ylh@grpthry.UUCP ( ylh ) (09/10/86)
I posted the following message to net.micro a few days ago
but obviously this is a better place for it (Apologies for not having
followed this group before, if you think my questions have
already been answered):
Symmetric in California recently advertise in various magazines
(Byte, Unix/World) their '375' computer. It is based on
the 32016, has 2 Mb ram, 50Mb winchester, 4 serial ports,
runs 4.2BSD and is loaded with languages. All for just $4,995.
(GnuEmacs, TeX, SPICE for $10 each!).
We ran a battery of benchmark programmes and found that it
performs at the level of a microPDP11/73. Seems a little
poor for a 32016-based machine.
Does anyone have any comments? (On the machine, the company, etc).
-----------------------------------------------------------------------
Y L Huang, grouptheory systems incorporated (613) 594 0227
uucp:
{pesnta,lsuc,prcrs}!nrcaer!grpthry!ylh
{allegra,decvax,duke,floyd,ihnp4,linus}!utzoo!dciem!nrcaer!grpthry!ylh dan@rna.UUCP (Dan Ts'o) (09/11/86)
In article <174@grpthry.UUCP> ylh@grpthry.UUCP ( ylh ) writes: >I posted the following message to net.micro a few days ago >but obviously this is a better place for it (Apologies for not having >followed this group before, if you think my questions have >already been answered): > >Symmetric in California recently advertise in various magazines >(Byte, Unix/World) their '375' computer. It is based on >the 32016, has 2 Mb ram, 50Mb winchester, 4 serial ports, >runs 4.2BSD and is loaded with languages. All for just $4,995. >(GnuEmacs, TeX, SPICE for $10 each!). > >We ran a battery of benchmark programmes and found that it >performs at the level of a microPDP11/73. Seems a little >poor for a 32016-based machine. > >Does anyone have any comments? (On the machine, the company, etc). Why do you think it is poor for a 32016. That level is about what I would expect for average UNIX utilities. For 32-bit and VM stuff, the 32012 should do better. But the 11/73 is a pretty fast machine, many operations on a 11/72 are as fast or faster than an VAX 11/750, which is approximately the level of performance I'd expect from both 68010 and 32016 machines (at 8 to 10Mhz). Especially with a relatively slow disk. In addition a comparatively configured 11/73 would be about $10000, including Ultrix. The real competition, I believe, is a PC AT (at 8-12 Mhz) running Xenix, which one can also find for about $5000. A PC AT running Xenix can also perform close to a VAX 11/750. No VM, though. Cheers, Dan Ts'o Dept. Neurobiology Rockefeller Univ. 1230 York Ave. NY, NY 10021 212-570-7671 ...cmcl2!rna!dan rna!dan@cmcl2.arpa
chongo@amdahl.UUCP (Landon Curt Noll) (09/13/86)
In article <537@rna.UUCP> dan@rna.UUCP (Dan Ts'o) writes: >In article <174@grpthry.UUCP> ylh@grpthry.UUCP ( ylh ) writes: > Why do you think it is poor for a 32016. That level is about what >I would expect for average UNIX utilities. For 32-bit and VM stuff, the 32012 >should do better. But the 11/73 is a pretty fast machine, many operations on >a 11/72 are as fast or faster than an VAX 11/750, which is approximately the >level of performance I'd expect from both 68010 and 32016 machines (at 8 to >10Mhz). Especially with a relatively slow disk. > In addition a comparatively configured 11/73 would be about $10000, >including Ultrix. One thing you should look at is a Symmetric with a better disk. They sell drives with an ave. seek time of 27 ms. The 50 mb drive is the slowest drive they sell. Another thing to keep in mind is which version of the Unix you looked at. If it was an older version of 4.2, then it was likely something which came from Old-Genix land. Under Genix 4.1 the language tools were MEGA-MEGA BOGUS, SLOW and POOR code generators. It is a credit that the Symmetric did as well as it did with that stuff. Symmetric has a new set of language tools. They produce good-old BSD a.out (not Genix strangeo a.out) as well. They have, or will have soon 4.3BSD which has a number of performance improvements over 4.2BSD. Symmetric did a bunch of in-house tuneups as well. These language tools provide a good performance upgrade over the old tools. Another speed factor can be the chip set. Old NS32000 chip sets were; well "i signed the form saying I can say". The chips that are being shipped today, even though they are at the same clock speed (10 Mhz) work faster because the compiler does not have to do "funny-things"... I have heard rumors of 8mb systems, but I don't know when or how many $'s. All these factors can play into how much performance you will see out of a 375 box. > The real competition, I believe, is a PC AT (at 8-12 Mhz) running >Xenix, which one can also find for about $5000. A PC AT running Xenix can also >perform close to a VAX 11/750. No VM, though. I have seen a Symmetric out perform a PC AT/Xenix system. chongo <> /\oo/\ -- [people at Amdahl Corp never say things like this]
sean@ukma.uky.csnet (Sean Casey) (09/15/86)
In article <174@grpthry.UUCP> ylh@grpthry.UUCP ( ylh ) writes: >We ran a battery of benchmark programmes and found that it >performs at the level of a microPDP11/73. Seems a little >poor for a 32016-based machine. I'm evaluating one of the things for the company I work for. So far, all I've done is talk to a lady at Symmetric over the phone and request more info. The lady I talked to at the company said that it benchmarked anywhere from 10% to 30% faster than a Vax 750 with an FPA (both int and floating pt). I'm not familiar with the speed of the 11/73. Does her claim stand up in the light of your benchmarks? Sean -- ------------------------------------------------------------------------------- Sean Casey UUCP: cbosgd!ukma!sean CSNET: sean@uky.csnet University of Kentucky ARPA: ukma!sean@anl-mcs.arpa Lexington, Kentucky BITNET: sean@ukma.bitnet
pase@ogcvax.UUCP (Douglas M. Pase) (09/15/86)
In article <amdahl.3676> chongo@amdahl.UUCP (Landon Curt Noll) writes: >In article <537@rna.UUCP> dan@rna.UUCP (Dan Ts'o) writes: > [...] >> The real competition, I believe, is a PC AT (at 8-12 Mhz) running >>Xenix, which one can also find for about $5000. A PC AT running Xenix can also >>perform close to a VAX 11/750. No VM, though. > >I have seen a Symmetric out perform a PC AT/Xenix system. > Besides, who would ever accept a Xenix OS over 4.2BSD? 4.2BSD is MUCH nicer! -- Doug Pase -- ...ucbvax!tektronix!ogcvax!pase or pase@Oregon-Grad
dan@rna.UUCP (Dan Ts'o) (09/23/86)
In article <1092@ogcvax.UUCP> pase@ogcvax.UUCP (Douglas M. Pase) writes: >In article <amdahl.3676> chongo@amdahl.UUCP (Landon Curt Noll) writes: >>In article <537@rna.UUCP> dan@rna.UUCP (Dan Ts'o) writes: >> [...] >>> The real competition, I believe, is a PC AT (at 8-12 Mhz) running >>>Xenix, which one can also find for about $5000. A PC AT running Xenix can also >>>perform close to a VAX 11/750. No VM, though. >> >>I have seen a Symmetric out perform a PC AT/Xenix system. Can you be more specific ? I have seen a PC AT out run an 11/780. I don't mean to be a PC fanatic, but perhaps someone with access to a Symmetric could run a UNIX system benchmark suite and we could compare it to the PC AT and other machines in their price class. I have such a benchmark suite which I would gladly mail or post. This suite has been run on dozens of UNIX machines. I have trouble believing that a Symmetrics will out perform a 12Mhz PC AT substantially (more than say 20%) except in non-typical or VM benchmarks. I could believe that a PC AT would out perform a Symmetrics by as much as 30-50% on average while costing only $3000-4000. Let's run the benchmarks. BTW, the PC 386 machines will gain by 2-4X. > >Besides, who would ever accept a Xenix OS over 4.2BSD? 4.2BSD is MUCH nicer! I hear that someone is porting 4.3BSD to the PC AT and the cost will be dirt cheap. Cheers, Dan Ts'o Dept. Neurobiology Rockefeller Univ. 1230 York Ave. NY, NY 10021 212-570-7671 ...cmcl2!rna!dan rna!dan@cmcl2.arpa
jk@utastro.UUCP (John Krist) (09/23/86)
Perhaps the 12 Mhz AT can outperform the VAX 11/780 (how many users
were on it?) for small things, but let's see what it'll do when given
thing like arrays greater than 64k, which is probably THE main reason for
using a 32 bit processor over a 16 bitter. That 80286 will be choking
bits when it hits the 64k segmentations. If someone wanted to do only
small stuff, like word processing, I could see using the must cheaper
AT's, but given a 800 by 800 array to process and I will refuse to do
it on anything smaller than a 32 bitter!
John Krist
CCD Detector Lab
Astronomy Dept, U of Texas @ Austin
{allegra,ihnp4}!{noao,ut-sally}!utastro!jk
jk@astro.AS.UTEXAS.EDUylh@grpthry.UUCP ( ylh ) (10/04/86)
> In article <174@grpthry.UUCP> ylh@grpthry.UUCP ( ylh ) writes: > > > >Symmetric in California recently advertise in various magazines > >(Byte, Unix/World) their '375' computer. It is based on > >the 32016, has 2 Mb ram, 50Mb winchester, 4 serial ports, > >runs 4.2BSD and is loaded with languages. All for just $4,995. > >(GnuEmacs, TeX, SPICE for $10 each!). > > > >We ran a battery of benchmark programmes and found that it > >performs at the level of a microPDP11/73. Seems a little > >poor for a 32016-based machine. > > Why do you think it is poor for a 32016. That level is about what > I would expect for average UNIX utilities. For 32-bit and VM stuff, the 32012 > should do better. But the 11/73 is a pretty fast machine, many operations on > a 11/72 are as fast or faster than an VAX 11/750, which is approximately the > level of performance I'd expect from both 68010 and 32016 machines (at 8 to > 10Mhz). Especially with a relatively slow disk. > In addition a comparatively configured 11/73 would be about $10000, > including Ultrix. > The real competition, I believe, is a PC AT (at 8-12 Mhz) running > Xenix, which one can also find for about $5000. A PC AT running Xenix can also > perform close to a VAX 11/750. No VM, though. > > Cheers, > Dan Ts'o > Dept. Neurobiology > Rockefeller Univ. > 1230 York Ave. > NY, NY 10021 > 212-570-7671 > ...cmcl2!rna!dan > rna!dan@cmcl2.arpa My surprise came from the fact that the 11/73 generally performs well below 68000-based supermicros such as the Plexus (which, admittedly has 2 other processors for disk & i/o -- effectively a 'distributed' kernel) or Sperry and I had thought that the 32016 machines are in the same league as the 68000 machines. Anyway, the 11/73 perhaps has the best price/performance ratio. Talking about competition, Plessey's new line of 4-slot Q-bus machines start at around $5,000 (alas, you run into limitations in the 73's architecture). ----------------------------------------------------------------------- Y L Huang, grouptheory systems incorporated (613) 594 0227 uucp: {pesnta,lsuc,prcrs}!nrcaer!grpthry!ylh {allegra,decvax,duke,floyd,ihnp4,linus}!utzoo!dciem!nrcaer!grpthry!ylh