Power Mac 6100

The Power Macintosh 6100, a.k.a. the Perform 6110 series, came in two speeds: a 60 MHz version introduced on March 14, 1994 as one of the first Power Macs, and a 66 MHz version introduced on January 5, 1995. Until I benchmarked a 6100/60, 6100/66, and 6100/66 with 256 KB level 2 (L2) cache, I assumed the 6100/66 was just a slightly faster version of the 6100 - but there seem to be other differences, as well.

All three systems were similarly configured with Mac OS 8.6 on an Apple/Quantum 230 MB hard drive. Disk cache was at the default of approximately 1.2 MB. The drive was not optimized before benchmarking.

Remember that benchmarks are arbitrary. They measure certain types of performance that may or may not reflect the way you work.

Virtual memory was tested, but the difference in scores was insignificant, as much because of Apple's improvements in running virtual memory as the fact that none of these benchmark programs pushed the system to the point where it needed more memory than was installed.

Speedometer 3.06

The systems were tested on 17 October 1999 using Mac OS 8.6 with the normal complement of extensions. No third-party software was added. Computer attached to a 14" color monitor and tested in 8-bit video mode at 640 x 480 resolution. Results are relative to a Mac SE or Classic, which rates 1.0. Numbers rounded off to two decimal places. (Tests with a 256 KB L2 cache in the 6100/60 were performed on 22 Oct. 1999 with the system attached to an 832 x 624 monitor.)

These scores compare the three models tested using the default cache of about 1.2 MB.

model     CPU  graphics  disk    math
6100/60   8.15     7.5   4.46    91.1
w/256L2   8.19     7.7   3.90    92.5
6100/66   8.94    21.4   4.08   100.4
w/256L2   8.94    26.9   4.38   101.9

CPU results bear out the 10% faster processor, as do math results. Both should be unaffected by the L2 cache. The 6100/66 with the L2 cache scores about 7% higher on the disk test, which could be due to drive variance or due to the cache.

The great surprise is the difference in graphics performance between these three machines - I ran results twice to verify the difference. Based on a 10% faster design, graphics performance on the 6100/66 would score about 8.3. Instead, it scores 21.4; Apple evidently made a change in the graphics circuitry when moving the 6100 to 66 MHz.

Instead of running 10-12% faster, graphics is 187% faster - and faster yet with the L2 cache. I'd heard that the L2 cache made a great difference on the 6100, since it uses regular system memory for video. This bears it out. Graphics performance improves by 25% with the addition of a 256 KB L2 cache when running in 8-bit mode at 640 x 480. Very impressive!

However, adding the L2 cache to the 6100/60 did not result in significant improvement, increasing my suspicion that there are more changes than clock speed between the 6100/60 and 66.

Speedometer 4.02

The systems were tested on 17 October 1999 using Mac OS 8.6 with the normal complement of extensions. No third-party software was added. Computer attached to a 14" color monitor and tested in 8-bit video mode at 640 x 480 resolution. Results are relative to a Quadra 605, which rates 1.0. Numbers rounded off to one or two decimal places. (Tests with a 256 KB L2 cache in the 6100/60 were performed on 22 Oct. 1999 with the system attached to an 832 x 624 monitor.)

These scores compare the three models tested using the default cache of about 1.2 MB.

model    CPU  graphics  disk     math
6100/60  3.11   1.44     1.23   106.3
w/256L2  3.12   1.48     1.51   106.3
6100/66  3.40   1.68     1.54   116.5
w/256L2  3.43   2.14     1.73   118.4

Again, the CPU and math scores reflect the 10% faster processor. Curiously, the same 230 MB hard drive on the 6100/60 that outperformed the 6100/66 under Speedometer 3 came in behind them on Speedometer 4 - and by a significant margin. In this case, the 6100 with the L2 cache did score appreciably higher on the disk test, about 13% higher (vs. 7% with Speedometer 3).

But under Speedometer 4, we find the graphics difference between the 6100/60 and 66 is only 14% - just a bit more than you'd expect from a 10% faster CPU.

But the 256 KB L2 cache again makes an appreciable difference, in this case scoring 27% higher on the graphics test. Your mileage may vary depending on screen resolution, bit-depth, and size of the L2 cache. In fact, later tests of the 6100/60 with a 256 KB L2 cache show almost no difference in benchmark scores. It seems there's more difference between the two models than just clock speed.

Speedometer 4, 1 MB cache

I'd like to thank Matthew S. Carpenter for benchmarking his 6100/60 under Mac OS 8.1 on September 2, 2001. The first pair of results reports performance with a base system installation, and the second pair is his normal setup. The final row of numbers is repeated from our tests above to show 6100/60 performance with the more common 256 KB L2 cache. Note that our machine has a much slower hard drive than Matt Carpenter's.

model    CPU  graphics  disk     math
6100/60  3.08   1.46     2.76   105.5
w/1MBL2  3.12   1.88     2.97   107.4
normal   3.06   1.46     2.79   105.5
w/1MBL2  3.12   1.89     2.96   107.2
w/256L2  3.12   1.48     1.51   106.3

The CPU benchmark is insignificantly higher with either cache, a bit over 1%. Even the math score sees less than a 2% improvement with the 1 MB cache - and that's twice the improvement with a 256 KB cache. The disk benchmark sees a somewhat bigger improvement, about 6.5%, although you might not notice even that in day-to-day use.

Where you really notice the improvement is the graphics benchmark, which correlates with user reports that a 6100 with a 1 MB cache seems a lot snappier. Where the small 256 KB cache had improved performance less than 2%, the big 1 MB cache boosts graphics performance by 29%. That's significant.

The explanation for this improvement comes from the design of the Power Mac 6100, which uses up to 615 KB of system memory for video display. The 1 MB cache is big enough to buffer the entire display, allowing much faster access to video information and thereby speeding graphics performance significantly.

MacBench 5

These systems were tested on 17 October 1999 using Mac OS 8.6 with the normal complement of extensions. No third-party software was added. Computer attached to a 14" color monitor and tested in 8-bit video mode at 640 x 480 resolution. Results are relative to a Power Mac G3/300 (beige), which rates 1000. Numbers rounded off to two decimal places.

I also tested the 6100/60 with 256 KB L2 cache on 22 November 1999. The only other difference was a 832 x 624 monitor, which also let me run the graphics benchmark.

model       CPU  math  disk graphics
6100/60     100   161    n/a    n/a 
w/256L2     112   171    n/a    371 
6100/66     112   177    n/a    n/a 
w/256L2     135   192    n/a    n/a 

As above, the difference between CPU and math results on the 6100/60 and 66 is almost exactly what you'd expect with a 10% faster processor. Unlike the Speedometer CPU and math tests, which are relatively unaffected by the presence or absence of a level 2 cache, under MacBench 5, the L2 cache makes a measurable difference - about 20% on the CPU score and 8% on the math score on the 6100/66, but only 6-12% on the 6100/60. Yet another bit of evidence that there's more difference between these two machines than mere clock speed.

Because Mac OS 8.6 takes about 150 MB, there was insufficient room to run the disk tests. Likewise, the use of a 640 x 480 monitor prevented running the graphics test, which requires an 800 x 600 resolution.