srp@ethz.UUCP (Scott Presnell) (07/19/87)
Hello, I would like to ask a couple of really simple, stupid questions. If you're irritated by stupid questions, maybe you better hit the 'n' key... I consider myself a pretty good programmer in a variety of languages, however when it comes to hardware, I only know a little, and that's mostly about PC/AT's and the like. So my questions are... What are the CPU's in the following computers: PDP-11/45 Vax 11/750, Vax 11/780, Vax 11/785 Vax 8600 Do they have individual names? (as in '386). Are they proprietary products of DEC? Was I correct to group them the way that I did above? I am confused because most people refer to these processors as "vax", what exactly does that mean? Furthermore what is the difference between an 11/750, 11/780 and 11/785? What are the associated coprocessors available? References to books or articles would be great. I am not interested in starting some sort of wierd argument about which might be 'better' etc. So I would suggest mailing responses rather than posting them. Thanks for your patience. Scott Presnell Organic Chemistry Swiss Federal Institute of Technology (ETH-Zentrum) CH-8092 Zurich, Switzerland. uucp:seismo!mcvax!cernvax!ethz!srp (srp@ethz.uucp); bitnet:Benner@CZHETH5A
ron@topaz.rutgers.edu (Ron Natalie) (07/21/87)
What are the CPU's in the following computers: PDP-11/45 Vax 11/750, Vax 11/780, Vax 11/785 Vax 8600 Those machines refer to specific computers made by Digital Equipment Corporation. The PDP-11/45 is a specific model of the PDP-11 processor. The CPU is commonly referred to as the PDP-11/45 CPU (although it does have some other name as well like K?-11). Like others of the PDP-11 family, it has the basic PDP-11 instruction set. It supports Kernel, Supervisor, and User mode (each with their own memory management registers) and up to 256 Kbytes of memory. It uses the DEC UNIBUS (also being the finest implementation of that bus) for I/O. It is differentiated from most others in the PDP-11 line by having SPLIT-Instruction and Data Address maps. It was near the top of the line (second only to the PDP-11/70, which supported 4 meg of memory and had the capability of using MASSBUS peripherals). This machine is very old The VAX 11 is the follow on to the PDP-11 line. VAX stands for Virtual Address Extension. Unlike their predecessors, they did paging and supported 32 bit addressing. The CPU in the 750 was about a third the speed of the 780. The 780 is usually the basic unit of compute speed (having been a common computer that had a processor speed very close to the mythical MIPS). While the 750 and the 780 used different CPU buses (the CMI and SBI respectively), they both had interfaces to connect MASSBUS's and UNIBUS's. These machines are contemporary to each other and were introduced in the late seventies. One other CPU, the 730, was also available, this was again a third as slow as the 750 and was hence pretty unusable (it supported UNIBUS only). The 782 was DEC's composition of two 780's and a shared memory controller. One CPU was the master and had all the peripherals. One could get about a 70% speed up for a lot of money. Purdue University found that there was plenty of bandwidth left on the SBI, so they removed the terminator and installed a second 780 CPU in its place for a comparable speedup for much less money than a 782. The 785 is essentially a fast CPU upgrade for the 780 CPU yielding a 70% speed-up. Everything else was the same. This was an easy way for people stuck with 780's to get some extra performance since DEC was getting behind the industry with their new machine. The 8600 is one of the new VAX products. The bus changed again (now called BI) and the box is generally much better mechanically. -Ron
chris@mimsy.UUCP (Chris Torek) (07/21/87)
In article <13413@topaz.rutgers.edu> ron@topaz.rutgers.edu (Ron Natalie) writes: >... The CPU in the 750 was about a third the speed of the 780. Ron, you should know better than that. The 780 is usually rated at 1 MIPS; the 750 is typically called about .6 or .7 MIPS, not .3. It is not *that* much slower. (But any slower is too much slower!) >... One other CPU, the 730, was also available, this was again >a third as slow as the 750 and was hence pretty unusable The 730 is about .3 VAX MIPS, I understand. >The 785 is essentially a fast CPU upgrade for the 780 CPU ... running at a nominal 1.5 VAX MIPS. >The 8600 is one of the new VAX products. Not all that new. >The bus changed again (now called BI) No, the BI machines are the 8200/8300/8500/8700/8800 series. I could never keep the details straight, but some of them are multiple CPU systems. The 8600 uses its own internal bus (the `A' bus) and has bus adapters. The only adapters available now are SBIAs, which present 780-style SBI interfaces. One can then plug in 780 Unibus adapters or Massbuss adapters or whatever. From the 4.3BSD kernel code, it is apparent that DEC could come out with new IO adapters (perhaps a BI adapter a la the 8800's DB88?). The 8600 is rated about 4 VAX MIPS; the 8650 (a `tweaked' 8600) is about 6 VAX MIPS. Rumour has it that, using properly hand-selected and tuned up boards, one can push the clock all the way up to where an 8600-series machine runs at 12 MIPS. I wonder how long it will take for DEC to come out with the 8675. . . . We have an 8250. This is supposedly rated around 1.2 or 1.3 VAX MIPS. It is hard to tell whether this `feels right', as yet: We were supposed to get Ultrix and a boot RX50 with the machine, but we did not, so I borrowed an Ultrix 2.0 field test tape and managed to boot it from a TU80. The field test kernel will not build, and the generic kernel's TCP or BVPNI Ethernet driver does not work (hard to tell which); and the machine crashes under any sort of load anyway. After much of hacking on my part, 4.3BSD finally boots, then promptly crashes (right after mapping the BUA I/O and register space ... very strange). -- In-Real-Life: Chris Torek, Univ of MD Comp Sci Dept (+1 301 454 7690) Domain: chris@mimsy.umd.edu Path: seismo!mimsy!chris
ed@mtxinu.UUCP (Ed Gould) (07/21/87)
>The 8600 is one of the new VAX products. The bus changed again (now called >BI) and the box is generally much better mechanically. The 8600 (and 8650) is not a BI machine, it's a multiple-SBI machine with its own bus (called an Abus). The BI Vaxen are the rest of the 8000 series: the 8200/8250, 8300/8350 (a dual-processor 8200), 8500/8550, 8700 and 8800. (The 8x50 machines are faster versions of the 8x00's, but are otherwise the same.) The 8200 and 8300 are simple (single-) BI machines; the 8500, 8700, and 8800 support multiple BI's. -- Ed Gould mt Xinu, 2560 Ninth St., Berkeley, CA 94710 USA {ucbvax,decvax}!mtxinu!ed +1 415 644 0146 "A man of quality is not threatened by a woman of equality."
chris@mimsy.UUCP (Chris Torek) (07/21/87)
In article <7622@mimsy.UUCP> I wrote: >We have an 8250. (Actually, we have two, but one is still in shipping boxes, and needs to stay that way until we have something that runs one the machines.) >... I borrowed an Ultrix 2.0 field test tape.... The field test >kernel will not build, and the generic kernel's TCP or BVPNI >Ethernet driver does not work I should mention that field test 2.0 is not the same as Ultrix 2.0 (good thing too!); we just happened to have one such tape handy. >... After much of hacking on my part, 4.3BSD finally boots, then >promptly crashes (right after mapping the BUA I/O and register space >... very strange). Not so strange after all. This is what I get for being careless while installing Mike's new Unibus code. My UMEMmap was 16 entries too small, so the moment I clobbered the TLB, the machine read through Usrptmap to find the kernel stack, but Usrptmap now pointed off into Unibus space. Oops. Well, now I know what those things in the 8200 machine check frame are all about. -- In-Real-Life: Chris Torek, Univ of MD Comp Sci Dept (+1 301 454 7690) Domain: chris@mimsy.umd.edu Path: seismo!mimsy!chris
stevens@hsi.UUCP (Richard Stevens) (07/22/87)
> The 8600 is rated about 4 VAX MIPS; the 8650 (a `tweaked' 8600) is > about 6 VAX MIPS. Rumour has it that, using properly hand-selected > and tuned up boards, one can push the clock all the way up to where > an 8600-series machine runs at 12 MIPS. I wonder how long it will > take for DEC to come out with the 8675. . . . One item that no one has mentioned on the net was the incredible price *increase* that DEC put into effect for the 8600 around April of this year. The base price of the system building block went from $350,000 to $447,000 !! All you got for the additional $97,000 was an Ethernet card and 1 year's warranty. Most (if not all) of the other VAX'es went down in price, but the 8600 went WAY up. I think DEC wants to get rid of these non-BI systems that easily support 4.3 BSD and third-party peripherals (that's exactly why we got an 8600 and not and 8500/8550). Richard Stevens Health Systems International, New Haven, CT ihnp4 ! hsi ! stevens
ctp@pop.utexas.edu (Clyde T. Poole) (07/22/87)
In article <13413@topaz.rutgers.edu> ron@topaz.rutgers.edu (Ron Natalie) writes: > [lots deleted] >The 8600 is one of the new VAX products. The bus changed again (now called >BI) and the box is generally much better mechanically. > >-Ron Wrong! The 8600 does NOT have a BI bus. It was the first of the 8xxx series of machines and it along with the 8650 are the only ones that do not have the BI bus. (I think :-) ) ----- Clyde T. Poole -- Technical Coordinator, Facilities and Equipment Univ. of Texas at Austin ARPA/CSnet: ctp@sally.utexas.edu Dept. of Computer Sciences UUCP: {harvard,ihnp4,seismo}!ut-sally!ctp Taylor Hall 2.124 BITNET: ctp@UTADNX Austin, TX 78712-1188 VOICE: (512) 471-9551 "Life is a bitch ... and then you die"
mangler@cit-vax.UUCP (08/01/87)
In article <463@mtxinu.UUCP>, ed@mtxinu.UUCP (Ed Gould) writes: > (The 8x50 machines are faster versions of the 8x00's, but are otherwise the > same.) According to the February 1987 "Digital Technical Journal", the 8500, 8550, 8700, and 8800 all use the same CPU (with different microcode). The 8800 is a dual 8700. I understand that the 8200/8250/8300/8350 series is based on a 7-chip implementation of the VAX-780, which was originally intended to be the the Microvax, but when it became clear how much longer it would take, they started on a one-chip subset design, and the latter project was completed first. (Does anyone know where the CVAX fits into this?) Rumor also has it that the 750 was intended as the first VAX, but DEC had so much trouble with the 750's LS-TTL gate-array design that they ended up selling the engineering prototype (the 780) as a product. Is there any truth to the legend that the designer of the 750's FPA did such a great job that it outran the 780 at floating point, and Marketing forced him to divide the clock by two? Where is the jumper? (Most 750 timings seem to be multiples of 320ns, except a few, like writes, that are multiples of 160ns - is this related?) Don Speck speck@vlsi.caltech.edu {ll-xn,rutgers,amdahl}!cit-vax!speck
rbl@nitrex.UUCP (08/03/87)
In article <3426@cit-vax.Caltech.Edu> mangler@cit-vax.Caltech.Edu (System Mangler) writes: >In article <463@mtxinu.UUCP>, ed@mtxinu.UUCP (Ed Gould) writes: >> (The 8x50 machines are faster versions of the 8x00's, but are otherwise the >> same.) > > ... > > >Rumor also has it that the 750 was intended as the first VAX, but DEC >had so much trouble with the 750's LS-TTL gate-array design that they >ended up selling the engineering prototype (the 780) as a product. > >Is there any truth to the legend that the designer of the 750's FPA >did such a great job that it outran the 780 at floating point, and >Marketing forced him to divide the clock by two? Where is the >jumper? (Most 750 timings seem to be multiples of 320ns, except a >few, like writes, that are multiples of 160ns - is this related?) > >Don Speck speck@vlsi.caltech.edu {ll-xn,rutgers,amdahl}!cit-vax!speck Stanford University Computer Forum offers up a videotaped lecture on Design Choices in the VAX-780. I'll review it as soon as I spin it up ---- sometime in the next week or so. Rob Lake