Observations during PM8500 Accelerator Testing

Testing Observations - XLR8-180mhz 604e:

Free & Easy Speed Boost:
As previously stated - the Motorola math extension (LibMoto ) makes a tremendous difference in floating point performance. Since its free at their web site, its something you should at least try for yourself. To learn more about LibMoto - read their excellent FAQ . LibMoto for end users is available at: Motorola's Web Site .

Static precautions:
XLR8 provides a disposable wrist strap and we recommend its use. I personally leave the power cord installed in the machine to assure that the chassis ground is connected to a real earth ground (power strip OFF). Touching the power supply metal housing will also assist in discharging any static charge before touching the system or addin card.

Cable clearance:
On the PowerMac 8500 there is a ribbon cable that connects to the motherboard very close to the processor slot. I recommend that you flex this cable away from the processor card connector to avoid contact with the dip switch solder side pins on the XLR8 card. By the time I realized this the ribbon cable had been scuffed from repeated contact during card insertions. Use care in inserting or removing the XLR8 card - use a straight insert/remove force as the card is not long enough to contact the stock apple card guides ( at least on the sample card). The card is supported adequately by the rotating latch which has a card support notch that engages on the center edge of the card. I'd like to see XLR8 increase the length of the card by about 1/16" on each end to better engage the outside card guides/support. Comparing the XLR8 card to the original Apple card visibly shows the shorter length. Again - this is not a serious concern as the rotating locking latch provides the main vertical card support.

Apple Memory Control Panel - Disk cache settings:
Disk scores were somewhat disappointing until I increased the Apple Disk cache from the default of 96k, to 512k. This made a huge difference not only in MacBench scores but in actual finder performance. I did not experiment with settings above 512k, but contrary to what many experts say about leaving the 96k setting alone, this change made a noticable difference in operational and benchmark speeds.

Bus Speed Adjustments:
The XLR8 card provides for bus speed settings from below 40mhz ( not recommended ) to 60mhz. We recommend you first install it at the default settings of 45mhz bus and 180mhz (x4 ratio) CPU speed. Most users will be content with the performance increase this provides if your system is a 120 or 132 mhz machine. Remember to press the CPU reset switch before reinstalling the card after changing any settings to ensure the new speed setting will be recognized ( also a must when initially installing the card). The manual from XLR8 has excellent instructions with pictures for the Powermac 7500, 8500 and 9500 showing the installation procedure and reset switch locations for each machine. No other card out there offers the potential for maximizing your system due to the number of bus speed settings. In our experience with this particular PM8500, having the 47.6mhz setting enabled us to increase performance by more than 10% over similar cards that did not have a setting between 45 and 50mhz. This testing takes time and patience, but can be rewarding in the final analysis. Our only wish is that the XLR8 card dip switches were accessible with the processor card installed. However, since a CPU reset is usually recommended after each speed setting it is best to remove the card to easily reach the reset switch. Careful examination of the 8500 with the XLR8 card installed showed that a non-conductive 1/8" rod of nylon or similar material could have been used - guided between the XLR8 heat sink and the processor board. It is possible that circuit layout defined the location of the dip switches as shorter trace lengths are required for better reliability in some circuits.

Apple L2 Cache limits:
If you have the stock Apple L2 cache, I suggest you leave the XLR8 card at its default setting of 45mhz bus/180 cpu. If you insist on attempting to increase the speed ( and XLR8 encourages this! ) I'd suggest trying the 50mhz bus setting first. Verify that all your applications run fine, and run the system for at least several hours before closing up the case. Should any problems arise ( lockups, failure to boot ) take the bus speed down a notch or two and repeat. With the system we had the Apple cache would not run (or boot) at bus speeds faster than the default 45mhz. Some users have reported running the bus speed at 50mhz with the stock Apple cache. The speed rating of the chips on this particular Apple L2 cache are fast enough for a 60mhz bus according to my calculations, but either the cache card design or another component in the system was the limiter. It is interesting to note that the Newer card with the stock Apple cache runs at 52.5 mhz bus. It seems there are timing/wait state differences between the XLR8 & Newer Cards. If in doubt that the L2 cache is the culprit ( most likely ), you can remove the L2 cache and retest. Unfortunately as all PM8500 owners know - the Apple case design makes adding/removing memory or cache a 20 minute disassembly exercise. The 7500 is much better in this respect as are the new PM 8600/9600.

OverClocking:
In my 15+ years of tech support and tweaking hardware I have never encountered a company quite like XLR8. Not only do they not care if you overclock the card - they encourage it! I found their tech support to be top notch, caring, and excellent overall. When the first cache I received would not run above 50mhz, they tested and shipped federal express a replacement module that they had run reliably at 53mhz in a PM7500. Sadly this module did not run any faster than the original one in this particular 8500, but their attitude and response was first class in all respects. The cards two year warranty will not be voided by overclocking the CPU. Try that with the other guys! I found XLR8's attitude refreshing to say the least. I'm still stunned, as most vendors would immediately cry "foul" and "warranty void" at a minutes notice. I hope the competition is listening. I think companies like this deserve your consideration next time you look to spend your money on a Mac upgrade. BTW - their Tech Support line is Toll-Free also.

Bus Speed is Key to Performance:
Obviously, the highest performance overall will result from using the highest bus speed that runs reliably in your system. A 51 mhz bus speed system at 3.5x cpu ratio will usually outperform a 45mhz x 4 cpu ratio on all tests . Installed hardware ( PCI cards, disks, etc. ) are another variable that can affect the maximum bus speed that your computer will sustain. The main advantage to the XLR8 card is the number of bus speed options it provides - many more than the competition. In the case of our particular PM8500, no other card we've seen would have allowed the 47.5mhz bus speed, others would have forced us to reduce the setting to 45mhz. Experimenting takes a little time up front - but will pay off in the long run. It is critically important to ensure that the system is running reliably - be sure to verify that the system is stable with all your applications for several days before "closing the case". Performance without stability is nothing but frustration.

Memory Interleaving:
If your motherboard supports it - always add dimms in pairs in the Ax/Bx slots. This interleaving effectively makes the memory bus 128bits wide. This interleaving really helps in the real world when you're "beating up" the processor, although true multi-tasking OS's benefit the most. To see if your system is using interleaving - download the DayStar Npower control panel. (link removed after daystar.com URL gone) This control panel is a dream come true! Very impressive, but I expected nothing less from the people who designed the original Macintosh Multi-Processor cpu design. (The Genesis MP series.)
Note: Recent Newer Technologies MaxPowr owners have reported that they had to stop interleaving memory (move dimms) to enable the cards to run at the higher processor speeds ( above 200 mhz). We did not do this during our attempts to overclock the XLR8 card as removing interleaving extracts a 5% to 15% performance penalty ( depending on who you talk to, and what app/benchmark you run ). I personally would not consider that a viable alternative. Why cut memory performance up to 15% to gain another 10-15% in cpu speed?? This problem has been noted on several web sites and appears to be related to only certain PM9500 systems ( not a common problem). My opinion is these users should verify that all their memory is of similar type and speed (at least in the matched pairs). Timing differences when two dimms from different manufacturers are in the A/B slots could possibly cause the problem. (I do not believe this problem is related to the Newer Technology boards.)

UPDATE:
See the Newer Review for important test result updates.

604 vs 604e:
In case you didn't know, another benefit of the 604e upgrades is the fact the "e" has twice the internal CPU cache of previous models.

= Back to the XLR8 Review page =
= Back to WWW.XLR8YOURMAC.COM =

Copyright © 1997.
No part of this sites content is to be reproduced in any form without written permission.
All brand or product names mentioned here are properties of their respective companies.
Users of this web site must read and are bound by the terms and conditions of use.


Testing Observations - XLR8-180mhz 604e: Free & Easy Speed Boost: As previously stated - the Motorola math extension (LibMoto ) makes a tremendous difference in floating point performance. Since its free at their web site, its something you should at least try for yourself. To learn more about LibMoto - read their excellent FAQ . LibMoto for end users is available at: Motorola's Web Site . Static precautions: XLR8 provides a disposable wrist strap and we recommend its use. I personally leave the power cord installed in the machine to assure that the chassis ground is connected to a real earth ground (power strip OFF). Touching the power supply metal housing will also assist in discharging any static charge before touching the system or addin card. Cable clearance: On the PowerMac 8500 there is a ribbon cable that connects to the motherboard very close to the processor slot. I recommend that you flex this cable away from the processor card connector to avoid contact with the dip switch solder side pins on the XLR8 card. By the time I realized this the ribbon cable had been scuffed from repeated contact during card insertions. Use care in inserting or removing the XLR8 card - use a straight insert/remove force as the card is not long enough to contact the stock apple card guides ( at least on the sample card). The card is supported adequately by the rotating latch which has a card support notch that engages on the center edge of the card. I'd like to see XLR8 increase the length of the card by about 1/16" on each end to better engage the outside card guides/support. Comparing the XLR8 card to the original Apple card visibly shows the shorter length. Again - this is not a serious concern as the rotating locking latch provides the main vertical card support. Apple Memory Control Panel - Disk cache settings: Disk scores were somewhat disappointing until I increased the Apple Disk cache from the default of 96k, to 512k. This made a huge difference not only in MacBench scores but in actual finder performance. I did not experiment with settings above 512k, but contrary to what many experts say about leaving the 96k setting alone, this change made a noticable difference in operational and benchmark speeds. Bus Speed Adjustments: The XLR8 card provides for bus speed settings from below 40mhz ( not recommended ) to 60mhz. We recommend you first install it at the default settings of 45mhz bus and 180mhz (x4 ratio) CPU speed. Most users will be content with the performance increase this provides if your system is a 120 or 132 mhz machine. Remember to press the CPU reset switch before reinstalling the card after changing any settings to ensure the new speed setting will be recognized ( also a must when initially installing the card). The manual from XLR8 has excellent instructions with pictures for the Powermac 7500, 8500 and 9500 showing the installation procedure and reset switch locations for each machine. No other card out there offers the potential for maximizing your system due to the number of bus speed settings. In our experience with this particular PM8500, having the 47.6mhz setting enabled us to increase performance by more than 10% over similar cards that did not have a setting between 45 and 50mhz. This testing takes time and patience, but can be rewarding in the final analysis. Our only wish is that the XLR8 card dip switches were accessible with the processor card installed. However, since a CPU reset is usually recommended after each speed setting it is best to remove the card to easily reach the reset switch. Careful examination of the 8500 with the XLR8 card installed showed that a non-conductive 1/8" rod of nylon or similar material could have been used - guided between the XLR8 heat sink and the processor board. It is possible that circuit layout defined the location of the dip switches as shorter trace lengths are required for better reliability in some circuits. Apple L2 Cache limits: If you have the stock Apple L2 cache, I suggest you leave the XLR8 card at its default setting of 45mhz bus/180 cpu. If you insist on attempting to increase the speed ( and XLR8 encourages this! ) I'd suggest trying the 50mhz bus setting first. Verify that all your applications run fine, and run the system for at least several hours before closing up the case. Should any problems arise ( lockups, failure to boot ) take the bus speed down a notch or two and repeat. With the system we had the Apple cache would not run (or boot) at bus speeds faster than the default 45mhz. Some users have reported running the bus speed at 50mhz with the stock Apple cache. The speed rating of the chips on this particular Apple L2 cache are fast enough for a 60mhz bus according to my calculations, but either the cache card design or another component in the system was the limiter. It is interesting to note that the Newer card with the stock Apple cache runs at 52.5 mhz bus. It seems there are timing/wait state differences between the XLR8 & Newer Cards. If in doubt that the L2 cache is the culprit ( most likely ), you can remove the L2 cache and retest. Unfortunately as all PM8500 owners know - the Apple case design makes adding/removing memory or cache a 20 minute disassembly exercise. The 7500 is much better in this respect as are the new PM 8600/9600. OverClocking: In my 15+ years of tech support and tweaking hardware I have never encountered a company quite like XLR8. Not only do they not care if you overclock the card - they encourage it! I found their tech support to be top notch, caring, and excellent overall. When the first cache I received would not run above 50mhz, they tested and shipped federal express a replacement module that they had run reliably at 53mhz in a PM7500. Sadly this module did not run any faster than the original one in this particular 8500, but their attitude and response was first class in all respects. The cards two year warranty will not be voided by overclocking the CPU. Try that with the other guys! I found XLR8's attitude refreshing to say the least. I'm still stunned, as most vendors would immediately cry "foul" and "warranty void" at a minutes notice. I hope the competition is listening. I think companies like this deserve your consideration next time you look to spend your money on a Mac upgrade. BTW - their Tech Support line is Toll-Free also. Bus Speed is Key to Performance: Obviously, the highest performance overall will result from using the highest bus speed that runs reliably in your system. A 51 mhz bus speed system at 3.5x cpu ratio will usually outperform a 45mhz x 4 cpu ratio on all tests . Installed hardware ( PCI cards, disks, etc. ) are another variable that can affect the maximum bus speed that your computer will sustain. The main advantage to the XLR8 card is the number of bus speed options it provides - many more than the competition. In the case of our particular PM8500, no other card we've seen would have allowed the 47.5mhz bus speed, others would have forced us to reduce the setting to 45mhz. Experimenting takes a little time up front - but will pay off in the long run. It is critically important to ensure that the system is running reliably - be sure to verify that the system is stable with all your applications for several days before "closing the case". Performance without stability is nothing but frustration. Memory Interleaving: If your motherboard supports it - always add dimms in pairs in the Ax/Bx slots. This interleaving effectively makes the memory bus 128bits wide. This interleaving really helps in the real world when you're "beating up" the processor, although true multi-tasking OS's benefit the most. To see if your system is using interleaving - download the DayStar Npower control panel. (link removed after daystar.com URL gone) This control panel is a dream come true! Very impressive, but I expected nothing less from the people who designed the original Macintosh Multi-Processor cpu design. (The Genesis MP series.) Note: Recent Newer Technologies MaxPowr owners have reported that they had to stop interleaving memory (move dimms) to enable the cards to run at the higher processor speeds ( above 200 mhz). We did not do this during our attempts to overclock the XLR8 card as removing interleaving extracts a 5% to 15% performance penalty ( depending on who you talk to, and what app/benchmark you run ). I personally would not consider that a viable alternative. Why cut memory performance up to 15% to gain another 10-15% in cpu speed?? This problem has been noted on several web sites and appears to be related to only certain PM9500 systems ( not a common problem). My opinion is these users should verify that all their memory is of similar type and speed (at least in the matched pairs). Timing differences when two dimms from different manufacturers are in the A/B slots could possibly cause the problem. (I do not believe this problem is related to the Newer Technology boards.) UPDATE: See the Newer Review for important test result updates. 604 vs 604e: In case you didn't know, another benefit of the 604e upgrades is the fact the "e" has twice the internal CPU cache of previous models.
= Back to the XLR8 Review page = = Back to WWW.XLR8YOURMAC.COM =
Copyright © 1997. No part of this sites content is to be reproduced in any form without written permission. All brand or product names mentioned here are properties of their respective companies. Users of this web site must read and are bound by the terms and conditions of use.