Benchmarks, Finally
[My apologies for sitting on this for so long, but I wanted to get these numbers out before the new machines come out. As you can see, creating an ASCII table of so many numbers is a major chore, and for that I thank Ian Feldman. As with all benchmarks, take these with a grain of salt because as much as low-level benchmarks can point out things you wouldn’t otherwise notice and show how different systems compare, the speed that matters is the real-world speed that you experience running your programs under your precise environment. -Adam]
A couple of comments before I analyze the numbers. First, these timings were obtained on beta systems shown to me in late July, and are therefore preliminary. However, by the time the iron gets presented to the press, the hardware is pretty stable, so the figures should be within the ballpark. Second, tests were run with the screen in the black-and-white mode (1-bit) so the results could be normalized to the Classic II’s screen. Apple says that for the new systems, screen operations in color should be a tad faster than the same operations run in black-and-white, because QuickDraw has been optimized for a color screen. Fine, we’ll check that out when we have the shipping systems. Again, we have to calibrate the results to a reference Mac that has a black-and-white screen. Finally, check out the notes at the bottom of the spreadsheet; they’re important.
As the results indicate, the Mac IIvx just barely out-guns the Mac IIci. Note that the Mac IIci used in these tests doesn’t have a cache board. That’s because the IIci at BYTE was purchased long before Apple began shipping them with 5 MB of RAM and a cache board as standard equipment. It’s a safe bet that a IIci with a cache board will probably out-perform a IIvx.
The Performa 600 falls in the vicinity of the 20 MHz Mac IIsi in performance, while the Mac IIvi hangs out with the 16 MHz Mac SE/30 performance-wise. As I’ve mentioned elsewhere, the reason for the IIvx/Performa 600’s less-than-expected results is because although the CPU and FPU are clocked at 32 MHz, the bus operates only at 16 MHz. A glance at the 8-, 16-, and 32-bit memory moves of these computers compared to memory operations on the IIci confirm this finding.
Why was the slower bus used? First, to take advantage of readily available 80 ns DRAM. Second, Apple squeezed three computers out of a single design. A prototype system I examined could be a Mac IIvx, Performa 600, or Mac IIvi, depending upon the CPU board that plugged into the main logic board. I’m speculating here, but probably the only way to accomplish this easily was to lock in the bus clock at one speed. This would simplify the overall design, which in turn eliminates additional parts, and in turn lowers costs.
I had hoped the IIvx would fill the void in mid-range performance between the 25 MHz Mac IIci and 40 MHz IIfx. Instead, it looks like it’s up to the PowerBook 180 and PowerBook Duo 230 to plug that particular hole.
(The following Macintosh benchmarks table is 71 lines long and with little luck may therefore be printed on one sheet of paper, but do use a monospaced font!).
_________________ LOW-LEVEL BENCHMARKS in seconds unless noted _____
__________ CPU:____________ FPU: +-------------------- Sieve
32bit move ---------------+ | +-------------- Sort
16bit move ---------+ | | | +------- Math
8bit move ---+ | | | | | +- Sin(x)
Matrix | | | | | | | e^x
========== v v v v v v v v v
Classic II 24.2 98.7 54 41.2 36.4 45.4 189.5 97 106.1
LC II 24 92.1 52.7 39.6 37 43.3 186.4 95.3 103.4
IIvi 18.4 86.6 46.2 26.4 33.1 34.2 156 80.5 87.5
SE/30 16.4 82.2 42.1 22.8 31.3 29.8 143.6 70.6 94.5
Perf:a 600 14.1 65.5 39.3 26.2 19.6 25.1 136.6 66.5 71.7
IIsi 13.4 64.1 33.5 18.2 25.1 24.4 37.5 12.8 12.9
IIvx 8.3 52.5 26.3 13.1 15.9 14.5 30.1 9.4 9.7
IIci 10.7 51.1 26.7 14.5 19.9 19.9 29.8 9.9 10.2
PB 140 16.7 83.3 42.9 21.2 32 30.2 189.7 102.2 110.3
PB 145 11.1 52.8 27.8 15.5 20.4 20.3 135.6 70.2 76
PB 160 11.2 53.4 27.9 15.6 20.4 20.4 119.5 61.4 61.1
PB 170 11.1 52.8 27.8 15.5 20.4 20.5 35.4 11.6 11.9
PB 180 8.3 39.6 20.8 11.6 15.1 15.4 25.2 8.1 8.5
Duo 210 10.6 51.1 26.7 14.5 19.9 19.3 105.1 54.2 59
Duo 230 7.9 38.7 20.2 11 15.5 15 85.5 44.4 48
========== ----------------------------------------------------------
_____ DISK:_______________ FILE I/O: VIDEO:____________
Write -------------------+ 1 MB write +-------- TextEdit
Read --------------+ | | | +- DrawString
Seek --------+ | | | 1 MB | | Graphics
LLseek/32 blocks ---+ | | | | read | | Slow QD
LLseek/ 1 blocks | | | | | | | | | |
========== v v v v v v v v v v v
Classic II 1103 2165 0.2 18.2 9.8 4.1 4.8 5 2 47.8 0.4
LC II 974 2008 0.2 13.2 8.7 3.6 4.8 4.9 1.9 44.9 0.4
IIvi 735 1559 0.2 9.4 6.2 3.1 1.6 4.3 1.5 38.8 0.3
SE/30 921 874 0.1 22.4 9 4.6 5 4.6 2.3 26.6 0.3
Perf:a 600 720 1296 0.1 10.6 6.1 3 1.4 3.5 1.3 32.3 0.2
IIsi 883 1799 0.1 14.7 8.3 3.4 2.4 3.2 1.1 27.9 0.2
IIvx 720 1257 0.1 10.7 6.2 3 1.4 2.4 0.8 19.7 0.2
IIci 847 1551 0.1 16.5 8.9 4 3.1 3 1.3 19.6 0.2
PB 140 1086 2384 0.2 15.9 8.8 3.8 2.7 5.5 2.6 37.5 0.3
PB 145 754 1475 0.2 12.7 7.9 3.8 2.1 4.3 2.2 16.1 0.2
PB 160 1029 2157 0 19.2 6.1 3.3 2.6 3.7 1.4 24.7 0.2
PB 170 1219 2116 0.1 16.8 9.4 4 2.4 4.4 2.2 26.2 0.2
PB 180 771 1458 0.1 10.5 3.3 2.8 1.9 2.6 1.2 17.9 0.2
Duo 210 798 1591 0.1 12.6 7.5 3.5 4.4 3.2 1.3 22.8 0.2
Duo 230 676 1206 0 12.5 7.5 3.1 2.2 2.6 1.2 17.9 0.2
========== ----------------------------------------------------------
____ _________ INDEXES: +----------------------- CPU index
68020 Dhrystone --------------+ | +----------------- FPU index
68000 Dhrystone --------+ | | | +---------- Disk index
68020 Linpack --+ | | | | | +--- Video index
68000 Linpack | | | | | | | Dhrystone
========== v v v v v v v v v
Classic II 933 n/a 2000 2238 1 1 1 1 2000
LC II 883 n/a 2000 2309 1.03 1.02 1.12 1.07 2000
IIvi 678 n/a 2631 n/a 1.14 1.21 1.57 1.25 2631
SE/30 581 233 3125 3699 1.39 1.27 1.24 1.23 3125
Perf:a 600 568 n/a 3125 n/a 1.54 1.44 1.74 1.55 3125
IIsi 508 936 3571 2309 1.74 6.79 1.45 1.7 3571
IIvx 319 n/a 5000 n/a 2.26 8.91 1.81 2.31 5000
IIci 399 154 4545 5802 2.17 8.66 1.29 1.94 4545
PB 140 592 n/a 2941 3605 1.39 0.97 1.15 1.02 2941
PB 145 417 n/a 4166 5253 2.09 1.39 1.44 1.52 (missing)
PB 160 414 n/a 4545 n/a 2.08 1.63 1.56 1.67 4545
PB 170 415 155 4166 5244 2.09 7.37 1.27 1.37 4166
PB 180 309 n/a 7142 n/a 2.79 10.41 1.93 2.18 7142
Duo 210 399 n/a 5000 n/a 2.18 1.8 1.41 1.78 5000
Duo 230 304 n/a 6250 n/a 2.56 2.2 1.94 2.21 6250
========== ----------------------------------------------------------
Measurements by Tom Thompson, BYTE; ASCII table design by Ian Feldman
=====================================================================
Note 1: IIci had no cache board
Note 2: IIsi equipped with FPU
Note 3: PowerBook Duo 210 & 230 docked; had access to FPU