Author: Vincent Diepeveen
Date: 06:20:25 05/19/02
Go up one level in this thread
On May 18, 2002 at 22:41:10, Jorge Pichard wrote: Note that i hope you realize that the K7 is also a RISC processor. The practical situation is that you can be very happy if the G4 is performing at 33% more than the Athlon. So that would make a 750Mhz G4 *at most* as fast as a 750 + 33% = say 1Ghz K7. You may be very lucky if that's the case. Please look to the number of instructions a clock it can do versus the K7. Then you know something about absolute limitations of the G4. If no more than 4 instructions can pass a clock, and at K7 no more than 3 can go, thereby realizing it's harder to do 4 a clock than 3 a clock, then you will realize that the average performance will be never more than 33% in favour of the g4. In reality it will be less. I hope you realize when you look to dead old RISC processors, that those are a complete joke and doing 1 instruction a clock at most. SH7000 series is a single chip which is getting outgunned by a P5 at 60Mhz which if run at 33Mhz incredible, because P5 can do 2 instructions a clock. SH7020 can do only 1. Note that P5 has a branch mispredict penalty because of a pipelin. this SH7020 has ig uess a five-stage pipeline, which means also like 5 clocks of penalty or so in case of a misprediction. SH7020 series (staring at its programming manual) have about 16 registers of 32 bits and about 3 control registers and 4 system registers (all also 32 bits). That is real good for such a processor. But take K7. It has like 44 extra general registers or so? P4 will be near that too. The real problem of all those things is they run at 20Mhz and do 1 instruction/cycle for basic instructions. Even the best programmer in the world can't make a 2500+ rated chesscomputer from that, because already years ago i was used to play against programs getting that depth. 20Mhz will be like 5 plies or so? Perhaps 7 plies if you have a real dumb PSQ program. If you play those against a program running on a K7 or P4, then it's always 100% scores for the modern hardware. This single chip is getting produced perhaps at $0.50 a piece, that's why they put them in the machines. The manufacturers are simply too cheap. Instead of putting in a decent chip, they sell it as 2500, whereas it will never manage to beat this FM at a bit serious level (even rapid i already might win easily) and i'm not 2500 rated, i have 2312 and i'm even going down quite a few points coming list. >True RISC Processing > The new Star line uses Motorola's new RISC processor technology. RISC is an >acronym for Reduced Instruction Set Chip. A simple description of what that >means is the chip uses a more efficient group of internal commands to accomplish >its task. This technology, developed by Motorola, allows processors to process >information more quickly. Those of you familiar with Macintosh know of the G4 >RISC processors, which process information faster than Pentium or Athlon >processors of the same clock speed. A G4 750mHz processor will run 2 to 3 times >faster than a 1200mHz (1.2gHz) Pentium 4, partially due to the difference >between RISC and DISC (Pentium or Athlon) processors. The processor also >generally runs much cooler, extending the lifetime of the chips themselves. The >new Novag chess computers utilize this newer processor technology, and the gains >are seen in substantially higher playing strengths, at much faster speeds. >According to Novag, the Star Sapphire, utilizing the SH7020 16mHz RISC >processor, will have and estimated ELO rating of 2530. The Star Diamond, >utilizing the SH7020 20mHz RISC processor, will have and estimated ELO rating of >2550. All the Star series computers use 1 megabyte of Program Capacity (ROM) and >256k of capacity random access memory (RAM). > > >Pricing and Availability
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.