@RUTGERS.ARPA:rogers%albany.csnet@csnet-relay.arpa (04/20/85)
From: "Prof. Alan Rogers" <rogers%albany.csnet@csnet-relay.arpa> A couple of weeks ago, I inquired on the net about the AT&T 3B2 and the UNIX PC. What follows is a summary of the replies I received. Thanks to all who replied. ---------- >From: Stephen Kelley <kelso@maryland.ARPA> > >I've read some good/unpleasent/indifferent articles on the net >'bout the UNIX PC - looks interesting tho I haven't got my >hands on one yet. I have been using a 3B2 for 'bout 6 mos. >SYSV C-Compiler was horrific and the doc *was* worse. SYSV.2 >(now stnd) compiler & doc are fine. (SYSV had no *more* cmd). >The only current prblm is floating point (ycch!!). It's verry >slow (lib soft modules) and it's not documented as to how >these modules get included (a -f flag on compile/load is >needed). Rumor (from ATT) has it that a fl. pt. chip is in the >works. If one comes out, than this is a pretty nice machine. >Sys Admin cmds r esp. easy thru the "menus". > > >From: Doug Gwyn (VLD/VMB) <gwyn@BRL-VLD.ARPA> > >The 3B20 is fairly old and barely competitive. >The 3B5 is a nice alternative to a VAX-11/750. >The 3B2/300 is nice but has the following drawbacks: > (1) slow floating-point > (2) limited networking support > >From: Jordan Brown <lcc.jbrown@UCLA-LOCUS.ARPA> > >I did much of the work porting dBASE II/III to the 3b2 and the PC-7300... > >3b2: Processor is a WE32000, which is Bell proprietary. SysV. $10k >minimum with .5mb ram, 10mb winchester, 2 serial ports, 1 floppy. More >like $17k with 2mb, 30mb, 6 serial, 1 parallel, 1 floppy. 3B-net is >available, I believe; this is their Ethernet software/hardware. Reasonably >fast; for one test I ran it was about 1/2 as fast as my 11/750. >Unfortunately, at the time I tested it their floating point was >*atrocious* - it was slower than an IBM PC without HW floating point. >They promised a substantial improvement, but I don't know if that ever >happened. It's a pretty standard system - a box you plug terminals >into. Quite small; maybe 3"x24"x24". One nifty thing is that the >power switch is "soft" - when you flip it it starts a normal system >shutdown, and maybe 40 seconds later turns itself off, having killed >off all the processes, and so on. All of this is based on a >pre-release unit; I don't know exactly what the released units are >like. > >PC-7300 (the "Unix PC"): 68010. SysV. $5-6k. I'm not sure what the >standard configuration is; I believe .25mb, 10mb, 1 serial, 1 parallel, >1 data phone line, 1 voice phone line, 1 floppy. The machines I've >seen had .5mb and 20mb. I believe they have some proprietary low-cost >network available, but don't know details. FAST. Very close to my >750. Faster for a case where the vax compiler generates bad code >(register shorts, I believe). Haven't tested floating point speed. >Bitmapped screen, plenty of resolution, don't remember exactly how >much. Optional mouse. Standard shell is a windowing shell with a >"simple" user interface. Kinda neat; you can run several programs >simultaneously and switch between them. Terminal output can be either >very fast (>9600 baud) or very slow (<600 baud) depending on the >intelligence of the program. This appears to be truly one of the >machines where writing a single character is almost exactly as >time-consuming as writing 80; programs which do single-character writes >are terrible, programs which do big writes run good. This machine is >really intended to run single user, but supports two in a limited >fashion; I'd regard the second terminal as a stardard Unix terminal, >w/o fancy windowing, though they claim to simulate some windowing on >it. In general, a neat machine - I've considered buying one for myself, >but have no need for another computer except as a toy. > >From A-EPSTEIN@COLUMBIA-20.ARPA Thu Apr 11 00:14:15 1985 > >The 3B2 was thoroughly flogged in a review in the November '84 issue >of Unix/World ( Vol. 1, No. 6 ). We had one in our offices a little >before the review came out and we came to the same conclusions. The >3B2 ( in its original form - I don't know if FP hardware is yet >available ) is V E R Y S L O W in floating point stuff ( we measured >about 177 FP mults/sec ). It was our opinion that the 3B2 should have >been lots faster, considering that it is a full 32 bit machine. Our >Sun 120 ( with 16 bit internal ALU ) ran rings around the 3B2. > >Of course, FP operations are only one measure of performance, but the >Sun beat the 3B2 in every benchmark - sieve, postincrement, system call, >etc. > >I would be glad to supply the benchmark codes we ran to any and all >interested parties. I sent them to ATT-Napierville but never heard >back from those guys. > >Cheers, William Anderson > >From: packard!vax135!timeinc!dwight@TOPAZ.ARPA >Organization: Time, Inc. Edit Tech, New York City > > Unix PC: > This machine was originally called the PC-7300 (the >other AT&T PC is called the 6300). They changed the name to the >Unix PC literally the day before the product was formally >unveiled last week. > It's based on a 68010 Motorola mpu. It is running >full object AT&T Unix System 5, but S5R2, not S5R2V2 (V2 added >demand paging, record locking, and lots of other neat and very >useful enhancements). AT&T has written a very comfortable >window manager (wmgr) over Unix that makes it a very friendly >environment for the non-Unixphile. The shell, of course, >is always available if you prefer it, and it appears to be >the standard Bourne shell (I did not check for Korn extensions-- >I have a feeling it wouldn't have it--though it would be great >if it did). > I'm not certain about the standard configuration, although >here's my guesses: 1 MB main memory, 2 serial asynchronous >communications ports, 1 standard Centronix-type printer i/f, >10 MB hard disk standard, 20 MB hard disk optional, 10 MHz >clock speed on the 68010, 1 320/260 KB minifloppy halfheight >drive (can read but not write PCDOS disks). I'm not sure >about MSDOS availability or partioning. > It can be connected to the Star-LAN twisted-pair >local area network, which is not yet fully available (under >beta test). > It's VERY POWERFUL for its size. It can easily handle >three concurrent users and over 32 large, intense tasks at >the same time. I was amazed with its capabilities. It is >very fast, and certainly the best microcomputer implementation >of Unix I have ever seen. I would love to own one. > It costs about $5000 for the basic machine, as described >above, and another $800 (curiously) for the operating system. >I do not believe the o.s. includes any programming tools >like compilers, SCCS, text formatters, etc.; I believe these >are available at any extra cost. So for $6000 you can get a >really nice standard Unix box for up to, say, six light users >or three to four heavy users. > > 3B2 > See the most recent (May?) issue of Unix/World, in >the performance section, for some interesting comparisons of >this machine (which is in the $15K-20K class) with others >in the same price range. The author of that article on performance >tools claims that the machine is the best performer in its class. >Response time degradation with up to 16 users is "very flat"-- >and this guy recommends the machine to his own clients. > We're seriously thinking about buying several of these >boxes for many of our requirements at Time. > Again, I don't have a spec sheet in front of me (I'm at >home--the spec sheet is at work), but I'll hazard some guesses: >1 Bellmac (WE32000) MPU, 20 MB hard disk standard, 40 MB optional. >1 minifloppy drive, streaming tape available. 4? asynch serial >ports standard. Full S5R2 Unix (object only). 3BNet networking >capability, plus StarLAN, and interfaces for the AT&T Network >are also avilable. Makes a great server for Unix PC, 6300, >IBM PCs, whatever. > Both of these boxes are worth looking into seriously. >----------------------------------------------------------------------------- > --Dwight Ernest KA2CNN \ Usenet:...vax135!timeinc!dwight > Time Inc. Edit./Prod. Tech. Grp., New York City > Voice: (212) 554-5061 \ Compuserve: 70210,523 > Telemail: DERNEST/TIMECOMDIV/TIMEINC \ MCI: DERNEST >"The opinions expressed above are those of the writer and do not necessarily > reflect the opinions of Time Incorporated." >----------------------------------------------------------------------------- > >From: Snoopy <seifert@mako> >Alan, > I worked for AT&T for 5.5 years, 3 on the 3B project, and 2.5 >on the 5ESS project, which uses the 3B20D. 3Bs are v e r y S L O W. >Very very very s-l-o-w. The 3B20 is supposed to be similar in power to >a vax. More like an 6800 having a bad day. And the 3B2 is supposed to >be ~50% of a 3B20. > >If you need super-high reliability, the 3B20D (D for duplex) has very little >downtime, since it's unlikely that both processors would break at the >same time. > >Before you make your final choice, consider our (Tektronix's) 6000 >family. The 6130 in particular. 10MHz 32016, 1MB ram, expandable to >7MB, ethernet built-in, 2 rs232 ports built-in (can add more, and >can login over the LAN), 20MB wini built-in (40 or 80MB optional, >can also have external disks), 5.25 floppy built-in, runs "Utek" >operating system, based on 4.2BSD with many system V programs ported over, >distributed filesystem and "remote" commands, virtual memory with paging, >each process has a 16MB address space available. Optional streaming >tape drive, external disks, extra rs232 ports, extra memory, mono- >or color display with mouse, IBM-PC compatible co-processor board, etc. >Base price is US$9500. The 6130 benchmarks at 60% of a VAX 780 doing >large compiles (compiles have lots of cpu and lots of disk i/o). >For more performance, there's the 6200 series, with multiple 32032s. > >If this sounds interesting, send a physical mail address and I'll >send some literature. > > > _____ > |___| the Bavarian Beagle > _|___|_ Snoopy > \_____/ tektronix!mako!seifert > \___/ >From: "Dr. Joseph M. Leonard" <jmleonar@crdc.ARPA> > We've been using a 3b2 since last summer - some thoughts: First, the >machine does NOT have hardware floating point. There is a software emulator, >and the hardware mods are due sometime in 1985 (or '86...). This is not >a problem for SOME of our applications, but... > > There is a nice disk/cart. tape package that is now available for it. >The eXtended Memory unit has (up to) a 72 Mb winchester, and a 20(?) Mb >cart. tape, that can be used in a backup capability (with individual file >recovery). This requires some hardware mods that are included in all 3b2's >being shipped, but as ours is one of the first... > > The machine works. Leaving the question of 4.2/3 vs. V, the box holds >2 Mb, and the /400 model that is rumored will hold 4 Mb. With 100-ish Mb hard >disk, either 3B2 is a nice machine for 5-10 people. Oh, yes, there is no >on line manual, but that is par for the course for a supermicro. > > The original manuals were the PITS, but the Release 2 manuals are MUCH >better. In fact, they are as nice as the VAX/VMS ones (I think that this is >good, but...). They are NOT in the Vol.1/Vol.2 format but that is not too >bad. > > Finally, the unix pc... This was initially to be released as the PC 7300. >Convergent Tech built this machine (it's a 68010), and must have needed the >money to act as an OEM. I saw the product announcement (go to one if you >get the chance - AT&T really sets a nice spread), and the machine looked good. >The trade rags say that this is a good machine, but that the OEM arraingement >might not be all that advantageous for Convergent. It looks like it is aimed >at the 2-8 user market. 2 Mb of core fills the box (in 512 Kb incs), and there >is some kind of Virtual Memory that allows up to 6(?) Mb of adddress space. I >am weak on this, though. > > I hope that this is of some help - let me know if you hear something that >contradicts my poop. Don't worry, I'm used to being corrected... > > Prices - Fed prices for the pc - in the 6.5 - 8.5 K$ range > Our 3B2 (full box) - 19 K$ > The /400 should be in the 27-32 K$ range... > > Joe Leonard >Arpanet: jmleonar@CRDC.ARPA ><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><> >Opinions expressed by my employer are their own, and have nothing to do with >those expressed here, or any others that I may have. ><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><> >____________________ >From cadtroy!schoff Sat Apr 13 18:01:32 1985 > >Your are confusing two products, the 3b2 uses a WECO chip and has been >out on the market for about a year. the unix pc from AT&T uses a 68010 >and was jointly developed with Convergent Technology, it is a much nicer >(and cheaper) machine. AT&T is giving the 3b2's away because they can't >sell them on the commercial market, (ie they are giving them to universities). > >marty >cadtroy!schoff >------------------- >From: Jordan Brown <lcc.jbrown@UCLA-LOCUS.ARPA> >The 7300 is a kinda neat machine; I keep being tempted to buy one. > >It's not terribly fast to watch - terminal I/O ranges from ~60cps for stupid >programs that do single char writes to maybe 2000cps if you write big lines. >(these are not measurements; they are guesses - I'm assuming my 750 was really >sending at 9600 baud and then comparing that with the 7300) > >I think it needs more than the base 512k - I believe it was swapping a lot. > >Their windowing shell is interesting - I wouldn't want to use it for any length >of time, but it is occasionally useful - it has a menu-driven interface to >L.sys, for one thing. > >They do not include a windowing shell for programmers (theirs is targetted >to compete with Macintoshes, which it does badly). However, I thought about it >for a couple minutes and wrote a small program (initially ~20 lines, eventually >grew to ~40-50) which would run a specified command in a window, detaching from >the original window so that your shell continues. You can then get some >semblance of job control, with virtual terminals. (In fact, I think you >*could* start multiple gettys on the console, but I know of no reason to) > >There are bitmap graphics which I know little about except that there are >raster primitives and GSS available. They have a business graphics >application. > >The mouse response is lousy, I believe because it is waking up the process >fairly often to do things. I strongly prefer to use the keyboard equivalents >rather than the mouse, and I'm usually a mouse fan. > >dBASE III, Microsoft Word, and Microsoft Basic are available for it; I don't >know what others.