rs4u+@ANDREW.CMU.EDU (Richard Siegel) (08/12/87)
There were also some technical inaccuracies in the benchmarks. Notice that they listed the Mac SE with HyperCharger having a 68881 clocked at 7.833 MHZ. This is not true -- the '881 in the HyperCharger is clocked at 12 MHz, and I understand GCC's going to kick it up to 16 MHz real soon. Also, they listed the plain Macintosh as having a math chip. Do these guys not know what they're talking about, or what? --Rich
hugo@gnome.cs.cmu.edu (Peter Su) (08/13/87)
BYTE is a rag, and I have no respect for the bozos that put it out.
They haven't had anytihng worth soap for years on end, and they publish
that idiot Pournelle who is still living in the 70s with his 8 inch
disks. Gawd!
Anyway, I wouldn't trust their benchamarks (or the ones in the august
issue with the Mac II vs. PS/2 model 80) as far as one of their
magazines would fly.
But, what do I know?
Pete
--
ARPA: hugo@gnome.cs.cmu.edu
UUCP: ...!{ucbvax,ihnp4,cmucspt}!hugo@gnome.cs.cmu.edu
"There are reports that many executives make their decisions by
flipping coins or by throwing darts, etc. It is also rumored that
some college professors prepare their grades on such a basis."
- Donald Knuthsteven@lakesys.UUCP (Steven Goodman) (08/13/87)
In article <AV87t=y00V4IyRA2MW@andrew.cmu.edu> rs4u+@ANDREW.CMU.EDU (Richard Siegel) writes: > >There were also some technical inaccuracies in the benchmarks. Notice >that they listed the Mac SE with HyperCharger having a 68881 clocked >at 7.833 MHZ. This is not true -- the '881 in the HyperCharger is clocked >at 12 MHz, and I understand GCC's going to kick it up to 16 MHz real soon. > >Also, they listed the plain Macintosh as having a math chip. > >Do these guys not know what they're talking about, or what? > > --Rich The 68881 was trimmed down to about 8Mhz in order to have it a valid comparison seeing as the 80287 ran at 8Mhz. But even then it was unfair to the Compaq seeing as they were using the "old" model which used this older FPU. Also Mac had cacheing where new Deskpro's has integrated. If this was a fair comparison should have had the 80387 in the new deskpro (6 times faster than 80287) and configured compaq with equal cache. Code on the Mac was compiled using a 68020 C compiler. So the Compaq was actually at a disadvantage seeing as it was not able to use their nice new 32 bit FPU where the Mac was and the Mac could speed up some of the disk action with it's cache and the Compaq was unable because they use older daekpro. I would say that if this would done you would even be more disapointed in the resaults. Must remember that Mac's have a definate disadvantage when it comes to many benchmarks, putting the CPU asside most systems offload the processes conserning gaining data from the disk, Mac's do not. Saving grace on a Mac II is you have slots and can add a card to do this off load processing. What it comes down to is Intel has FINALLY made a good chip, it DOES outperform the 68020 at equal clock speed. I am more interested in the next generation the 486 vs the 030, that should be very interesting. -- Steven Goodman Lake Systems Milwaukee, Wisconsin UUCP: {ihnp4,uwvax}!uwmcsd1!lakesys!steven
chow@monet.Berkeley.EDU (Christopher Chow) (08/13/87)
In article <170@lakesys.UUCP> steven@.UUCP (Steven Goodman) writes: >Also Mac had cacheing where new Deskpro's has integrated. >If this was a fair comparison should have had the 80387 in the new deskpro >(6 times faster than 80287) and configured compaq with equal cache. Equal cache size? What do you mean - the processor cache size is fixed as its a function of the chip design. You can't possibly mean disk cache -- it has nothing to do with floating point benchmarks. >Code on the Mac was compiled using a 68020 C compiler. So the Compaq was >actually at a disadvantage seeing as it was not able to use their nice >new 32 bit FPU where the Mac was and the Mac could speed up some of the >disk action with it's cache and the Compaq was unable because they use What makes you think the Mac code was compiled on a 68020 compiler? The article didn't mention which compilers which they used, not did they give any details to what compiler options were enabled. It was my _impression_ that they used a non-68020 compiler w/o direct 68881 calls. But I can't be sure since they didn't provide this information. > What it comes down to is Intel has FINALLY made a good chip, it >DOES outperform the 68020 at equal clock speed. I am more interested in >the next generation the 486 vs the 030, that should be very interesting. Has it really? I hope you didn't use that article as your only source of information. Anyway, there's no point to continue debating the contents of this two page article in Byte. I think we can all agree that the Byte article was not worth the paper it was printed on as it didn't give alot of information (compilers, compile options, etc). Perhaps someone could write Byte a nasty letter and tell them to perform another benchmark and report all the relevant information. Christopher Chow /---------------------------------------------------------------------------\ | Internet: chow@tcgould.tn.cornell.edu (128.84.253.35) | | Usenet: ...{uw-beaver|ihnp4|decvax|vax135}!cornell!batcomputer!chow | | Bitnet: chow@crnlthry.bitnet | | Phone: 1-415-643-2953, USPS: 2299 Piedmount Av, Berkeley CA 94720 | | Delphi: chow2 PAN: chow | \---------------------------------------------------------------------------/ Christopher Chow /---------------------------------------------------------------------------\ | Internet: chow@tcgould.tn.cornell.edu (128.84.253.35) | | Usenet: ...{uw-beaver|ihnp4|decvax|vax135}!cornell!batcomputer!chow |
twleung@picasso.mit.edu (Theodore W. (Ted) Leung) (08/14/87)
> What it comes down to is Intel has FINALLY made a good chip, it >DOES outperform the 68020 at equal clock speed. I am more interested in >the next generation the 486 vs the 030, that should be very interesting. Having just returned from Macworld Expo, I have a different perspective on this entire debate. It may be true that the 386 is a "faster chip", but faster is not necesssarily better. The key to all computers is the quality of the application software (whatever the application is). If you happen to be crunching numbers, or doing something that is really computationally intensive, then maybe you really do care if the the 386 is faster than the '020 at the same clock speed. Most people are only concerned with getting their work done as quickly as possible. For a good many people, machines "as slow as the Mac II" running software like Hypercard (no jibes about the quasi multi-tasking Multifinder), are much better tools for getting their work done, than "fast machines like the Deskpro 386" running whatever. I don't mean to demean the engineering and scientific markets, as I'm at a school where > 60% of the students are engineers. But even here, I think that many students who are using VaxStation II's running 4.3BSD with X would give them up if they could have Mac II's with the current generation of Mac software, because the applications are that much better. Fast hardware is nice, but I'd rather have second place hardware and first place software. Ted Leung MIT/Project Athena Disclaimer: My employers don't even CARE about personal computers.....
radford@calgary.UUCP (Radford Neal) (08/15/87)
In article <170@lakesys.UUCP>, steven@lakesys.UUCP (Steven Goodman) writes: > The 68881 was trimmed down to about 8Mhz in order to have it a valid > comparison seeing as the 80287 ran at 8Mhz... > What it comes down to is Intel has FINALLY made a good chip, it > DOES outperform the 68020 at equal clock speed... This is ridiculous. Do you think that users get a nice warm feeling from knowing their chip is getting "more for the megahertz" than the next guy's? Would you buy a $10000 Intel machine that ran at 8MHz in preference to a $10000 Motorola machine that ran at 24MHz, just because the Intel machine does more per cycle? Crippling one machine to make a "fair" comparison does not help answer the real question - "Which machine should I buy (for my application)?". Radford Neal
dchen@rainier.UUCP (08/15/87)
First of all, all of the benchmarking I have seen comparing the '020 to the '386 has been pretty lame. Everything (so far, I should say. Optimism in action...) is either a) biased on purpose (run by Motorola or Intel), b) biased by accident (i.e., comparing a real machine against a paper machine, or comparing radically different configurations), or c) pretty useless (millions of nops per second, mega-linpack-i'm-sorry-but-99%- of-comp.sys.mac-inverts-less-than-one-matrix-a-year-anyway, megachars- compared-to-a-null-string-per-second, etc.). Most benchmarks commit more than one of the above. I suspect most people's interest in benchmarking is not based on a need to answer the question 'how should I design my next microprocessor' (In which case bus-cycles-per-light-furlong is probably an interesting statistic) but instead on the question 'gee, I wonder how fast a '386 box feels? I wonder just how well a Mac II justifies the cost difference over a Mac +/-? I'm going to write a CPU-bound program to do <x>, which machine should I buy?' Another words, it's the application speed that matters. So I kinda sorta think that an application benchmark would be pretty interesting. You know, Word on a Mac II, Word on a Compaq '386. Excel. PageMaker (I heard somewhere that the first PageMaker release running running on a '386 was slower than on a Mac Plus. Hey, it's just a rumor, don't believe it til you see it yourself). Anyway, I think the jury is pretty far out right now. We'll start to get the results when the applications start getting ported, til then, well, I think the two machines, running with the same speed DRAMs and compiler technology, run at about the same speed. My personal bias is that Motorola machines can be made to run faster (the 25 MHz '020 beat the 20 MHz '386 to market by about a year) if you've got the bucks for a serious cache or fast memories, and that the 030 will tip the scales to Motorola's side, around the end of '88. Don't hold your breath for the '486 to come out, I hear they're trying to get it to support Ada in silicon :-) And, yes, I do think comp.sys.mac is an appropriate forum for benchmarking discussions. Applications tend to dominate in this industry, but speed is important too. Also, the software markets are beginning to converge, or at least overlap, so that we are beginning to have a choice (Make Mine Mac). David Chenevert
breck@aimt.UUCP (08/15/87)
I've watched this "debate" go on for, what, a hundred messges? It seems like forever. Of course I only "arrived" on the net a couple days ago. So for me it is effectively forever. Anyway I just broke out laughing at the hilarity of the whole issue. Byte published the results of their benchmarks. One guy (sorry, I forget his name and don't want to go looking for the message) even posted the results from an ieee benchmark. Yet still with these numbers no one can even agree which chip is "faster". One message interprets the results as indicating that the 386 is the faster chip. The next message rejoices that the 020 is faster. Sort of goes to show the real value of benchmarks. :-) By the way, none of this is in any way the opinion of AIM Technology or any of its employees or officers.... except me that is. -- Breck Beatie uunet!aimt!breck
alan@pdn.UUCP (Alan Lovejoy) (08/19/87)
In article <53@gnome.cs.cmu.edu> hugo@gnome.cs.cmu.edu (Peter Su) writes: >BYTE is a rag, and I have no respect for the bozos that put it out. > >Anyway, I wouldn't trust their benchamarks (or the ones in the august >issue with the Mac II vs. PS/2 model 80) as far as one of their >magazines would fly. > >But, what do I know? > >Pete > More than you think. Carefull consideration of the results of the two infamous articles in question will reveal various inconsistencies. I leave the details as an exercise for the reader (I sent a letter to BYTE detailing the gaffs; perhaps they'll print it). But what can you expect from people who think that systems should be benchmarked using an interpreted language (or any HLL--assembly is the only fair way to benchmark iron)? At the very least, they should use a compiled language and publish the dissasembled machine code for the benchmark (on BIX, if not in the mag itself). By the way, did anyone notice that Infoworld rated the Mac II as 3.72 times faster than an 8Mhz IBM AT (faster than the Deskpro 386, slower only than the Intel Inboard (3.78) and PC's LTD 386 (4.00))? Unfortunately, Infoworld provides almost no information as to how they generated these numbers. At least you can't PROVE that their methods are whacko. Alan "Check out those WEIRD Fibonacci results!" Lovejoy
alan@pdn.UUCP (Alan Lovejoy) (08/19/87)
In article <170@lakesys.UUCP> steven@.UUCP (Steven Goodman) writes: >The 68881 was trimmed down to about 8Mhz in order to have it a valid >comparison seeing as the 80287 ran at 8Mhz. But even then it was unfair >to the Compaq seeing as they were using the "old" model which used this >older FPU. Also Mac had cacheing where new Deskpro's has integrated. Just a minute here. Is this a benchmark of CPU chips or computer systems? On the surface, BYTE made it appear that they were benchmarking CPU chips, but the text of the article made it clear they were benchmarking specific computer systems. If they're benchmarking systems, then they should have left the chip speeds alone, and "fair" can only be defined as comparing shipped product to shipped product. If they really were benchmarking CPU chips, then the whole think is so drastically unfair to everyone that it's beyond fixing. Alan "you can't argue from both sides of an issue at the same time" Lovejoy