Page 1 of 2 1 2 LastLast
Results 1 to 20 of 24

Thread: 750GX in 9500 - Vigorous

  1. #1
    Join Date
    Jan 2004
    Posts
    237

    Default 750GX in 9500 - Vigorous

    Respectfully declining to take “not supported” for an answer, John ordered the Powerlogix 750GX G3 zif for his 9500.

    Having learned (the hard way) the teachings of the Prophet of Ram, TZ, the board had already been pulled, ram slots cleaned, all ram not able to be confirmed as 60ns pulled, chips of the same mfg interleaved, the rest laid to the side. There was a good deal of either unidentifiable or “jive” ram in this machine and this left him with 4ea 128MB, 60ns EDO sticks. More will have to be had to meet the planned software demands.

    The GX was installed in a XLR8 Carrier Zif card. The card was toggled On=1,2,4,6,7,8,10,11. The 750GX has it’s own set of 6 toggles that apparently relate to the start-up speed. That was set to on=3,5.

    Having understood that these chips run a bit hotter than other G3 zifs, some extra care was taken to see that the heatsink was well mated to the chip and the thermal compound was correctly applied. The IDE ribbons had previously been exchanged for the small serial cables and the IDE drive fitted with a converter board - DK-0719SA. This removed air movement blockage and added some speed to the drive.

    Operating systems: X 10.3.3; 9.2.2
    Initial cache software on 9.2.2 - XLR8's MachSpeed
    In X - Powerlogix CPU Director 1.5f2

    The GX has a 1MB backside cache running at processor speed.

    The first boot was in 9.2.2 with no apparent problems - but the MachSpeed software limits the CPU to 800Mhz (16x). This won’t do for the long term, thus sayeth John. The XLR8 software offers adjustment of the backside cache but not on the CPU multiplier. Disables mobo cache and speculative processing.

    In 10.3.3, CPU Director 1.5f2 permits adjustment of the multiplier all the way to 20x (1GHz), offers options on start up speed and disables speculative processing. The mobo cache is not included and is apparently still in the works.

    It will boot into 9.2.2 without incident. In X, booting is bandy-legged. Often there is a panic, termed a “memory exception,” that the debugger is unable to trace, at or near the time CPU Director shifts from start-up speed to full tilt. Given the many options for adjustment - Carrier card, zif and software. Figuring out what will work for the booting problem will take some time.

    After booting with cache off, the cache can be set with the software, and you’re cooking.

    At this point, 20X causes instant crash. It will run at a maximum 19x multiplier (950MHz per software). The degree of stability at that speed we’ve not had enough time to determine. At 18x, (900 MHz) it does appear stable and vigorous. In OSX, finder functions are quite brisk, animations are smooth, video performance appears to be better (not sure how that works), CPU temp is reported at around 56C after 3 hours of continuous use. No lock-ups, no PCI problems. Yet.


    We’ll be doing some more revealing tests this week-end and will report back later.

  2. #2
    Join Date
    Jan 2002
    Location
    NW Montana
    Posts
    8,197

    Default

    This is sweet. Nice work. Randy

  3. #3
    Join Date
    Aug 2003
    Location
    Under the Midnight Sun
    Posts
    656

    Default Great!!

    This is VERY well done and a service to the 7300-9500 and clone community!

    Would it be possible to see your hardware configuration (slots) and cards and maybe an Xbench in 10.3.3?


    Thanks so much!!
    RG

  4. #4
    Join Date
    Jan 2001
    Location
    Mobius Strip
    Posts
    13,045

    Cool

    I may have to dust off my 7300 some day and pop one of those G3/1000s in there ;-)

  5. #5
    Join Date
    Jan 2004
    Posts
    237

    Default

    Howdy

    We’ll have some Xbench scores for you, and slot configuations, rguising. Probably in the next day or two. At this point the slot placement dosen’t seem to make any more difference than when it was running a 604e. We may learn better because there are some stability problems which, of course, become more pronounced at the higher multipliers. We’ll hit it this week-end - try and figure some of this out.

  6. #6
    Join Date
    Jan 2004
    Posts
    237

    Default

    We’ve gone a few more rounds with the 750GX, and other than a couple of black eyes and a few loose teeth, we’re still standing.

    The big problem is still stability at a speed that makes having the chip worthwhile. We’re talking about around $380 bucks here just for the zif. You can get a Sonnet G4 running at 800 for a bit less than that, get your AltiVec, and whatever support you can squeeze out of them. We’re not posting this information as a suggested upgrade. At this point, I could not recommend it. That may change. This is just a road we’re going down. Not necessarily sensible.

    We are still going through cache software in an effort to find one or a combination of them that will permit sable operation at 18x. On this unit that would be about 900 MHz. This has proven to be more of a challenge in OS 9 than in X. The Powerlogix CPU Director 1.52f works quite well in X (10.3.3) but only after you have booted with cache off. That is a nuisance. We’ll be working on that. In OS 9 there is a conflict with the Quicktime (6.04) extension.

    After booting with the cache off, and letting the desktop load, we can run this chip at 950 MHz (18x) without too many problems. The cache can be enabled on the fly. Now the 750GX has two PLLs that regulate the multiplier. PLL0 is for start-up and PLL1 is a setting after startup. This transference takes place on chip. The idea is that the ratio should be lower at start-up and once the PLL0 locks in, PLL1 negates it and takes over at a higher speed. It can also be done via software, which is where CPU Director comes in. The problem we are having with it is that it crashes, without fail, after enabling the cache, and before the desktop is formed if you let CPU Director set the cache. From what I am seeing, it appears that CPU Director is attempting to set the cache before the chip is ready to receive the instructions. Haven’t figured out how to change that yet.

    We do have some Xbench results. We borrowed a Sonnet G4 800 and a G3 466 XLR8 CPU and stuck them in the machine for perspective. All CPUs were operated by CPU Director. The machine is a 9500 with about 1.4 Gig of ram. PCI configuration is:

    Top - Generic USB
    2nd - Seritek 1S2
    3rd - Generic Firewire
    4th - nothing
    5th - nothing
    6th - Radeon Mac Edition (not 7000)

    750GX Sonnet G3500
    CPU 63.63 94.56 34.04
    GCD Loop
    95.88- 3.74 Mops/sec 84.78 - 3.31 49.44 - 1.93
    Floating Point Basic
    136.87 - 494.96 Mflops/sec 95.86 - 346.68 80.74 - 291.98
    vecLib FFT
    26.10 - 405.16 Mflops/sec 100.46 - 1.56 Gflops/sec 13.71 - 212.89
    Floating Point Library
    146.88 - 5.88 Mops/sec 95.08 - 3.81 Mops/sec 85.54 - 3.34
    AltiVec Basic 98.27

    We’ll post more later.

  7. #7
    Join Date
    Aug 2003
    Location
    Under the Midnight Sun
    Posts
    656

    Default Thank you Wizzard!!

    This is wonderful information (yet again...). I am truely impressed by your dedication to the old world macs some of us are 'still' using.

    my intention was to replace a Sonnet G4/450 which was starting to show some age doing iPhoto and iMovie. Your information has been invaluable, as it is exactly the road I was going down. It appears the $$ spent may not justify the upgrade.

    The Sonnet G4 / 800 is out of the question, it WILL not work with a S900 and any version of OSX despite what Sonnet tells us. Unless I get rid of the Mac Edition 7000 which is the only thing I have yet to try...

    Xbench on my machine (yes, not highly accurate, but somewhat repeatable)

    Sonnet G4 / 450
    CPU 53.26
    GCD Loop 47.51 - 1.86 Mops /sec
    Floating Pt. Basic 67.19 - 242.99 Mops /sec
    Vec Lib 53.08 - 824.99 Mops /sec
    Floating Pt. Lib. 82.14 - 3.29 Mops /sec

    Xbench final 42.46 / found as Marathon S900 on the xbench comp. site

    Panther 10.3.3
    Sonnet G4-450 w/ cpu director 1.5f2
    1M backside enabled - 300 / 3:2
    256mB interleaved
    ATI Mac Radeon Retail 7000 - Slot A
    Linksys Wireless PCI 54G - Slot B / off
    Acard 133 6280M - Slot F
    Seagate 7200RPM 80GB ATA ST380024A

  8. #8
    Join Date
    Jan 2004
    Posts
    237

    Default

    Those scores didn’t post anywhere near how I had it. Let’s try it this way:

    750GX@950
    CPU 63.63
    GCD Loop - 95.88 @ 3.74 Mops/sec
    Floating Point Basic - 136.87 @ 494.96 Mflops/sec
    vecLib FFT - 26.10 @ 405.16 Mflops/sec
    Floating Point Library - 146.88 @ 5.88 Mops/sec

    Sonnet G4@800
    CPU - 94.56
    GCD Loop - 84.78 @ 3.31 Mops/sec
    Floating Point Basic - 95.86 @ 346.68 Mflops/sec
    vecLib FFT - 100.46 @ 1.56 Gflops/sec
    Floating Point Library - 95.08 @ 3.81 Mops/sec
    AltiVec Basic - 98.27


    XLR8 750G3@500
    CPU - 34.04
    GCD Loop - 49.44 @ 1.93 Mops/sec
    Floating Point Basic - 80.74 @ 291.98 Mflops/sec
    vecLib FFT - 13.71 @ 212.89 Mflops/sec
    Floating Point Library - 85.54 @ 3.34 Mops/sec

    Last night we ran the chip in X at 1GHz for about 20 min before lock up. I'll be stomped if it didn't look like a ram failure again! Could it be that we ain't got ram good enough for these speeds?

    If so, this is where the costs begin to add up.

    Will post any elightenments.

  9. #9
    Join Date
    Aug 2003
    Location
    Under the Midnight Sun
    Posts
    656

    Default Could be as simple as ram...

    I can only assume you have removed / or disabled the onboard cache?

    Am using a pair of OWC 128mb FPM Dimms in the S900. Also have jumpered J38 which effectively removes the cache from the S900 board.

    I am sure you know all the ins and outs of the memory configs. My only suggestion might be to remove all but 1 Dimm module and see if it is stable, this will effectively eliminate the interleaving as well. Do you know if you have 60ns Dimms available?

    RG

  10. #10
    Join Date
    Jan 2004
    Posts
    237

    Default

    Onboard cache is still operational - as CPU Dir doesn't appear to contemplate it's disablement. We could physically remove the R31 resistor near the middle of the A1 ram slot - it's my understanding that this will disconnect the mobo cache. Problem is that we don't know and have not been able to find the proceedure for that. I'm not going to take the pliers to it until I've got some idea what's supposed to happen.

    We just got back together from another tear down. We'd cleaned the thing well before we started. We had OWC ram in it, had pulled some other sticks from the 9600 as well. We assumed that the OWC ram would all be the same. It's had the dickens tested out of it. But when we got down on it though - there were some sticks that were constructed differently. I can't imagine that it would matter - but we wound up pulling all of the OWC ram and using 8 sticks that I got from ramdirect on e-bay. They look better in quality and they are all exactly the same in construction. All of those are 60ns EDOs 4k.

    We'll fire it up again this evening and see if we did any good. The mobo cache is the wild card. Wish I knew how to clip it off.

  11. #11
    Join Date
    Jan 2002
    Location
    NW Montana
    Posts
    8,197

    Default

    I don't know enough about the 6 slot machines to advise for sure but - all the 73/75....85/86 needed the L2 Cache pulled out.

    Look here. http://www.macgurus.com/products/mot.../mbppc9500.php

    Do you have anything in the ROM slot?

  12. #12
    Join Date
    Jan 2004
    Posts
    237

    Default

    Nothing in the Rom Slot.

    I think the onboard cache is a serious limiting factor. We've spent most of the day trying to get OS 9 to act right. We had the multiplier on the chip set to 16x. That's a startup speed of 800 MHz. You can boot this chip in OS9 with XLR8 MSC for 9 - it will "disable" the mobo cache by software, but MSC has the limitation of not allowing manipulation of the multiplier on the fly. Plus - at 800MHz start up sometimes you get there and sometimes you don't. End up having to re-boot about have the time. We tried Sonnet's software as well, and it will work but - does not change the results - We also tried earlier versons of CPU director some of which will not get you even to the desk top. Even tried to see if a Newer Tech Cache Control would work, it would load, but failed to initialize the cache. Like booting with no software at all. Slow chip.


    For a while there, it seemed like we were having conflits with some of the extensions and Powerlogix's CPU Director 1.52f but after a bushel of boots you see that it's first one and then another - wherever the loading happens to hang.

    It would seem that if the mobo cache were THE reason for the instability, things would go better when booting at 16x in XLR8's software because that does disabled the mobo cache as well. Sonnet's may too. It's really about the same.

    The solution that keeps shining through is to lower the startup multiplier til you get a stable startup and try and use Powerlogix's Cache Control to increase the multiplier once in. It's the only software that has that feature that I know of. This is working. But the fastest reliable bus speed that we've been able to sustain with our set up is 750MHz.

    We ran across MacBench 5 scores from when we had the 604e in it - ran the processor test on the 750GX set at 750 and the results were that the GX is 650% better. Quite a chunk.

    We'll run some of the other CPU's and post a spread.

  13. #13
    Join Date
    Jan 2004
    Posts
    237

    Default

    As it stands at this point, the upper limits for stable operation of the 750GX zif, as we have it configured, are 15x (750MHz) in OS 9 and 19x (950MHz) in OSX. As I mentioned the required cache enabling software that we are using is the Powerlogix CPU Director 1.5f2 in both X and 9. This 9500 will not boot reliably in 9 at a 16x multiplier. We found it was best to toggle the zif for 12x (600) and then reset the multiplier on the fly with CPU Dir once booted in.

    In X (10.3.3) we are still having the problem of not being able to boot with CPU Dir enabling the cache on boot. That will freeze at the point the desktop tries to load without exception. We are using XpostFacto 3.17, whatever, to boot X, no need to slow down the processor, as that seems to have no effect on the freeze. We have not been able to determine what this problem is. We have tried to set out the XpostFacto kexts (there are two) but that appears to have no effect. We have tried removing the extensions.kextcache and extensions.mkext from the X system library on reboot when extensions are removed or added. But that has not offered any change in the result.

    We’ll be working on this for a while, I think.

    Now for the good side:

    The 750GX chip is hot - there’s no doubt about it. Finder operations in OS 9 border on violent. Opening windows move so fast, the mouse “click” seems like it’s the window slapping the inside of the screen.

    Here are some test results for a comparative view in OS 9.2.2 running MacBench 5:
    The disk we are using here is a Maxtor 120 GB ATA133 (9.3ms seek) with a Serial ATA Bridge to a Seritek Card in the second slot from the top.


    750GX@800(this is not a reliable speed for us)
    Processor (raw score) - 2991
    up from 604e - 830%

    Floating Point (raw score - 2630
    up from 604e - 396%

    Disk (raw score) - 3022
    up from 604e on SCSI - 415%

    750GX@750
    Processor - 2719
    up from 604e - 755%

    Floating Point - 2460
    up from 604e - 371%

    Disk - 3038
    % up from 604e on SCSI - 417%

    Sonnet G4@800
    Processor - 1962
    up from 604e - 545%

    Floating Point - 1953
    up from 604e - 294%

    Disk - 2887
    up from 604e on SCSI - 397%

    XLR8 750G3@500 (512cache@250)
    Processor - 1605
    up from 604e - 445%

    Floating Point - 1658
    up from 604e - 250%

    Disk - 2621
    % up from 604e on SCSI - 360%

    How reliable are these scores? I have no idea. We may not have things set up right, but we are trying to hold everything else at a constant while we just swap the CPU. Given the comparitively low scores on the Sonnet G4, I am wondering if it would run better on the Sonnet software, rather than CPU Dir. The caches were enabled. This is what we got. Be nice if we knew more about what we were doing. Like a chimpanzee with a pocket watch, I’m a lot better at banging on it and yelling than figuring out how to wind it so that it will work. We’ll keep trying.

    Today we are testing the 750GX in OSX on some applications and will have some other Xbench scores. We’ll post that too.

  14. #14
    Join Date
    Aug 2003
    Location
    Under the Midnight Sun
    Posts
    656

    Default Thanks for the update..

    I for one am following your thread with much interest (am sure you could tell that).

    I know there are resources out there for telling you how to disable the onboard cache, but I am not advocating that... It will take a soldering iron and some experience (as you inferred before).

    Again, I may try this upgrade on my machine as the G4 /450 seems to be the end of the road for the Umax's (the S900's at least in X).

    Traveling this week, so will peer in when I can...

    RG

  15. #15
    Join Date
    Jan 2004
    Posts
    237

    Default

    We were able to get it to run at 1 Gig long enough to get some Xbench results at that speed. It is not a stable speed on the 9500. 50 MHz back and we can boogie all night. At 1 Gig, I think the chip gets hot pretty quick. The temperature indicators are goofy so it’s hard to tell. We’ve got four thermometers : XLR8's, Powerlogix’s Sonnet’s and Newer Techs. All show different temps. Sonnet’s shows the same temp all the time so it probably can’t work for some reason. Powerlogix’s flips back and fourth like it can’t make up it’s mind. XLR8 shows a rising CPU temp at 1 Gig, although the starting temp was below room temp. Newer Techs seems to be working but is way larger than the rest. Anyway, we pulled the zif, cleaned the thermal paste off, polished the heat sink contact point area, cleaned everything with 99% alcohol, and reapplied a thin layer to the zif with a razor blade and “tinned” the heat sink. We mated the heat sink so that the clips would hold the thing flat on the chip, rigged up a fan to blow some air right on it. With that set up, it will still only last for a few minutes. Don’t think the reason is heat. Could be the mobo cache again or . . . . just to darn fast for this old machine. But 950 is qucik.

    I really wanted to know what we were missing by not being able to run at 1 Gig . Here’s the differences:

    750GX@950
    CPU 63.63
    GCD Loop - 95.88 @ 3.74 Mops/sec
    Floating Point Basic - 136.87 @ 494.96 Mflops/sec
    vecLib FFT - 26.10 @ 405.16 Mflops/sec
    Floating Point Library - 146.88 @ 5.88 Mops/sec

    750GX@ 1 Ghz
    CPU 66.78
    GCD Loop - 101.04 @ 3.95 Mops/sec
    Floating Point Basic - 143.89 @ 520.34 Mflops/sec
    vecLib FFT - 27.39 @ 425.19 Mflops/sec
    Floating Point Library - 152.93 @ 6.12 Mops/sec

    Incidently, we did re-install the Sonnet software to test the G4 800 CPU with it’s own software running the card. The results were about the same - maybe 2 to 5 per cent higher.

    We’ve been comparing performance in some apps in both OS 9 and X. Using the 750GX and the Sonnet G4 800. Really can’t tell much difference expect in a couple.

    In OS 9 Photoshop - the GX has the edge on opening large files and reconfiguring. Some of the filters are better on the G4.

    We also noted in OSX, iMovie had some dropped frames downloading a movie through a Dazzle converter on a VCR player with both of these CPUs. The 750GX was definitely superior. Quite usable for home stuff. The G4 800 dropped way more frames and had an audio stutter that made it essentially unuseable. Could be how we have it set up or some of other thing.


    We hooked up a DVR-107 through the SeriTek card using another Serial to IDE bridge. Those do work quite well - I can detect no difference due to the bridge. Itunes will burn to it as if it were OEM. iDVD will not. We’ve got a patch for that but haven’t gotten that fixed up yet.

    The 750GX will import a 4 min CD track in just under 45 seconds. Most of that is probably due to the hardware, but it is almost as fast as the DigAud. It was a little strange on the burning - very quick, but did not do well at telling you were you were in the burn process for some reason.
    The CD was perfect.

    It’s beginning to look brighter.

  16. #16
    Join Date
    Aug 2003
    Location
    Under the Midnight Sun
    Posts
    656

    Default Multiplier

    Wizzard,

    I just wanted to confirm you are only trying this at 50 mhz bus speed, and a 20x multiplier to achieve the 1ghz mark.

    "We were able to get it to run at 1 Gig long enough to get some Xbench results at that speed. It is not a stable speed on the 9500. 50 MHz back and we can boogie all night"

    Your statement appeared that you were trying to run above 50mhz which many 9500's are not capable of.

    Thanks so much for your contributions on here

    Would you be able to upload your Xbench run at some point at the xbench.com website, so that a complete comparison can be made? I am interested in switching to the Serial ATA card here as well!

    Thanks much (again).

    RG

  17. #17
    Join Date
    Jan 2004
    Posts
    237

    Default

    Throughout all of this, the base bus speed has not changed. 50Mhz.

    There are three levels of adjustment in this setup with a Carrier card and this Powerlogix zif. 1) the card, sets the bus speed, 2) the toggles on the zif set the startup multiplier, and 3) The software, which can change both the startup and the operational multiplier.

    The Carrier card was set to 1,2,4,6,7,8 - which gives you the base bus speed of 50MHz. 9,10,11,12 supply power to the zif’s switches - here it is set to 0n=10,11.

    The zif itself has six very small switches J1 thru J6. You can see the numbering, barely. The Powerlogix website either didn’t have any information on how to set these or we were unable to find it. They did refer you to the IBM Data Sheet on this chip. It is written in tech tongue. But we were able to finally figure out that a chart on page 45-46 is a list of multipliers AND how to set these switches. Looking at the chart, you will see, in the left column, a binary code. Ones and zeros. 1 = off, 0 = on. You will see 5 numbers, but have 6 switches! That’s because J6 is not included because its setting does not change - off. We started out with 11010 (16x), which translates to J3 and J5 on. We went back to 10110 - 12x (2,5, on) and have worked back up to 11001 - 15x (3,4, on).

    I’m not really sure how much the zif settings matter, except for the first boot, because you can set this in the software as well and it will then boot at the new chosen speed. As long as you have it set to a speed that won’t crash the machine, then you can fine tune it with CPU Director 1.52f.

    We are still getting a crash when CPU Dir tries to set the cache with the Powerlogix zif. It is odd that CPU Dir can set the cache on the other CPUs, the Sonnet G4 800, and the XLR8 G3 500, and a XLR8 G4 450, without incident. Works on all CPUs except its own. Corollarys? Something related to this chip? Or this chip in the Tsunami board? Is it a timing problem? Tries to set the cache before the frist PLL has locked in? I saw no reports of a problem with this with the 750FX (900MHz) zif. Probably ought to get the data sheet for that chip and compare. How about the chip and OS X? Is it trying to set the cache at the same time that OSX is running cache related items? Other versions of CPU Director do the same thing. (Back to 1.3 something).

    Anyway, you can see how the thinking is going.

  18. #18
    Join Date
    Aug 2003
    Location
    Under the Midnight Sun
    Posts
    656

    Default Understanding a bit more....

    Thanks for the clarification... that helps here

  19. #19
    Join Date
    Aug 2000
    Location
    Concord, CA
    Posts
    7,056

    Default

    I have no idea why you are getting a crash when CPU Dir tries to set the cache with the Powerlogix zif but I just applied the 1.5f2 in 9.2.2 on my QuickSilver G4-800DP with PLogix PF G4-1.20GHz DP 7457 CPUs. The first CPU Dir version that recognized the 7457 was 1.4f1. When I do a Get Info on CPU Dir v1.5f2 it shows 1.5f2 at the top but further down the window it still shows 1.4f1 I guess because this latest CPU Dir is for multiple type CPUs.

    Looking at the 1.5f2 ReadMe, the versions for your PL zif (note the dates):

    Version History
    1.5f2 (13 Apr 2004) - Changed "Enable on Restart" to "Apply on Restart." This means that the On/Off setting for the cache will be applied at boot time, so you can now force the cache into disabled mode rather than whatever mode the firmware set up
    - The built-in L2 caches on various CPUs now have the option to enable or disable at boot time rather than always enabling.

    1.5f1 (12 Apr 2004) - Updated to support 750GX processors.
    - Fixed a kernel panic when executing "sysctl -a" from the command line.

    So 1.5f2 (13 Apr 2004) really did nothing serious to tweak the 750GX of 1.5f1 (12 Apr 2004). And anything earlier than 1.5f1 would not recognize your 750GX. k

  20. #20
    Join Date
    Jan 2004
    Posts
    237

    Default

    I think you’ve got me aimed in the right direction, kaye. It has to be in how the software operates. It is a complicated set of variables to hold in the mind for contemplation. I’m in way above my head.

    But then . . . . sometimes that’s a good way to learn. Unfortunately, in the process, my ignorance is exposed. Wasn’t it Abe Lincoln that said, “it’s better to be silent and thought a fool than to speak and remove all doubt.”

    At any rate, placing that wisdom to one side and operating on the proposition that ignorance can be overcome with brute persistence, I’ve got the notion that before I get out HexEditor and risk ruining something bigtime, I should eliminate hardware variables. There were only two, that I could think of, left this morning, and I just finished eliminating one of those.

    The Carrier card has four switches, 9,10,11 and 12. I’m not sure what these switches set, because the bus speed is set by switches 1-8 and the processor speed is set by the jumpers or jumper switches on the zif. My assumption was that it was the power supply, but since this application is “unsupported” I don’t have much information on how to set them. Nothing at all for this chip. But now I'm even more unsure about what they do.

    I came up with 16 unique combinations that you can get with a 4 switch set, with two possible settings. Here’s what I can report:

    1) None of the settings would permit CPU Dir 1.5f2 to enable the cache on boot. Always a crash in that case.

    2) None of the settings kept the machine from booting on the second attempt when CPU Director disables the cache because of a cache enabling related crash.

    3) None of the settings kept the cache from being able to be enabled after booting to the desktop with the cache disabled.

    In each of the 16 possible settings I removed Extensions.kextcache and Extensions.mkext, although perhaps superfluous, I did not want the boot history to influence the next attempt.

    Now I know that it is not in the settings of 9,10,11 and 12.

    The final issue, if I have them all, is whether the motherboard cache could be a the root of the problem? CPU Director does not affect on board caches.

    I understand that there is someone using a similar set-up, without problems, in a 9600 with a Kansas mobo, where there is no on-board cache. He is using the FX chip with CPU Dir 1.3fx. Correct me if I err here, but I understand that in the Kansas set-up the cache was on the OEM processor daughtercard and when that was pulled came out with it.

    Since we’re not going back to any configuration with this machine that would admit the onboard cache, I’d disable it if I knew how. R31 resistor needs to be removed. Anyone know how to do that? Pliers? Hot Iron? At this point "shotgun" has a nice ring.

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •