GUBBINS@RADC-TOPS20.ARPA (Gern) (05/16/85)
Okay, I propose this: Since neither the IBM-AT or Z-100 come with a math coprocessor as a standard feature (both are options), and mostly because I don't have easy access to a Z-100 that does have one, that the benchmark(s) not use the coprocessor support (8087/80287). The benchmark(s) must be the same identical code or be generated from the same compiler under PC/Z/MS-DOS. I recommend that MS-FORTRAN be used as both the IBM and Zenith packages will work on each other's systems. I do not know about the linkers. That the benchmarks do not use any disk or I/O, or a separate disk access benchmark be used (the formats/sizes/track access times/ etc, may be different and dependent only on which vendor was stuck in the production line at the time). I'm not worried, as I have just been told that the Z-100 again blows the AT away in this area also. That each machine be functioning in the fastest manufacter's supported mode. The Z-100 at 8MHz and the IBM-AT at 6MHz. I will even run mine at 6MHz if you want things to be more fair in this area. That the benchmarks do not use grahic output as you can't even come close to the maximum 640x512x8 colors supported by ZDS on the Z-100. That this challege is taken in the spirit of friendly fun, the quest for knowledge, and no cheating allowed. Any modifications? Cheers, Gern -------
GUBBINS@RADC-TOPS20.ARPA (Gern) (05/24/85)
The Zenith Z-100 (Z-110,120) is not IBM compatible (well, about 70% actually...) The Z-150/160 is the IBM compatible machine. Lotus, DBASE II must be tailored to each machine, but the compliers do not, the linkers do to a degree. The AT is slow due to IBM mucked the design, by having to use 1 memory wait state. (they put a V-8 400 engine in a volkswagon bug). This slows the machine down by about 20-30% (each memory reference takes 166 nsec longer than if they did it right. This takes their 6MHz machine down to about 5MHz throughput. A beefed up 8088 at 7.37 to 8 MHz with no wait states then has the edge with speed even though it is really an 8-bit machine (which slows it down a lot also, but apparently not as much as the wait state on the AT's 80286. Gern -------
curtis@WISC-RSCH.ARPA (05/28/85)
I repeat, Zenith's own literature advertises the Z-100 as 60% faster than the PC. This is basically because of the 8 MHz clock speed compared to the 6 MHz clock speed on the PC (we're talking the 8088 PC). The PC/AT is 250-300% faster than the 8088 PC, I still don't understand how the Zenith is faster than the AT. If it were, don't you think Zenith would be making a point out of it? Practically speaking, I would say my 2000 line PASCAL programs click away quite a bit faster on an AT than ANY 8088 machine, regardless of clock speed. A.C.
GUBBINS@RADC-TOPS20.ARPA (Gern) (05/28/85)
Here is new data on Z-100 vs. IBM-AT.
Note that the 5MHz Z-100 beats the AT 2 out of 4 benchmarks, is close in
a 3rd. A interpolated 8MHz Z-100 beats the AT by a good margin in 3 out
of 4 benchmarks. I will be able to verify the times as soon as I get
hold of the new Z-100 or a ZDS upgrade kit for mine.
25-May-85 19:46:26-EDT,2713;000000000000
Return-Path: <atm%deimos@cit-hamlet.arpa>
Received: from cit-hamlet.arpa by RADC-TOPS20.ARPA with TCP; Sat 25 May 85 19:46:08-EDT
Received: from deimos by hamlet with DECNET ; Sat, 25 May 85 16:45:17 PDT
X-ST-Return-Receipt-Requested:
Date: Sat, 25 May 85 16:44:49 PDT
From: atm%deimos@cit-hamlet.arpa
Message-Id: <850525164416.001@deimos>
Subject: Z-100 vs AT
To: INFO-HZ100@RADC-TOPS20.ARPA
Here is a bit of imput to the Z-100 vs. AT debate. I have run four benchmarks
on both machines. The Z-100 has a 5 MHz clock, but an 8087 has been added
(on a Hudson board). The AT has the standard 6 MHz clock and an 80287.
The programs are the following: The Sieve of Eratosthenes, 10 passes (see
BYTE, Jan 1983), which is entirely integer and logicaand logical operations. The
Savage benchmark is solid double-precision floating point function calls
(see Dr. Dobb's Journal, Sept. 1983 and Aug. 1984). Both of these have
been run on a very wide range of computers, from a Cray to an HP-15 !
The third is INTSUM, which does double-precision arithmetic only, no
function calls. The fourth is FOUR, which evaluates a Fourier series in
single-precision, calling SIN(A) and doing indexing.
All the benchmarks were compiled with MS-FORTRAN Version 3.2, using the
8087 mathematics library and the $NOFLOATCALLS metacommand, which is the
combination that gives the fastest code. On the Z-100 they were run under
Z-DOS Version 1.25, which is a few percent faster than MS-DOS 2.21,
since it takes less time to service the clock interrupt. Run times are in
seconds, and no allowance is made for loading time.
The results indicate that a Z-100 with standard clock doesn't quite beat an AT
on any program, but comes very close on SAVAGE, which runs almost entirely in
the 8087. (Note that the AT's 80287 runs with a 4 MHz clock.) If you multiply
the Z-100 times by 5/8 to simulate an 8 MHz clock, then the AT loses on any
program emphasizing floating point. However, it still wins on integer
operations. Compiles and links don't use floating point, so the AT seems
noticeably faster to use. The Compaq Deskpro, which has an 8086 and an 8087,
both running at 8 MHz, should beat either the 8 MHz Z-100 or the AT.
By the way, the AT can also be sped up to 8 MHz with a new clock crystal.
I think the contestants should adjourn to the bar for a friendly drink.
Program SIEVE SAVAGE INTSUM FOUR
Z-100 18.42 6.76 27.08 11.65
Z-100*5/8 11.51 4.23 16.93 7.28
AT 6.20 6.43 29.70 11.81
-------gordonl@microsoft.UUCP (Gordon Letwin) (05/30/85)
Recent articles have said that the IBM-PC-AT runs at an effective rate of 5 mHz due to wait states. This is true. They then suggest that an 8 mhz 8088 is therefore faster; this is not true. The 286 wins over the 8086 (both 16-bit buss interfaces) in a very important way - the address arithmentic is pipelined in the 286. In the 8086 instruction timings are listed as "N + EA", because the "effective address" is computed by running it through the ALU. In the 286 there is seperate EA computation hardware which computes the EA "on the way to the address pins", so to speak. Intel quotes most 286 instructions as running at "N", with EA classifications only for REGISTER/IMMEDIATE and MEMORY. I haven't done any calculations or actual measurements, but I think that a 5 mHz 286 is at least as fast as a an 8 mHz 8086 (not 8088). As for the 8088; the 286 wins over that again because of the buss size. This chip family tends to keep the memory interface nearly fully busy because of the fetch silo. in the 8088, this silo must be nearly always empty. (By the way, don't believe the 8088 timings in the books... it takes 4 cycles to fetch a byte, so if you add the # of bytes in an instruction with the number of other memory accesses it requires, and multiply by 4, you get the number of cycles that the opcode keeps the buss busy. The manual quotes a much smaller figure by assuming that the fetch-ahead silo is always full. Since there are almost no instructions that are quoted as slower than their memory fetch needs, the silo will nearly always be empty and the 4*<fetches> figure is the accurate one, not the manual figure. This is less true on the 8086 and the 286 because they fetch twice as many bytes in that 4 cycles.) Factoring in the doubled memory bandwidth of the 286, a 5 mHz 286 will considerably outrun an 8 mHz 8088. In fact, I benchmark an AT at about 2.5 to 3 times faster than a PC-XT, which means that you'd need a 12 mHz 8088 (without wait states) to keep up. As for the Z-100 measuring faster than the AT... I don't know whats going on, but I do know that you're measuring the performance of the basic interpreters. The interpreters are NOT identical. I don't know how they differ, but they are responsible if a high-speed 8088 is looking as fast as a 286. gordon letwin
ritzenth@bgsuvax.UUCP (ritzenth) (05/31/85)
I'll take that drink!