Clock Chipping the Daystar Turbo 601/66

By: Jeff Walther (trag@io.com)

Introduction:

The Daystar Turbo 601/66 for the IIci can be modified to run at 90MHz for approximately a 20% performance improvement in CPU related functions. Unlike other clock-chipping techniques this modification does not increase the bus speed of the Daystar Turbo 601 card. Instead, it uses the fact that all PowerPC 601 CPU's have the ability to clock double and clock triple depending on how they are configured. These instructions tell how to modify your Daystar Turbo 601/66 card so that the PowerPC 601 will triple the clock signal instead of doubling it.

The Daystar Turbo 601 uses a 33.3333MHz oscillator. The PowerPC on the Daystar Turbo 601/66 version doubles it to 66.6666MHz while the Daystar Turbo 601/100 version triples the signal to 99.9999MHz. The circuit boards and most components are identical on both cards. To convert a 66 to a 100 nine surface mount (SMT) resistors must be removed, eleven SMT resistors added, and a 44 pin integrated circuit must be added.

I found that the modified card would work for only a few minutes before over-heating. It was thus necessary to replace the 33.3333MHz clock signal with a 30MHz one. This drops the PowerPC down to 90 MHz and allows it to operate properly.

Disclaimers:

Summary:

  1. Solder an ICS 9178-02 in position U14 on the front of the card.

  2. Remove 000 (0 ohms) SMT resistors, R5, R10, R11, R12, R22, R24, R28, R48 and R65.

  3. Add 000 (0 ohms) SMT resistors to positions: R6, R8, R13, R18, R21, R23, R27, R44, R54 and R60.

  4. Add a 111 (110 ohm) SMT resistor to position R25.

  5. Test and if necessary replace the 33.3333MHz oscillator with a 30MHz one.

  6. Optional: Add some sort of fan to the portion of the heat sink that is not blocked by the power supply.

What You'll Need:

Where to get What You'll Need:


The Procedure:

It doesn't really matter in what order you modify the board. If you are a little rusty on soldering technique, I'd recommend working on the resistors before tackling the integrated circuit. Moving surface mount resistors is pretty easy and will build your confidence and improve your soldering technique.

Remember to take anti-static precautions.

  1. Remove a surface mount resistor by applying a hot soldering pencil to each side of the resistor simultaneously. The resistor will come loose and can be lifted off the board with the tips of the soldering pencils. The resistor will probably stick to one of the pencil tips so have something available that you can wipe it off onto. I used a Bounty paper towel. Save the resistors that you remove, because you will be installing them back in different positions. Sticking the resistor to a piece of cellophane tape is a good way to keep track of them.

    Remove resistors R5, R10, R11, R12, R22, R24, R28, R48 and R65. They can be hard to find. There will be some scanned in pictures of the board here that show their approximate positions in a week or so.

    Be patient and don't force the resistors off the board. They will come loose when they are hot enough. Also, for something so small, they can store a lot of heat, so be careful not to burn yourself after you've removed them.

    Now is probably a good time to take a little break.

  2. Place the resistors that you removed in step (1) in their new positions. You can solder a surface mount resistor to the board, by carefully positioning it, and then holding it down with the flat of a screwdriver or X-ACTO knife. Then apply a soldering pencil to each side of the resistor. Be careful not to gouge the circuit board.

    This technique takes a little practice. I found that applying the soldering pencil to one side of the resistor almost simultaneously with touching it with the blade worked best for me. If I pressed the resistor with the blade flat, without softening the solder on either side, the resistors had a tendency to pop out of place. Be patient, work slowly, and don't be afraid to remove the resistor and start over. Be Gentle.

    Place the resistors that you removed in step (1) in positions R6, R8, R13, R18, R21, R23, R27, R44 and R54. Place the additional 000 resistor that you obtained in position R60.

    Since these are 0 ohm resistors, you should measure 0 resistance across the pads at these ten positions after the resistors are installed.

    Place the 111 resistor that you obtained in position R25.

    After you have installed all the other components the resistance across position R25 should be about 34.1 ohms. Up until I installed the last couple of resistors (R54 or R60) the resistance across R25 with the 111 installed was about 110 ohms. I installed the 9178 first, so I don't know if the resistance across R25 will be 110 or 34.1 without the 9178.

    You may wish to use a continuity tester to ensure that all of your connections are good. I didn't find any problems when I did this part. It's really pretty easy, though you may have sore shoulders when you finish.

  3. Mounting the ICS 9178-02 is a little more nerve wracking. There are 44 pins on that chip and they are fairly closely spaced.

    First you must remove the large heat sink from the PowerPC 601. From the back of the card, pop the four silver colored clips through their holes toward the front of the card. The heat sink will now be loose. Carefully lift it up, pivoting it around the black metal pin that sticks through a hole in the card at the far end of the card from the PPC chip. When it is at about 70 degrees to the card, you can pull the pin out of the hole. Set the heat sink aside. Do not touch the white square in the middle of the heat sink or on the PPC chip. That white substance is heat sink grease and is best not disturbed.

    Next practice positioning the ICS 9178-02 chip on position U14. You want to get each pin on its own little solder pad. It takes a little work pushing it around to get all four rows of pins positioned properly. I used the tip of an X-ACTO knife to gently push it around. Some of the plastic TV tuner tools that can be bought as a set at Radio Shack might be a better choice for this task. The chip should be oriented as shown in the graphic. (as above, pictures will follow in a while) That is, pin one, the cut off corner, is in the upper left corner. The writing on the chip is oriented the same as on the PPC chip.

    Then, when you are comfortable getting the chip positioned, remove it and put a tiny dab of the silicone sealant in the exact center of position U14. Don't get any on the pads. Then gently place the 9178 on the dab of silicone as close to properly positioned as you can manage. You want to use enough silicone to stick to the chip, but not so much that it squishes out onto the pads, when you place the chip.

    Push the chip around until you have all the pins properly positioned again. The silicone is very viscous and should keep the chip from skittering around so much. Gently press the chip down, and solder one pin on each corner. The pads already have solder on them, so it should be enough to press the pin down onto the pad with a hot soldering pencil.

    Once you have one pin on each corner soldered, make sure that the chip is firmly pressed against the board, so that all the pins are touching their pads. Now take a deep breath, let it out slowly and admire your work for a few moments.

    Next you need to solder the other 40 pins. Work around the chip, soldering one or two pins on a side and then moving to the next side of the chip. This will prevent heating any one area too much. Be cautious, but don't worry too much. The ICS data book says the 9178-02 can tolerate 260 degrees C for ten seconds.

    To solder the pins, carefully press each one onto its pad with a hot soldering pencil and hold it there for a moment or two until the solder on the pad has enveloped the pin. Be careful not to touch two pins with the pencil simultaneously. You need a pretty fine tip for this. I found that the stock tip on my new soldering pencil was thin enough. The tip on my older pencil was too fat though. I occasionally slipped and did touch two pins, but it never created a short circuit. Be careful, but don't get neurotic.

    When you are done soldering, after letting the chip cool, it may provide some peace of mind to test your job with a continuity tester. I touched the end of each pad without touching the pin with one lead, and touched the other lead to the base of the pin against the chip. You should have a good connection for every pin.

    Next, I went around the chip touching adjoining pins looking for shorts. Place one lead on one pad or pin, and the other on an adjoining pin or pad. Pins 20 through 23 are common, so no you don't have a short there, or if you do, it doesn't matter. Pin one is the top pin on the left side of the chip. The pins are counted counter-clockwise from there.

  4. Take another deep breath and let it out slowly. You're done with the touchy part. Replace the heat sink on the card. You may want to add a tiny little dab of heat sink compound, if you accidentally wiped off any of the original application. Install the card in your machine and see if it boots up. You should update to the latest Turbo 601 software, 1.1, with the Turbo 601 Installer v1.1 from Daystar to use the modified card. 1.01 might work, but I'm sure the original version won't because it was released before the Turbo 601/100.

    At this point your Turbo 601 is tripling it's 33.3333 clock signal and running the PowerPC chip at 100MHz. My machine booted in this configuration, and froze up after about 2 minutes.

    This was the point where I attached the CPU fan to the heat sink. On the IIci, there is just enough room at the end of the heat sink to fit the CPU fan, so that it doesn't bump into the IIci's power supply when the card is installed. Adding the fan didn't really make any difference at 100MHz so you may want to skip this step at this point.

  5. If your Turbo 601 doesn't run reliably at 100MHz (and I'd be surprised if it does) replace the 33.3333MHz oscillator with a 30MHz one. (The turbo 601 also has a 31.3344MHz oscillator on board. Ignore it.) Either remove the 33.3333MHz oscillator with a soldering pencil and vacuum desoldering tool and solder a socket in it's place, into which you can try a 30MHz oscillator, or (much easier) use a clock chipper such as the Output Enablers RocketSocket to replace the signal with a slower one.

    Now test the turbo 601 again. At this point my Turbo 601 not only booted, but also ran reliably. The CPU fan was still attached to the heat sink, because that adhesive usually only sticks once, and I didn't want to remove it if there was a possibility that I would need it later. So I haven't tested the Turbo 601/66 at 90MHz without a fan on the heat sink.

    TattleTech 2.52 reports that I have a PowerPC Macintosh IIci and that my CPU speed is 90MHz.

  6. If your Turbo 601 doesn't boot at 90MHz, i.e. it doesn't make the reassuring chord at power on, check your work. I'd be surprised if it doesn't at least power up at 90. If it boots but freezes soon after, try installing a fan on the heat sink if you haven't already. If that doesn't work, try installing a 29MHz oscillator.

    I have tried clocking an unmodified Turbo 601/66 down with a 27MHz and 28.332MHz oscillator. It won't boot at 27. It will bong but not complete boot up at 28.332. The modified Turbo 601/66 might perform differently, but I doubt it.

    If the Turbo 601 won't work with the 29MHz oscillator, then it probably isn't going to work. It will be a pain, but reverse your work. You will probably have to clip the pins on the 9178 and remove them one at a time. This is one reason why having a socket would be nice.

    Odds and Ends and History:

    Traditional Clock Chipping:

    The thing that encouraged me to search for a way to clock chip the Turbo 601/66 is the fact that the PowerPC chip on my Turbo 601 is labeled as an 80MHz component. The die for the 66MHz and 80MHz PPC's are supposed to be the same, so if your PPC reads 66MHz there is still a good chance that this modification will work.

    Before researching this modification, I tried straight clock chipping. The Turbo 601 doesn't tolerate altering its clock speed much. The first thing I tried was replacing the 33.3333MHz signal with a 40MHz one. That should yield 80MHz. The Turbo 601 will not even boot in this configuration.

    At 38MHz, I got the start up bong once and then the machine froze. A 37MHz clock gave the same results as the 38.

    The 36MHz bonged and then after a few seconds gave an error chime. I wish I knew more about music. I think it was four rising notes. The screen came on and stopped at the E-Machine's splash screen that my Futura IISX provides. The machine then sits that way without change. I didn't let it sit more than about a minute though.

    The 35MHz bongs, and gets to the smiling mac icon. Then the screen momentarily does some narrow vertical bars and goes blank. Nothing more happens.

    Next I tried clocking the Turbo 601/66 below 33.3333MHz just as an experiment and in anticipation of developing a clock tripling modification. At 30MHz the Turbo 601 works fine. However, CPU performance decreases about 40% where you'd expect about 10%. This decrease doesn't appear with the tripling modification in place, thank goodness, otherwise after performing the modification and running the PPC at 90MHz, you'd get an overall performance decrease of about 15%.

    At 27MHz, the Turbo 601 won't boot at all. At 28.332MHz it performs much like it does at 37 and 38MHz.

    All of the above were tried on the unmodified Turbo 601/66. The only clock speeds I have tried with the tripling modification in place, are 33.3333 and 30MHz.

    The 31.3344 Clock: I don't know what the 31.3344 oscillator is for. I tried replacing it and it didn't appear to affect the board at all. I even disabled it completely and the board still worked correctly.

    Oscillator Package Sizes:

    The 33.3333MHz oscillator on my board is in a standard SG-615 package size. The Output Enablers RocketSocket fits this size oscillator. Under this oscillator there are four pads where a smaller surface mount oscillator can be mounted. There is no electrical difference in the pads used for the two sizes. Apparently Daystar put the two sets of pads on the board, so they could use whatever type of surface mount oscillator was most available at any given time. If you have the smaller oscillator, and you need to replace it, you will probably have to desolder it and solder in a standard surface mount oscillator on the other set of pads.

    Other Things to Try:

    The modified Turbo 601 should be tested at 29MHz (to see if the card will run that slowly) and at 31 and 32MHz. Additionally, some enterprising soul may wish to attach a Peltier Thermoelectric cooling element directly to the PowerPC 601 and see if it will run at 100MHz with the additional cooling.

    My Experience:

    I have been using my IIci with the 90MHz turbo 601 for about a month now without any problems, and it has been left on continuously for a full week without any problems duing that time. I have a fan attached to my heat sink as shown in the graphic (soon to be here...). It doesn't add any noise to the machine, since it is a very quiet fan.

    It's now been at 96MHz with a 32MHz oscillator for a while and it's been fine.

    Booting up goes noticeably faster with the modified card. Launching applications and many Finder functions are quite a bit zippier. Some games are much faster. Marathon in high resolution isn't any faster. In low resolution it is faster. Apparently my video system is maxed out in high-res mode.

    I am using a IIci, 32 MB RAM, a 2 GB IBM hard drive, E-Machines Futura IISX video card, 17" monitor, and System 7.5.1 with a minimum of extraneous extensions. I use Speed Doubler but not RAM Doubler.

    The Benchmarks:

    Using Ziff-Davis' MacBench 2.0 software, my turbo 601/66 scored about 26 in the CPU test and 130 in the FPU test. A stock Turbo 601/100 scores about 34 on CPU and 163 on FPU. My modified Turbo 601 at 90MHz scores about 31 on CPU and 154 on FPU. Disk and video benchmarks didn't change with the various cards. Apparently those systems are already at their maximum performance just with the 66MHz card.