Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP Path: utzoo!mnetor!uunet!seismo!rutgers!ames!elroy!jplgodo!wlbr!etn-rad!jru From: jru@etn-rad.UUCP (John Unekis) Newsgroups: comp.sys.ibm.pc Subject: Re: Intel Microprocessors Message-ID: <242@etn-rad.UUCP> Date: Fri, 7-Aug-87 16:14:45 EDT Article-I.D.: etn-rad.242 Posted: Fri Aug 7 16:14:45 1987 Date-Received: Sun, 9-Aug-87 08:24:12 EDT References: <1112@lznv.ATT.COM> <399@aucs.UUCP> <3225@cucca.columbia.edu> <234@etn-rad.UUCP> <789@unccvax.UUCP> Reply-To: jru@etn-rad.UUCP (0000-John Unekis) Organization: Eaton Inc. IMSD, Westlake Village, CA Lines: 122 In article <789@unccvax.UUCP> cbenda@unccvax.UUCP (carl m benda) writes: >In article <234@etn-rad.UUCP>, jru@etn-rad.UUCP (John Unekis) writes: >> In article <880@bdmrrr.bdm.com> davis@bdmrrr.bdm.com (Arthur Davis x4675) writes: >> make use of their opportunity to escape from the Intel tar pit >> and use the MC68020 to make the PS/2 into a REAL computer? >> > >Well John, can you say memory management? Why is it that a 4 meg Mac >can't run a real multitasking operating system? It has the wonderous >68000 microprocessor. Awww guess what? it can't manage its own memory. >Segmentation allows the Intel chips to manage their memory. I suggest >you put a MAC with 4 meg next to an AT with 4 meg of memory running >Microport system V or SCO Xenix and place your bets as to which machine >can start compiling 4 different programs simultaneously. You see John, This type of sleazy logic is typical of braindamaged intel vicitms, I picked a MAC for my comparison with a PC specifically because they are both single-tasking machines. If you want to compare a UNIX based 68xxx machine with a UNIX based 80x86, try a SUN 3 , which makes a UNIX based AT look like the toy it really is. >not even the 'powerful' 68020 can run UNIX without the aid of a WHOLE >different chip to aid it namely the 68851 MMU. Only ONE chip? To run UNIX on any processor requires DOZENS of support chips. There is the chip set to support the state machine for the BUS protocol, the (optional) floating point processor, the memory itself, etc. etc. ... OH, but the 80386 can run UNIX all by itself can it? When was the last time you had the cover off your computer? Or maybe you did know these things and just forgot to mention them (more intel sleaze). >Just ask computer engineers >which chip set they'd rather design a computer with. I have worked with dozens of computer engineers, and none of them that I know of consider Intel to be a serious contender in the super micro arena. After going for years without ANY intel processor with a 32 bit instruction set, they finally cough out the 80386 and guess what... it has errors in the chip mask that cause it to generate random bit errors when it does 32 bit math. WAY TO GO INTEL ! This is almost as confidence inspiring as their 82586 ethernet chip (which is probably up to rev Z by now). The MAC II isn't >even a real computer because it can not run a real multi-tasking OS. >at least not without and extra 600.00 chip (68851) If we are discussing Intel vs. Motorola, then the limits of the MAC II are pretty much irrelevant. There are dozens of vendors that supply myriad systems using the 68xxx class processor running multitasking operating systems. There are Motorla, Plexus, Heurikon, Ironics, Sony, Masscomp, Stratus, ...(too many to list here), running operating systems like Versados,Pdos,OS9,UNIX V, UNIX BSD4.3, VOS, etc. How many vendors are there for 80386 systems? Well theres Intel, and IBM, and Intel, and IBM(clones don't count). And multitasking operating systems? Well there's UNIX, and UNIX, and UNIX. What about real-time applications? The 68xxx has Pdos, OS9, and Versados. The 80386 ... do I hear crickets chirping? And on the subject of memory management- the whole point of an MMU is to allow each task to run in its own memory space independently, this allows the task to behave as if it owned all the memory in the system, without being aware that other tasks are there. It also provides security by keeping tasks from trouncing each other's memory. In the 68020 world , you load the MMU with the start of a memory area in user space, (lets say address 0) then where the memory starts in real address space, and how big it is. From then on the program runs as if it had the whole memory (up to 4 gigabytes) to itself. No segment changes are ever needed. And for security no user state process can modify the MMU, so that no user process can ever physically modify another's memory. In the Intel scheme, you first have to decide whether you are running in protected mode or not. If not, then you are constantly having to juggle your segment registers to be able to address memory across 64 K boundaries. There is no protection to keep a user task from waltzing into anothers memory whatsoever. If you decide to go into protected mode, then you must first set up a global segment table in supervisor state to tell you where to find the segment tables for each task. Then within the context of the global table, you set up a local segment table for each task. This does not free you from segmentation, however. You must still juggle the segment registers, only now they must be translated as an index into a segment table. The segment register contains an index into the table which gets you a segment descriptor, which contains the base address of a segment which is added to the remainder of the segment register to get an actual segment base address which is then added to the offset register to get to real memory. Piece of shi... sorry, I meant piece of cake. The Intel memory management is to Motorola Memory Management what the three stooges are to the Bolshoi ballet. > >One final note, the 386 has a 1meg segment limit. What a limit. The application that we are dealing with at my facility is digital image processing. Typical images are 2048 by 2048 pixel elements, where each pixel is 8 bits deep. That's four megabytes for each image. And some operations require us to combine two images. Trying to do this on an Intel processor is like running with your shoelaces tied together. >What size segmentation would you like? > I would like no segments at all, A flat linear address space that is practically infinite. Which is what I get with Motorola. REMEMBER CAMPERS - When you buy Motorola , you get a 32 bit micro with a mainframe hidden inside. When you buy Intel, you get an 8 -bit 8080 chip disguised as a 32 bit microprocessor. --------------------------------------------------------------- The opinions above are practically perfect in every way. Naturally they belong to me. ihnp4!wlbr!etn-rad!jru