Info-IBMPC@USC-ISIB.ARPA (Info-IBMPC Digest) (01/24/86)
Info-IBMPC Digest Friday, 24 January 1986 Volume 5 : Issue 12 This Week's Editor: Billy Brackenridge Today's Topics: RT PC Public Announcement ---------------------------------------------------------------------- Date: Thursday, 23 January 1986 02:07:01 EST From: Joe.Newcomer@a.sei.cmu.edu Subject: RT PC Public Announcement I was at a presentation at IBM Pittsburgh on 22-Jan-86 in which the RT PC (note: NOT PC/RT!) was presented. The following is the contents of the "Facts Folder" (IBM G320-0909-0) which was distributed, and the notes I took. I will clearly indicate where my notes are used, since I may have errors of transcription and/or memory and thus these should not be construed as being representative of IBM. [I deleted the "facts folder" as we published an exact copy the other day. I think it is worth hearing some opinions of this new machine. I'd also like to hear from some of the Beta testers who are probably released from their non-disclosure agreements at this point. -wab] All opinions which appear below are my own personal opinions, and not those of my employer, including the U.S. government, the U.S. Air Force, or CMU. We had in attendance a hardware and a software person from Austin to answer questions from the audience. We were first given a videotape presentation which started out with lots of fast race cars (apparently to grab the attention of the watchers) then shifts to "another high performance machine, the IBM RT PC". Interviews with many key people starting with John Cocke (establishing the history including the 801 project) and going all the way to the RT PC product managers, development team, etc. Shots of VLSI layout workstations, mask fabrication, chip assembly lines, all that stuff. Only moderate puffery [but little acknowledgement of events which happened outside IBM (never mind that the window software resembles PARC's early work...)]. A slide show full of numbers, which I frantically wrote down (during the questioning period I finally said "Aren't we going to get some hardcopy of all this data?" and I therefore got one of the three fact sheets that were available; one of the local salesmen ran to get them; that is the "Facts Folder" quoted above). Fortunately the non-videotape presentation was very factual, with none of the handwaving I've had to put up with from other IBM presentations, so I gather they are getting used to technically sophisticated audiences. No wool even in evidence, let alone attempts to pull it over eyes, unlike mainframe presentations I've seen as little as two years ago. The most interesting part was the question-answer session with the people from Austin. Some of the questions were detailed IBM environment questions that were lost on me, but the askers seemed both technically sophisticated and to be satisfied with the answers. Overall, I'd rate it as a satisfactorily informative presentation, at least at the level of detail I was interested in. What was missing was enough hardcopy for the 15 or so attendees to take away, including a cover sheet saying who was actually involved; I didn't get the names of the people from Austin, although the hardware person had a name something like "Cates". I apologize both to them and to netland for not having the names. What follows, in no particular order, are my notes. Again, I state that there may be transcription errors; caveat lector. I have quoted answers only when I believe that I am giving either exact wording or close enough (within a couple words) that the people who made the statements would not disagree; if I have misquoted the responsibility is mine. I certainly cannot attribute the quotes to specific people, as both sales and technical people were responding and I didn't track any of the names. I have no future as a reporter (or perhaps I do; after reading in the local paper about the new "risk architecture" used in the joint CMU/IBM project I may have a lower bogosity coefficient here). ----------- There is no "server support" in the RT PC for a network of PCs other than that provided by the AT coprocessor. In response to a question from the audience, there is no improvement in thruput by using an RT PC as a server node. There is currently no token ring support; only PCnet support. [And presumably "no Ethernet support" is also implied by that answer. -jmn] The processor has 170ns cycle time; many instructions (although not all) execute in one cycle time. Certainly the most frequently used ones do. How fast is it? The thruput rate is about 2 MIPS. How much faster than an AT is this? Although there was much of the "caveat, apples and oranges" waffling that is essential when such questions are asked, the nominal multiplier over the PC/AT is considered to be ("conservatively") 2x to 3x. Data path is 32 bit; a 40 bit virtual memory (1 terabyte) is supported by the MMU. The expansion slots are compatible with the PC/AT bus structure, and most PC/AT peripherals will plug in directly. The CPU has approximately 50,000 devices, is horizontally microprogrammed, contains a lot of random logic. (This from the videotape presentation). The MMU is more complex than the CPU, containing about 60,000 devices. The AIX is based on release 1 of AT&T Bell Laboratories Unix System V, with some features from release 2, some Berkely extensions, some extensions from INTERACTIVE Systems, and further extensions by IBM's Austin TX laboratory. [whew! -jmn] What about software support? "There were 35 announced packages for this product when it was announced, which is the largest number of announced packages for any product at its introduction". Many of these are from third party developers. The open architecture of the PC line is maintained; device drivers may be written and configured in by the user. A sample device driver written in C is provided. The Technical Reference manual provides information on both the bus structure (PC/AT bus) and software interfaces ["BIOS" ? -jmn] When will the technical specs be released? "They come with the system when it is delivered" [Obviously the asker once dealt with AT&T. Did they ever publish the tech ref man or DOS tech ref on the 6300? I own a PC/AT because the AT&T salespeople looked at me like I had three heads when I said I wouldn't buy a machine without them]. How many of the option slots are actually available? "32-bit add-ons do not absorb option slots, i.e., the Floating Point Accelerator and additional memory do not use option slots." However they think "the disk controller uses one of them." The Professional CADAM licensed product [is apparently a port of and -jmn] is compatible with the mainframe CADAM product and can share files with it. [This was said to be a 3-D engineering design system; I am not familiar with it but several attendees asked detailed questions about it and seemed content with the replies. -jmn] The AT/370 extension card "has not been tested with this configuration". However "We have not at this time ruled out its inclusion" but "the shopping list is endless and we have limited people (and time); we have to make priority decisions based upon where we see the market" Will there be a retrofit of the RT PC to the PC/AT. "Ah, you read PC Week, don't you?" "We cannot at this time comment on any product that has not been announced." Up to 19 terminals [I think this includes the system keyboard/display] can be installed, and up to 8 users can be concurrently logged in. There are 4-port RS232C cards. Will the PC AT Coprocessor be able to take advantage of these additional ports? "The good news is that it is 100% compatible with the PC/AT. The bad news is that you have all the same limitations". What is the maximum physical memory which can be supported? "16 megabytes". What is the maximum virtual memory per process? "256 megabytes". What is the DOS shell? "A user familiar with DOS can type commands that look like DOS commands and have them executed by the AIX kernel...No, this is not to be confused with coprocessor support". Does the AT Coprocessor (80286) run virtual or real? What about memory use? "We map the AT Coprocessor addresses to the main system memory. Yes, it runs in real, not virtual mode. You can add [private] coprocessor memory." Is the current memory expansion the maximum? "No, in a year or so when memory packing technology improves you will see denser memories". When will shipments start? "In March". No, when will QUANTITY shipments start? "In March". [Some devices will not be available until September, however. I did not note which ones; memory suggests that one of the displays and the streaming tape are delayed until then. Check with your local IBM rep if you care. -jmn] How many of these computers are actually in operation? We saw in the [Pittsburgh] paper that CMU has 23 of them. "We have quite a few at vendor sites as well as our own internal development." [With respect to the CMU number] "Multiply that by 10. At least." Will the CMU Andrew software be available? No, that comes from the ACIS division, and will not be available commercially; only to academic sites. That product includes a port of 4.2bsd Unix, a C compiler, a Common Lisp interpreter, EtherNet support and Token Ring support. [Also a distributed file system. CMU CSD is having a series of presentations on the ITC/CMU developed software for the academic release -jmn]. [The videotape shown showed a user interface with "classical" overlapping windows of the PARC or MacIntosh flavor; Andrew is a tiling system. -jmn] Will the 1.2 MB diskette be compatible with the PC/AT? "It is the same drive that is used in the PC/AT". Is the 3270 emulation similar to that of the PC/AT? "It is exactly the same card." The chip is proprietary. Will it be released independent of the PC RT? [Gist of answer was "No"]. Will this chip be used in other (future) IBM products? "We would like to think so; we've invested three years of effort in this and would not like to think of it as a one-shot effort" but of course "We cannot comment on any product which has not yet been announced". Will the RISC instruction set be available to users (who may want to write compilers)? "Yes, we are maintaining the idea of the open architecture. The instruction set is documented and available". An assembler? "Yes". Are the languages compatible? [My recollection is yes, you can call procedures in C, FORTRAN or Pascal from any other language, including BASIC. This is somewhat at variance with the facts folder; if you really care, ask. -jmn] Is the Fortran-77 compatible with the Fortran-77 which operates on IBM mainframes? Gist of answer was that they are separate implementations and therefore there could be differences. [Conspicuously absent are COBOL and RPG-II. -jmn] [Lots of questions from the CADAM afficionados about the 5080 support. Significance of questions and answers lost to me. No retention.] Price? Austin: "We don't know the prices." Pittsburgh sales: "We just got the price sheets this morning and haven't had a chance to look at them in detail". Ballpark? "As little as $10K for a minimal configuration, up to $29K for a 'fully-loaded' system, exclusive of the 5080". With the 5080? "Up to $59K". Quantity discounts? "A different QPA [Quantity Purchase Agreement -jmn] than the PC line" [but apparently good, and whether good "better" or good (but) "worse" wasn't clear. Anyway, if you care, call your rep -jmn] We understand it will be offered thru "selected dealers". How can we find/become one? "This is a very sophisticated product and we wanted to insure only the very best support and service was available. Right now the number of qualified dealers is like this [holds up 10 fingers]. There must be a procedure, but we [Austin] don't know it." [several acronyms for apparently IBM divisions, arrangements, or something flew by, and these seemed to answer the questions by referring people to somewhere within IBM]. How hard is it to install and configure? "I [the software person from Austin] was the test site; my technical people brought it into my office and unpacked it...no, correction, took it out of the box. It took me 35 minutes to assemble, and that's because I made a mistake...I hadn't followed the instructions. I was able to start it running, and in 10 minutes I was doing the configuration. In another hour or so I was running it." What was the configuration? [I don't recall everything, but it involved one of the memory configurations, 40MB disk, diskette, two 4-port asynch line cards, two printers, ... -jmn] So you had it running in under 2 hours? "Yes". What about third-party devices? "We maintain the open architecture that made the PC series successful. We fully expect to see devices available for the RT PC in the same way they have come available for the PC". Does it support ---? [various people asking various devices, various answers. Important point was that certain application-specific devices, such as electrostatic printers, are supported by the application packages, e.g., CADAM]. We note the conspicuous absence of the 51[-- 53?] display. Is this not supported? [I missed the number, but it is the color display for the PC/AT which is not in the list given earlier; I think it was the 5153 but am not sure. -jmn] "Correct, it is not supported". What is the underlying mechanism of your database system? "We worked with the people from Oracle on the database and SQL interfaces". Other comments indicate that a multiuser shared relational database with locking down to the field level is provided. Nobody asked about distributed database so this point wasn't clarified. LAN configurations which included RT PC/25s RT PC/10s and PC/ATs were shown on various slides. Why is the floating point accelerator an OPTION on something called a "scientific" computer???! [great indignation on the part of the asker]. "Not everyone who does technical work or who needs a multiuser or multitasking workstation needs floating point, and there seemed to be no reason to have them pay for a feature they don't need". [I have written megabytes of code in the last 10 years and have written perhaps a half-dozen floating point operations -jmn] I didn't get the "Fact Folder" until a few minutes before I left the nearly 90-minute presentation, so I didn't get to ask what are "extended program development support services". SCCS? Programmer's Workbench? No idea. [Several complicated networking questions involving SNA, 3270 emulation, etc. whose significance and the significance of the answers was lost on me -jmn] IBM has written a book on the development of the RT PC. Architecture, RISC, etc. Sort of the collection-of-conference-papers style we have seen before for other projects. This book was made available to IBM offices with the announcement. The Pittsburgh office says they will have more than the one copy which came with the announcement in about two weeks. I got my name on the list. Book is about 10" or so square and about 5/8" or 3/4" thick. If you like that sort of thing (I do) contact IBM. I didn't ask how much it cost (well, they didn't know, and I said I didn't care, just get it for me). So that's what I have. Hope this is informative. joe ------------------------------ End of Info-IBMPC Digest ************************ -------