Artigos

(Adobe Reader., o qual pode ser baixado gratuitamente em: http://www.adobe.com/)

 

  How Fast is My Chess Computer's"Brain"?

By Larry Kaufman, I.M

When purchasing a chess computer, many people want to know how fast is its processor (a.k.a. "chip" or "C.P.U."). Of course this is only one factor in determining playing strength, along with RAM size, ROM size, and program quality, but it is a legitimate question. Unfortunately it is apt to be answered by quoting the Megahertz, which is only a meaningful number when comparing identical processors. A 16 MHz chip performs 16 million clock cycles per second, but different chips require widely differing numbers of cycles to execute one instruction, Therefore MHz is only useful for calculating the true speed of the processor and is not a measure of speed by itself. I think the reason that is widely used in advertising is that MHz is a clearly defined, non controversial number whereas more meaningful speed measures are not universally accepted. Nevertheless, we must try. I would like to introduce here the notion of "MaxMips". The idea for this article and much of the data come from Ron Nelson, Fidelity's hardware designer. "Mips" stands for Million of Instruc­tions Per Second, while "MaxMips" refers to the highest possible value of Mips. The point is that the Mips value for a processor depends greatly on the nature of the program. Ideally we should quote average Mips, but this is a poorly defined quantity. MaxMip should be simple to calculate, dividing the MHz by the minimum number of clock cycles per instruction, although for the more expensive processors there are complications that make this calculation a bit fuzzy. Bear in mind that some processors, may come much closer to MaxMips on average than others, so Maxchips is only a guide, but at least, unlike MHz, it is a useful number to consider.

Let's start with the 6502 chip, used in most of the models over $100.00 until recently. It takes two cycles to do anything, so a 5 MHz model like Fidelity Chesster, Saitek Corona, or Mephisto Milano would have a maximum speed of 2.5 Mips. A 6 MHz model like Novag Super Expert C or Fidelity Designer 2100 display would do 3 MaxMips. A Polgar 10 MHz would do 5. Fidelity´s "Little Chesster", which uses a "single chip" version of the 6502, does 1 MaxMip. For the h-8 chip, the situation is similar - divide by 2. The 10 MHz chip used in Fidelity Travel Master and Saitek Blitz and Prisma does 5 MaxMips. Actually it is considered to be slightly slower than the 6502 on average, but not enough to worry about. Thus it can be seen that it is only the limited memory in these models that prevents them from achieving master status.

A popular chip for models in the $100 ballpark is the single chip 6301. It takes 4 cycles to accomplish anything, so the MHz must be divided by 4. Thus, the 8 MHz Novag Primo, Mephisto Marco Polo, and Fidelity "USCF Academy" do 2 MaxMips, the Novag Super VI.P. (9.8 MHz) does 2.45, and the Novag Beluga and Super Novag (each 16 MHz) do 4. These numbers somewhat overstate the true speed of the chip, the difference between average and MaxMips being unusually large in this case. As with the H-8, the small memory in some of these models limits the strength more than the processor speed.

Some examples of MaxMips in low priced models are: Fidelity EI Dorado 0.4, fidelity Microchess 0.5, Novag Uno 0.53, and Fidelity Marauder 1 MaxMip. They use a variety of single chip processors. Now we come to the Motorola 68000 family. For the 68000, no less than 5 cycles are required per instruction. so a 12 MHz machine like the Mephisto 16 Bit Lyon or Vancouver does only 2.4 MaxMips, about the same as the much cheaper 8 bit models like Chesster, Corona, and Milano. Its superiority is due to using 16 bit instruction, to its larger ROM, to superior programming, and especially to its vastly greater RAM, which permits hash tables. Even the 16 MHz Fidelity Elite (version 2 or the new Premiere) and Fidelity Mach III only do 3.2 MaxMips, only marginally more than the Fidelity Designer 2100 Display, but again the larger memory and the 16 bit instructions make the difference. As for the 68020, 3 cycles is the minimum, so the Mephisto Vancouver 32 bit 12 MHz does 4 MaxMips and the Fidelity Designer 2325 at 20 MHz does 6.67. For the 68030 and 68040 the situation is rather complicated, but the effective minimum seems to be 2.5 and 1.25 cydes respectively, 50 the Mephisto Vancouver 68030 36 MHz does 14.4 MaxMips and the Fidelity Version 10 (68040/25 MHz) does 20. Finally, for RISC processors the minimum is normally 1 cycle per instruction, so the "ChessMachine" at 15 Mhz should be rated at 15 MaxMips, though in practice I don't believe it is really as fast as the 68030 at 36 MHz, perhaps because RISC chips sometimes require more than one instruction to do what a CISC (Le. standard) chip can do in one. When comparing 8 bit and 16 bit chips, it is very difficult to say how much the ability to use 16 bit instructions is worth, but I think you won't be too far off the mark if you multiply the MaxMips of 16 and 32 bit chips by 1.5 when comparing to 8 b Thus, the Elite Premiere and the Travel Master processors are of about equal power, with the 200+ point strength difference being attributable primarily to the huge difference in RAM and ROM, since each has the world's best program for its hardware.

 


Inside or outside, wonderful computer - Elite


 

  "Construindo seu livro de aberturas - dicas e ferramentas"          livrodeaberturas.pdf 


 

1990 - Torneio de máquinas x máquinas na Escola Jogo Aberto                        Torneiobrilde90.pdf


Vendas no ebay - pequeno negócio?

salesinebay.pdf


 

A problemática do rating nos jogos de xadrez dedicados.

 

Na edição de Jogo Aberto de 1988, um dos poucos exemplares que consegui resgatar (pioneira revista de xadrez cujo diretor era o MF Dirk Dagobert van Riemsdijk), dizíamos que um dos problemas que afetavam a análise e comparação dos níveis dos diversos modelos de computadores de xadrez prendia-se ao seu rating.

 

Cada fabricante procurava vender seu produto da melhor maneira, e como o rating na época era em tese um dado objetivo a respeito da força enxadrística, o mesmo era usado para impressionar os compradores.Todavia, outras informações serviam mais para confundir do que para esclarecer. Por exemplo: qualificar uma máquina de expert poderia significar apenas que ela teria muitos níveis de jogo, e não que sua força seria equivalente a de um grande mestre.

 

Entretanto, foi inegável a contribuição dos programadores como Julio Kaplan, David Kittinger, Richard Lang, Frans Morsch, Ron Nelson, Ed Schröder dentre outros.

 

Este site tem por objetivo resgatar a memória do que representou os computadores de xadrez na década de 80 e 90, e porque continuam a exercer um facínio imenso entre os vários colecionadores do mundo.