[Editor's note: Many thanks to Tom Thompson and BYTE Magazine for this, and, we hope, future articles. Tom and BYTE have provided us with this information because of our speedy distribution and because BYTE has limited space for Macintosh coverage. Tom feels that disseminating the otherwise wasted information through TidBITS is an excellent way to share it with the Macintosh community in a timely fashion. We agree, and hope everyone finds BYTE's tests, which would be impossible for us to duplicate, useful. -Adam]
"How fast is that Performa 600?" I've heard this question a lot recently, now that the Performa line has been out for about a month, and the prices of the Mac IIsi and IIci have fallen. Some time ago, I ran BYTE's low-level Mac benchmarks on a prototype Performa 600. These low-level tests exercise various computer subsystems (processor/memory, floating-point, disk, and video) to gauge their performance. We won't have definitive results until we run BYTE's application test suite on a shipping Performa 600 (we're waiting on a loaner from Apple), but these preliminary low-level test results do provide a rough performance estimate. The tests occasionally pin-point substantial system design changes, and measure their impact on the Mac's performance. For example, when a PowerBook 170's floating-point processing at 25 MHz easily bested a 40 MHz Mac IIfx, it didn't take long for Apple's engineers to point out that the System 7.0.1's new Omega SANE routines caused the performance boost.
So where does the Performa 600 stack up? Here are some preliminary results:
BYTE low-level test results Mac IIci Performa 600 IIsi SE/30 CPU Matrix 10.7 14.1 13.4 16.4 8-bit move 51.1 65.5 64.1 82.2 16-bit move 26.7 39.3 33.5 42.1 32-bit move 14.5 26.2 18.2 22.8 Sieve 19.9 19.6 25.1 31.3 Sort 19.9 25.1 24.4 29.8 FPU Math 29.8 136.6 37.5 143.6 Sin(x) 9.9 66.5 12.8 70.6 e^x 10.2 71.7 12.9 94.5 Video TextEdit 3.0 3.5 3.2 4.6 DrawString 1.3 1.3 1.1 2.3 Slow Graphics 19.6 32.3 27.9 26.6 QuickDraw graphics 0.4 0.2 0.2 0.2 Mac IIci Performa 600 IIsi SE/30 Indexes CPU Index 2.17 1.54 1.74 1.39 FPU Index 8.66 1.44 6.79 1.27 Disk Index 1.29 1.74 1.45 1.24 Video Index 1.94 1.55 1.70 1.23
All values are in seconds, unless noted. For each test index, a Macintosh Classic II = 1, and higher values indicate better performance. Disk I/O benchmarks removed for brevity. All systems ran System 7.0.1, except the Performa 600, which ran a beta version of 7.0.1P. Note that (a) the IIci had no cache board and (b) the IIsi was equipped with an FPU.
Discussion -- As you can see in the CPU test suite, the Performa 600 with its 32 MHz 68030 doesn't come close to the 25 MHz 68030-based IIci, and in some instances the 20 MHz IIsi does slightly better. Why is this? Take a closer look at the memory move tests, which measure how fast data can be moved about in RAM. The times fall somewhere between the IIsi's and the 16 MHz SE/30's results. Apple explains that although the Performa 600's CPU is clocked at 32 MHz, the main bus only operates at 16 MHz. This impacts performance, since the CPU must wait for reads and writes to memory to complete. Final performance is hard to gauge here, since the 68030 has two 256-byte on-chip caches. For example, look at the Sieve times, where the benchmark code fits into the 68030 caches. This is the only test where the Performa 600 outpaced the IIci.
The slower bus doesn't make the Performa 600 a slouch in other areas, however. The Performa 600, without an FPU, performed faster floating-point processing than an SE/30, even though the latter has a 16 MHz 68882 FPU. (Remember that the Performa 600 has an FPU socket.) Video timing results were mixed, again due to whether or not the benchmark code fit into the 68030 caches. Looking at all of the indexes, the Performa 600 appears to fall in the Mac IIsi range of processing power. Again, we'll know more when we run application benchmarks on a shipping system.
Even if the Performa 600 does no better than a IIsi in terms of processing power, the computer certainly has other advantages. It has three NuBus slots (versus the IIsi's one adapter slot), a beefier power supply (112 W vs. 47 W), a 5.25" bay for a CD-ROM drive or other SCSI peripheral, and built-in video that supports 16-bit pixels on a 13" monitor - something not even the IIci can do.
Tom Thompson -- firstname.lastname@example.org