|Home||Non-Fiction||The Pentium Pro|
Intel's New Chip off the Old Block
Hello everybody, this is your friendly neighbourhood Cyberian coming back to the fast paced action of Cyberia after a trip to the "real" world. Boy, it's a jungle out there! So here I am back at my old, faithful computer terminal, telling you about what's hot and what's not in the land of Cyberia.
Intel has come out with another chip (actually two of 'em - the P6 and a 133MHz Pentium) and the downward slide of computer prices has already begun, but before I go into the latest tidbits about Intel and their latest brain-child, a brief look at Microsoft's Windows '95 and how it is doing.
Microsoft has finally released Win95 on the scheduled (actually rescheduled for the umpteenth time) date of August 24th but the sales of his royal highness Billy's latest OS are supposed to be quite disappointing. One vendor reportedly told that he had received a couple of thousand copies of Win95 but had barely sold about 20. There is also a rumour to the effect that a new and improved (and as usually claimed by Microsoft, bug free) version 1.1 will be released within the next few months. One interesting factor is that Sri Lanka really seems to have come a long way on the road to computerisation because there are already several final release copies of Win95 in Sri Lanka. Gone are the days when we had to wait several months (if you were lucky) to get hold of a copy of a new software package.
Another twist to the Win95 story was the banning of the OS by the Australian Navy because of the rumoured "spy" program embedded in the OS by Microsoft. The rumour is that this "spy" program would note down all the software run on that computer and upload the information to the Microsoft Network (sort of Microsoft's corner on the Internet) when the user connected to the network. Micrsoft denies the existense of such a "spy" in their OS and say that their software is merely designed to register the user on the Microsoft Network when he first logs on. Whatever the case maybe, Micrsoft, as usual, is certainly in the middle of another controversial dispute.
Incidentally, while on the subject of Windows '95, I have heard of a person here in Sri Lanka who has managed to install Win 95 on a 386DX machine with just 4MB of RAM! I am not too sure whether this is factual nor do I know how fast the response from the system is but I guess that the machine probably crawls along. Anyway, to anybody who would like to try and duplicate the feat, I wish the best of luck but me, I'd rather do without all that trouble.
Coming back to the main story of the day, Intel has started releasing sample machines containing their next-generation chip, codenamed P6, for testing purposes. They have lent one such machine to the unforgettably named Professor Thomas Nicely, who discovered the floating point bug in the Pentium chip and started the public outcry which ended in Intel losing millions of dollars in replacement costs. Intel is supposed to have challenged the Professor to find a bug in their new chip. According to Intel, they had collected a list of people who were using Pentiums for leading edge applications, when the Pentium bug (popularly known as the Fdiv bug) was reported and they had given sample P6 machines to some of the people on that list. It is obvious that Intel is determined not to make the same mistake that they once made but as to whether they will succeed, only time can tell.
Though Intel has been testing the P6 for several months now, it will not be generally available for some time to come. The most optimistic predictions say that it will start shipping in volume late this year and a name change is also said to be on the cards. The first version of the chip is to be a 133 MHz one and it is supposed to be twice as fast as a 100 MHz Pentium and up to five times quicker on certain operations due to changes in architecture which enable the new chip to process instructions much more efficiently. This enhancements to the 80x86 family is known as Dynamic Execution and is based mainly on multiple branch prediction, data flow analysis and speculative execution.
I can almost hear the sound of your chins hitting the floor as you go "Huh?" but before you start writing to the editor about meaningless gibberish being published in a computer column, let me hasten to explain. Though the terms are fancy, what they do is quite simple and not very new. In fact, the development work on the P6 started in 1990 around the same time as the work on the Pentium but the better strategies were reserved for the later chip. The whole philosophy behind the P6 is that it is to be a chip that is never idle. While a Pentium will wait for results from the prior instruction to arrive before it executes the next instruction, the P6 will keep on working and look ahead for something else that it can get on with. This is made possible by multiple branch prediction.
Branch prediction is simply guessing which instruction is likely to be carried out next. Though I say "simply guessing", it is neither simple nor is there any guessing involved. It is done through a complicated algorithm which basically does something like saying: "I recognize this instruction! Now what did I do after this the last time I executed this instruction...." and picks the same instruction if certain conditions are right. The Pentium too has the same facility but while it can look ahead by 256 branches, the P6 has a branch entry table which is twice as large and thus, is capable of producing much more accurate predictions. Though the P6 improves on the way multiple branch prediction is done, it is not new but speculative execution is.
What speculative execution does is a bit similar to playing the stock market - you know, buy shares today in the hope that they go up tomorrow! Here, the processor runs the instruction it has selected through branch prediction on the chance that it is the right one. If the prediction was correct, the P6 wouldn't be just saving time, it would have already done the work that needs to be done next! This keeps the processor ahead in the game but if it chose the wrong instruction, it has to discard the work it had done and carry out the correct instruction. This is achieved through register renaming. When the processor works ahead, is stores the results of its work in a spare set of registers and if the instruction was the correct one, the registers are renamed to reflect the correct result. Otherwise, they are ignored. This enables the processor to carry out out-of-order executions, that is, it can grab bits and pieces from here and there in the program and carry them out if they don't depend on the results from a prior calculation or instruction. This leads directly to the third part in the new, smarter P6 and this process is known as Data Flow Analysis.
Intel engineers have looked at a vast number of typical programs and have analysed the way in which most programs are executed and have come up with methods to speed up program execution in the processor but this might prove to be a stumbling block later when operating systems and programs advance beyond the point they are currently at.
With all this jazzing up, the P6 is expected to be looking ahead about 20 to 30 instructions on average for something to do and this should really get things sizzling on the P6!
Here are the complete cheats as well as the save game edit values for ID's medival DOOM II clone - Heretic. You can simply type the cheats when you are in the game to achieve the given result or, if you are of an adventurous nature (or dream of being a great hacker (but this really isn't a hack you know!)) you can use a hex editor on a saved game and change the values at the given offsets to increase your heatlh, armour etc. or to get an item you need.
The first few of the cheats given below are actually the cheats for DOOM II but in Heretic, they do the completely opposite of what they were meant to do DOOM. Here are the codes:
Offsets in SAVEGAME files in hex
60 - Health -(0-200)
Have fun with these and if any of you do write a save game editor for Heretic using the above values, just send me a copy as I would be interested in seeing what you had done.
Here are some Pentium jokes which were culled from the Internet and posted on a local bulletin board. I have collected a few that I think are the best. Let me know what you think.
1. An acronym for Pentium: