jdm@boulder.Colorado.EDU (James D. Meiss) (10/17/89)
Worstation speed comparisons:
DataGeneral Aviion
(88000,16.7MHz) 17MIPs/$8K = 2.1
DecStation 3100 (MIPS, 16.7) 14MIPS/$12K = 1.2
Sun SPARC (SPARC, ?) 12MIPS/$10K = 1.2
Iris ( , )
MacIIci (68030, 16.7)
MacIIx (68030, 12) 4MIPS/$8K = 0.5
Apollo 4500 ( , ) 7MIPS/$19K = 0.4
These numbers come from 2/27/89 Electronic Engineering times
and ComputerWorld.
Can someone fill in the missing numbers? What happens
if you soup up the IIci with a cache card?
Does anyone have the AIM Tech. Workstation Benchmarks Suite
V for these machines?
Is Apple going to introduce an 88000 model?
Jim Meiss
jdm@euclid.Colorado.edunoah@Apple.COM (Noah Price) (10/17/89)
In article <12850@boulder.Colorado.EDU> jdm@boulder.Colorado.EDU (James D. Meiss) writes: > MacIIci (68030, 16.7) ^^^^-- 25 MHz > MacIIx (68030, 12) 4MIPS/$8K = 0.5 ^^-- 16 MHz > Can someone fill in the missing numbers? What happens >if you soup up the IIci with a cache card? If someone can steer me toward a universally acceptable "MIPS" determining program, I can run it on the IIci and IIci with a cache card. noah ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ noah@apple.com Mac IIci Hardware Design Team ...!{sun,decwrl}!apple!noah Apple Computer, Inc.
grunwald@foobar.colorado.edu (Dirk Grunwald) (10/17/89)
Using MIPS ratings doesn't work out very well for actual performance.
What follows is from comp.windows.x; it's the results for a heavily
scalar floating point intensive program. For scientific applications,
it'll show you the benefits better than raw MIPS ratings.
For simple interactive use, all of these machines are pretty fast &
you should really be looking at things like IO & context switches/second.
----------from comp.windows.x----------
Here is the 2nd updated list of xfroot fractal-points/processor_second
measured on various clients. The number, a count of trips/second
through the 8 line "hopalong" loop in xfroot, is a rough index of scalar
double-precision floating point uniprocessor speed. New items since the
last posting are marked with ">".
Cray X-MP 157,000 to 194,000* (*=per processor)
Cray 2 129,000 to 183,000*
> Convex C2 (gcc) 117,000 to 151,000*
> Convex C2 (vc3/fastmath) 108,000 to 138,000*
> Convex C2 (vc3) 99,000 to 118,000*
DEC DS5800 95,000 to 115,000
DEC DS5400 77,000 to 91,000
DEC DS3100 58,000 to 75,000 <<<<<<<<<<<<<<
Gould NP1 44,000 to 60,000*
DEC Vax 6400 (vcc) 50,000 to 57,000
Convex C2 (vc2) 49,000 to 55,000*
Convex C2 (cc) 41,000 to 47,000*
> Dec Vax 8650 28,000 to 33,000
Sun Sparcstation 1 ~25,000 to <<<<<<<<<<<<<<
DEC MV3900 (vcc) 22,900 to 26,100
> DG AViiON (88k 16.7 MHz) 17,200 to 24,200 <<<<<<<<<<<<<<fellman@celece.ucsd.edu (Ronald Fellman) (10/17/89)
The comparisons are VERY misleading since they didn't take into consideration number crunching or graphics applications. For example, I took the sin of a number on a Data General 88K (at 20MHz), DecStation3100, and a Sun 4/280. The DecStation was about twice the speed of either the Sun or DG machines. Many simulation programs, such as Spice, spend nearly all of their time doing double-precision floating-point computations. Thus you can probably double the price/performance ratio of the Decstation for those applications. For graphice applications, the SparcStation can be purchased with an inexpensive graphics accelerator board. This cant be done with either the Dec or DG machine. For graphics applications, the SparcStation might be best. -ron
jim@calvin.EE.CORNELL.EDU (Jim Providakes) (10/17/89)
the MIPS numbers to tell how well the computer will perform, i.e, in floating point applications the Decstation 3100 is rated at 3.5 MFLOPS (real world) which is twice the preformance of the SUN Sparcstation. I believe the Data General floating point performance is about equal to the Decstation. The floating point performace of the MAC IIci is about .2 MFLOPS (from MIPS magazine, October 89). So if you are considering perchasing a workstation for such floating poing intensive applications as CAD, Mathematica/Macsyma, data analysis programs (MATLAB,SAS,SYSTAT,etc), or modeling then you may want to think about MFLOPS/$ and not MIPS/$.
earl@wright.mips.com (Earl Killian) (10/18/89)
In article <12850@boulder.Colorado.EDU>, jdm@boulder (James D. Meiss) writes: > DataGeneral Aviion > (88000,16.7MHz) 17MIPs/$8K = 2.1 > DecStation 3100 (MIPS, 16.7) 14MIPS/$12K = 1.2 > Sun SPARC (SPARC, ?) 12MIPS/$10K = 1.2 > Iris ( , ) > MacIIci (68030, 16.7) > MacIIx (68030, 12) 4MIPS/$8K = 0.5 > Apollo 4500 ( , ) 7MIPS/$19K = 0.4 > > These numbers come from 2/27/89 Electronic Engineering times >and ComputerWorld. Several of these "MIPS ratings" are based on the dhrystone benchmark. That makes this a pretty meaningless table. To give you an idea, take the DECstation 3100, which is listed above as 14 MIPS, and the DG Aviion, which is listed as 17. Then look the machines on a set of real application programs, such as the SPEC benchmarks. The DECstation 3100 is 10.1 SPECmarks. The Aviion wasn't included in the SPEC report, but the the Motorola Delta 8608, which contains a 20MHz 88100 (i.e. same micro at a higher clock), is 7.8 SPECmarks. --
adam@mit-amt.MEDIA.MIT.EDU (Adam Glass) (10/18/89)
jdm@boulder.Colorado.EDU writes: > > Worstation speed comparisons: > > MacIIx (68030, 12) 4MIPS/$8K = 0.5 > [...] I *highly* doubt that a 12Mhz Mac IIx could even come close to 4 MIPS. How many percentage points faster is it than the stock Mac II? I had heard that the II preformed somewhere around 1 MIPS. I honestly doubt that the IIx could be 400% faster than the II, but the 1 MIPS figure could be wrong. Still, it doesn't seem likely. And shouldn't the cost be the numerator of the ratio, such that lower cost:MIPS ratios are better? I mean, it seems counterintuitave that lower numbers are worse, even if it is MIPS:cost. Oh well. Minor detail. Adam -- Adam Glass, ex-"hacker" at the Media Lab. "Something is going to happen... Email to adam@media-lab.media.mit.edu ...something wonderful."
wilkins@jarthur.Claremont.EDU (Mark Wilkins) (10/18/89)
In article <1406@calvin.EE.CORNELL.EDU> jim@calvin.spp.cornell.edu.UUCP (Jim Providakes) writes: >for such floating poing intensive applications as CAD, Mathematica/Macsyma, >data analysis programs (MATLAB,SAS,SYSTAT,etc), or modeling then you may Depends on what you are doing, of course, but most of the time a symbolic math program is not going to be spending a significant amount of time doing numerical work. Mathematica, for example, even when plotting, spends more than 85% of its time doing symbolic evaluation. In these cases, MIPS are MUCH, MUCH more important than MFLOPS. -- Mark Wilkins
bukys@cs.rochester.edu (Liudvikas Bukys) (10/19/89)
Or, try this on for size: (list prices as of last June)
list $/MIP configuration
price (specified 16MB memory, 19" mono monitor, 100MB local disk)
----- -----
15996 1333 SUN 4/60 "12 MIPS", 16M, 19" mono, 104MB scsi disk
19291 1135 DG AV300 "17 MIPS", 16M, 20" mono, 179MB scsi disk
21684 1549 DEC DS3100 "14 MIPS", 16M, 19" mono+swivel, 104MB scsi disk
23000 1642 MIPS RS2030 "14 MIPS", 16M, 17" mono, 172MB scsi disk
(The above prices may or may not be true.)
DG has been advertising low cost per MIP, but that's for a machine with
a ridiculously small memory. And you can't call up your favorite
chip merchant to populate the machine at a decent price, because they use
non-standard plug-in memory modules. (Same with DEC, though I've read that
Clearpoint has DEC-compatible 2MB boards available now.)
Also never forget to consider the other factors in total-cost-of-ownership:
maintenance contracts, software update service, etc. SOME vendors offer
attractive almost-flat fee software update service (at least for universities),
others still think that you're happy to pay 30 times the update contract cost
if you have 30 machines. SOME vendors include a license to run the software
with the machine; some vendors charge EXTRA for that privilege ("steering wheel
not included").
Also never forget the wonders of competitive bidding. Those are list prices
up there, but amazing things can happen in a free market. Take list prices
with a grain of salt. Some vendors have freedom to be creative.adam@mit-amt.MEDIA.MIT.EDU (Adam Glass) (10/20/89)
bukys@cs.rochester.edu (Liudvikas Bukys) writes: > list > price $/MIP > ----- ----- > [...] It's not "dollars per million instructions per" it's "dollars per million instructions per second" - thus, it is not $/MIP, but rather $/MIPS. It's merely a pet peeve, but I really hate it when people think MIPS is only plural. It is both plural and singular. Saying "1 MIPS" is just as correct as saying "5 MIPS." Saying "1 MIP" is simply wrong. Sorry for wasting your time. I just thought it needed to be said. Adam -- Adam Glass, ex-"hacker" at the Media Lab. "Something is going to happen... Email to adam@media-lab.media.mit.edu ...something wonderful."
earl@wright.mips.com (Earl Killian) (10/20/89)
In article <1989Oct19.153249.25491@cs.rochester.edu>, bukys@cs (Liudvikas Bukys) writes: >Or, try this on for size: (list prices as of last June) > >list $/MIP configuration >price (specified 16MB memory, 19" mono monitor, 100MB local disk) >----- ----- >15996 1333 SUN 4/60 "12 MIPS", 16M, 19" mono, 104MB scsi disk >19291 1135 DG AV300 "17 MIPS", 16M, 20" mono, 179MB scsi disk >21684 1549 DEC DS3100 "14 MIPS", 16M, 19" mono+swivel, 104MB scsi disk >23000 1642 MIPS RS2030 "14 MIPS", 16M, 17" mono, 172MB scsi disk > >(The above prices may or may not be true.) I won't argue with the prices, but I will argue with the MIPS ratings, just as in another recent posting. Those ratings are close to what the individual vendors claim, but the scales that the vendors use are not identical. In fact, the scales different vendors use differ by as much as 2x. Such differences make tables such as the above extremely misleading. For example, if you call the DG AV300 "17 MIPS", then you should call the DEC DS3100 22+ MIPS. -- UUCP: {ames,decwrl,prls,pyramid}!mips!earl USPS: MIPS Computer Systems, 930 Arques Ave, Sunnyvale CA, 94086