: network accelerator (or burst)!! This name is nothing more than a fantesy created by the marketing departament for the unnkowingly John Smith how will presumably go in a store and choose a Pentium 4 machine because he is not content with the speed of his DSL connection. Actually Intel is trying to capitalize on the recent Internet development. Every Internet surfer has been at least once in his life frustrated by the speed (or the lack of...) of his connection. Intel is trying to get in his boat all those frustrated and poorly informated web surfers. Unfortunately for them their pages will not show up any faster and their downloads will not 'burst' either.

Actually Intel went pretty close to false advertising with this name.

PR rating

This is a marketing trick used by Intel's competitors in order to make up for the Mhz gap between their processors and those produced by the rival Intel.

This is not an Intel trick; it is here for completeness.

Both Cyrix and AMD have used in the past a PR (Performance Rating) system. The problem with this performance rating system is that it is based on a restrained set of benchmarks carefully picked to stress the strong points of the proceesors put to test. This is the reason for which the PR rating was completely compromised. Cyrix has done most on the 'work' in this area. Cyrix is a dead company today; just before it died it has released completely unrealistic PR ratings for its processors.

Today AMD is using a renamed PR rating for their latest processors - Athlon XP. At the same time Intel has developed the Pentium IV processor which was designed for high clock speed in the detriment of work done per cycle:

Productivity = Mhz x Work per cycle

And the end both systems: PR rating or Mhz rating are targeted at the naive customer. Before you buy a new computer it might be useful to search the web for real benchmarks in order to see how various processors are doing in your favourite application; take into consideration the price difference then decide what is best for you.



Best regards,
Razvan MIHAIU



07/03/2001 - Mohammedia, Morocco


Updated:
11/04/2003 - Bordeaux, France
- rewrote the PR rating module




Razvan Mihaiu � 2000 - 2024