The most complex part of any adjustable CPU card is the myriad of possible speed settings (bus speeds, ratios, CPU speeds and backside cache speeds). This is where most owners have problems with their cards if they get too impatient, are not careful, or assume that high speeds will be reliable. Here are some very general tips for speed settings:

1. Isolate one factor at a time when testing for reliable speeds.
2. Start with a low bus speed - 45MHz or less.
3. Keep the Bus/CPU ratio low so that you are not overclocking the CPU during testing for bus speeds.
4. Keep Cache speeds below rated speeds. Most ZIFs have 2:1 rated cache (1/2 the CPU speed)
5. Run tests overnight (game demo, RAM test) before you assume any speed is reliable
6. Don't push the ragged edge or you lose any tolerance for temperature or voltage variations. Speed without relibility is nothing but frustration.
7. Don't overclock the Cache. Cache speed is not a primary factor of real world performance. Despite what a cache utility may tell you, I suggest keeping the cache at 1/2 the CPU speed or within its rated speed (whichever is less).
8. When you reach a limit of bus speed, etc. - I back down at least two settings (1-2 MHz in bus speed) to ensure I have some 'headroom' left for reliability. (I.E. as temperatures rise, etc.) Remember you won't 'see' even a 10% difference in CPU speeds, but it might be a lot more reliable in heavy use and long term.

The photo below shows the CarrierZIF speed settings switch which are used to select bus speeds (in all cases) and Bus to CPU ratios (for ZIFs that don't have jumpers or switches on them for ratio settings). If your ZIF CPU module has jumpers or a rotary switch then refer to the manual for your specific ZIF module for how to adjust Bus/CPU ratios, as these ZIFs will ignore the CarrierZIF's switches 9-12 (bus/cpu ratio) setting.

As shown in the photo, switches 1 - 8 are used to select the main bus speed (see the CarrierZIF for a table of settings from 40MHz to 60MHz). The table will show which switches to set (flip down) for your desired bus speed. In the latest version of the manual the table is on page 72, but this may change if the manual is revised but just look in the table of contents for the 'MVP CPU Frequency Chart'.

This MVP Frequency Chart table also has the settings for bus/CPU ratio (switches 9-12 above). If your ZIF module does not have jumpers, then use the CarrierZIF's switches 9-12 to set the bus to CPU ratio which determines CPU operating speed (bus speed times bus/cpu ratio = CPU speed). Think of the bus/cpu ratio as a multiplier of the bus speed that determines the clock speed of the CPU. The CarrierZIF manual has a table of switch 9-12 settings that correspond to ratios from 4.5x to 8x in .5x increments. (Later models show the 10X ratio setting but that only works if you have a 10X G3 CPU module installed).

Tips: The first step I perform in any review of an adjustable CPU card like this is to find a reliable bus speed. I start with a low bus speed and CPU ratio. When you're trying to determine reliable bus speeds keep bus/CPU ratios low so that you have no chance of running the CPU at too high a speed. You need to isolate one factor at a time and therefore keeping the CPU at or below rated speed is a must. Ditto for the cache speed - keep it well below rated speed to make sure you're not overclocking the cache. Otherwise you'll not know what is the reason for any stability problems. Again I want to isolate bus speed only in this phase.

I run a battery of applications tests (see my reviews) at each speed. If these tests show no errors I increase bus speed by 5MHz initially and repeat the tests. If problems are seen I decrease the bus speed until the system is 100% reliable. After seeing a stability problem I run the mac overnight at the last bus speed that passed the apps tests reliably. For a production environment I often back off even that speed a bit as its foolish to run on the "razor's edge" of stability since there would be no margin left for increased room temperatures, voltage variations, etc.

Once I've determined the bus speed that is reliable in the test system I then try to select a bus/CPU ratio that results in the closest speed to the CPU's ratings. A simple way to do this is to divide the bus speed by the rated CPU speed - and then chose the closest bus/cpu ratio to that number.

Example: The card is set to 50MHz bus speed and the CPU is rated at 400MHz.

400Mhz CPU divided by 50MHz bus = 8

So I'd set the ratio to 8x. Often the math isn't that 'clean' - say your bus speed is 46MHz and the CPU rated speed is 350MHz.

350/46 = 7.0608xxxx.

I'd round to the nearest ratio that didn't overclock the card (7X). Remember bus/cpu ratios are in .5X steps, I.E. - 4.5, 5, 5.5, 6, 6.5, 7, 7.5, 8 and 10X (if you have a 10X G3 CPU module). BTW: Even with a 10X CPU there are no 8.5-9.5X ratios.

In some cases it may be better to lower bus speed and use a higher ratio to be able to run the CPU nearer the rated speed. And remember that unless you have a 10X G3 CPU module (still rare as of this article date), you'll need to have a Mac that can run a 50MHz bus speed to be able to run a 400MHz CPU speed due to the 8X maximum ratio in most G3 ZIFs on the market now.

Guide to Using the Speed Settings Table: The image below is from the CarrierZIF manual, modified to show the logical steps in determining speed switch settings.

1. Choose a Bus speed from the far left column in the table. Verify your switch settings on the card agree with those shown in the table for that speed.

2. Read across the table from that bus speed to find the nearest CPU speed to your ZIF module's rating.

3. Read up to the top of the column with that CPU speed to determine the correct ratio settings.

Bus Speed Tips: As noted in the troubleshooting sections of my reviews and FAQ, one aid to higher bus speeds is to remove any L2 cache dimms on the motherboard (a must in many cases) and/or deinterleaving RAM. I personally remove any L2 cache dimm from any Mac before adding a G3 CPU card, especially in cases of Apple Macs which often have very slow cache dimms installed. If the L2 cache is less than 1MB it is of no real benefit anyway and only serves to add another potential for instability.

Backside Cache Speeds: Unlike bus speed and bus/CPU ratio which use hardware switches, Backside Cache speed is controlled by the XLR8 software control panel. See the Software Controls page of my CarrierZIF review for more details. The following is a screenshot of the 'Advanced controls' tab of the XLR8 Cache control (v1.3.2 shown). Note the cache speed in this sample is set to 'automatic' (2:1 ratio in this case) but I could select 'manual' and adjust cache speeds from a list in the dropdown menu.

Most all ZIFs on the market now have cache rated for 1/2 the CPU speed (often listed as xxxMHz w/2:1 cache or for instance 400/200). I normally leave the cache setting at 2:1 (CPU/cache) unless I experiment with a CPU speed that would result in overclocking the cache if run at half the CPU speed. For example: during fmax testing if I'm running a 400MHz 2:1 cache upgrade at 440MHz CPU speed then a 2:1 cache setting means the cache is running at 220MHz - which may not be reliable for a 400/200 rated module.

I don't recommend overclocking your backside cache despite what some utilities may report as a possible reliable speed based on startup testing. In some cases these simple software tests will suggest that speeds far higher than the cache rating may be reliable (I saw one case where it reported 260+MHz tested ok, but in actual use at that speed it locked up within minutes). Cache speed is not a major factor in real world performance. Overclocking the cache can lead to stability problems and it's not worth squeezing out the last drop of cache speed at the expense of reliability.

About CPU Overclocking: Of course all the above warnings also apply to overclocking the CPU. It's risky and may lead to data loss and premature hardware failure. Although I usually test for 'fmax' (freq max) speeds with review samples and have run many macs overclocked, I cannot recommend the practice to others for obvious reasons. According to the entries in the Rate Your G3 Upgrade database the practice is fairly common for those willing to run the risk and exercise care in selecting speeds they have proven as reliable (at least in the short term). The long term effects of overclocking are not really proven for the G3 as the CPU is still fairly new. I've heard some CPU manufacturers claim that over time the 'headroom' in an overclocked CPU will dwindle, forcing you to lower speeds. For more on the risks of overclocking see the FAQ's Overclocking Risks topic area.

Now that you hopefully have an understanding of speed settings, the next page of this guide covers installation and troubleshooting tips as well as Mac system compatibilty. Or you can use the links below to jump to another part of this guide or other related links.