Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP Posting-Version: version B 2.10.2 9/18/84; site callan.UUCP Path: utzoo!watmath!clyde!burl!ulysses!allegra!mit-eddie!genrad!decvax!decwrl!greipa!pesnta!pertec!scgvaxd!wlbr!callan!tim From: tim@callan.UUCP (Tim Smith) Newsgroups: net.micro.68k Subject: RE: (Intel disbelief) and net.micro.best Message-ID: <1029@callan.UUCP> Date: Mon, 17-Jun-85 15:36:45 EDT Article-I.D.: callan.1029 Posted: Mon Jun 17 15:36:45 1985 Date-Received: Thu, 20-Jun-85 11:34:09 EDT References: <155@soph.UUCP> <> <157@soph.UUCP> <594@intelca.UUCP> <446@oakhill.UUCP> <1557@amdcad.UUCP> <411@enmasse.UUCP> Organization: Callan Data Systems, Westlake Village, CA Lines: 101 > POINT 2 -- WHY CAN'T ANYONE READ??? Let's make a deal; I will learn to read if you learn to write. > > Tim Smith, <1012@callan.UUCP> writes: > > > > [ He is talking about 1K block transfers ] > > > NO I WAS NOT !!!! See below, this is the thing that makes > the numbers interesting in the first place!!! > > > > My numbers, using the fastest block move instructions possible: > > > Intel family processors: > > > 8088, 4.77 MHz 130 copies/sec > > > 8086, 8 MHz 260 copies/sec > > > 80286, 6 MHz 450 copies/sec (real mode) > > > Motorola ones: > > > 68000, 12 MHz 340 copies/sec (SLOW block move loop) [ & waitstates ] > > > 68010, 10 MHz 425 copies/sec > > ...... > > I tried this on my 10Mhz 68010 with no wait states. > > ..... > > 68010, 10Mhz ~1600 copies/sec ( word aligned buffers ) > > ..... > > Are you SURE you don't have wait states? > > If you think I was comparing block copies you miss the WHOLE point > of the numbers, the only thing that makes them interesting. > They do SIGNIFICANT PROTOCOL PROCESSING between the block copies, > including APPLICATIONS CODE also. I'd never seen what I > consider a representative comparison of these processors on > IDENTICAL SOURCE CODE, so I posted the results. Why didn't you say so? > > Yes, block copy times dominate the calculation (typical of > how many applications???) ... no, they are not block copy Now you say that block copy times dominate the calculation. Then my original reply is still valid - My 68010 at 10Mhz no wait states seems to be faster than yours. ( for those who didn't see my results, worst case - assumes no alignment on buffer - was 650 copies/sec ). > benchmarks, which I consider pretty boring. The reason I > posted those numbers in the first place was because they > showed something MORE than either vendor's propaganda, and they > matched my actual impressions using the various machines. > Why didn't you say so? Here is the article I am being flamed at for quoting: (enter sarcasm mode) > I agree with Dan -- I do NOT believe those Intel benchmarks. Anyone who's > seen the 68000/8086 Benchmark Wars between Motorola and Intel has to be > skeptical about any 68020/80286 figures put out by Intel. I was quite > surprised at an informal benchmark I came up with last week on 1 Kb core ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ > copies on a set of machines, namely that the 286 processor came out tad ^^^^^^^^^^^^^^^^^^^^^^^^^^^ Well, excuuuuuuusssssee me! How ever could I have gotten the impression you were doing a benchmark on 1Kb core copies? > better than a 68010. There was only a little additional code (Ethernet ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ > protocol processing) going on between the copies. ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ Forgive me if I got the impression that only a little additional code was going on between copies. How silly of me not to have noticed that "SIGNIFICANT PROTOCOL PROCESSING" and "APPLICATIONS CODE" was being done between each copy. (exit sarcasm mode) > > My numbers, using the fastest block move instructions possible: > Intel family processors: > 8088, 4.77 MHz 130 copies/sec > 8086, 8 MHz 260 copies/sec > 80286, 6 MHz 450 copies/sec (real mode) > Motorola ones: > 68000, 12 MHz 340 copies/sec (SLOW block move loop) > 68010, 10 MHz 425 copies/sec > > My impression is that the 68010 and 80286 are roughly the same overall. > > Disclaimer: this was NOT an overall benchmark, etc., so no flames please. > No wait states except on the 68000, so far as I know. Yes, this gives > the 80286 an advantage since it then has the fastest memory system. (?) Look, using a simple dbra loop with a movb a0@+,a1@+ I get 650 copies a second for arbitrarily aligned buffers, and you get 425. You have some other code running that I don't. It seems to me that either your other code is taking up a significant amount of the time, or you have wait states in your memory. I was simply curious as to which it is. -- Tim Smith ihnp4!{cithep,wlbr!callan}!tim