Q: Some Macs can be "chipped" to a higher speed. Is this a safe option? Is it possible to chip the Quadra 605?
A: A little over a year ago, I picked up a Quadra 605 (a.k.a. LC 475, Performa 475/476) at an auction for $7.50. It came with 36 MB of RAM, which alone made it worth the price, but I decided to do a little experimenting. I'm often asked if clock-chipping is a viable performance enhancement, and I was curious to try it out myself and see if I would really see the speedup others had claimed.
Marc Schrier's Clock Chipping Home Page is a must-read for anyone interested in clock-chipping a Mac. I'll take this opportunity to reiterate a warning Marc gives several times -- if you aren't totally comfortable with using a soldering iron on your Mac's motherboard, either do not do the modification or get someone to do it who is comfortable. Neither I nor this web host take any responsibility for any damage you may do to your computer as a result of any information contained herein or linked in this article.
I visited Marc's page and read up on the modification for the LC 475 and Quadra 605 (there are several links on Marc's page). It turned out I needed a couple of low-wattage grounded soldering irons -- I only had a 40W soldering gun that was far too imprecise to do the work, so I went to Radio Shack and picked up a pair for $8 each. After performing the 10-minute modification (most of which was soldering iron warm-up time), the 25 MHz LC040 chip worked just fine at 33 MHz -- but I wanted a bit more performance. For $20, I picked up a 33 MHz 68040 chip, heatsink, and clip from a guy on LEM-Swap. After I installed those, I ran Speedometer 4.02 on the Q605/33 (running System 7.1.3):
With the 33 MHz increase (a roughly 33% increase in clock speed), the '040 benchmarks roughly as expected — about 33% faster in non-FPU-related tasks. The inclusion of the FPU in the full '040 bumps up FPU-related tasks immensely -- assuming a 33% increase based on clock speed alone, the addition of the FPU improves the Whetstone benchmark by nearly 400%.
Discounting the FPU-intensive Whetstone benchmark, the average on the Q605/33 is 1.325, right around the expected 33% increase over the 1.007 average of the Q605LC/25. The FPU increase is a factor of about 5.4, accounting for the CPU speed increase.
Since the clock-chipping modification bumps up the whole motherboard to 33MHz (and not just the CPU itself), across-the-board improvements ought to be apparent. They are. Here are the graphics numbers (the stock 605 was running 7.5.5, which shouldn't affect performance at all):
This is anywhere from a 35 to 50% boost in benchmarked graphics speed. In real-world usage, graphics are noticeably faster and smoother to animate on screen. Anecdotal reports from some users on Quadlist say that graphics display on the Web improves markedly with the addition of the full '040 to an LC-based Quadra, and the FPU may indeed account for this improvement, since it's apparent clock speed isn't the only factor working here.
The FPU benchmarks, which are relative to a Quadra 650, a 33 MHz full '040, show a noticeable improvement, too. (Speedometer will not run FPU benchmarks on any machine without a hardware FPU.)
For all practical purposes, this Q605/33 is equivalent in CPU and FPU performance to a stock Q650.
One last test: the Performace Rating. Five iterations each on CPU, Graphics, and Math (the Disk tests were skipped so that variations in HDs wouldn't affect the test).
A 31% improvement in CPU speed, 42% increase in graphics speed, and a whopping 366% improvement in FPU speed (not a big surprise due to the addition of the hardware FPU) over a stock Q605 -- not bad numbers for a 15-minute modification to a $7.50 computer.