Results 1 to 13 of 13

Thread: Hankern’ for Horsepower, Pt. 2 -Triumph of the Ancients- 9600

  1. #1
    Join Date
    Jan 2004
    Posts
    237

    Default Hankern’ for Horsepower, Pt. 2 -Triumph of the Ancients- 9600

    My son came home Friday and pulled out an XLR8 G4 450 CPU he’d traded some “stuff” for with a classmate. He wanted, not to put it in his 9500, but to “test” it on my old 9600. Smarter than I thought. I pulled the one sacred SCSI HD out, backed up the IDE 120, and we commenced.

    The CPU seemed almost new, Daystar in-box notice. Has the 7400 G4 chip. Allegedly never used. It came jumpered to 9x and dip switched to a 50 MHz bus. We sat that back to 49 for the initial try. I recently bought some OWC 128 ram chips 60ns and one of them was bad. In the process of isolating the bad stick, all of the ram in this machine was thoroughly tested. Real hassle. We made sure the sticks were in matching pairs, interleaved. All of the ram is 128 and recent.

    The PCI configuration:

    Top - Sonnet Tempo 133 IDE card to a Sony CRX140E Bus 1 Bus 2 Segate 120
    Second - Generic USB
    Third - Generic Firewire
    Fourth - nothing
    Fifth - nothing
    Bottom - Radeon Mac Edition (the older non-7000)

    We partitioned the IDE 120 into 8 and 112 GB. Installed 9.2.2 (OS9 Helper). Put in the CPU, cuda, trip, etc. Machine booted on the G4, installed the MSC OS9 cache control software, disabled motherboard cache, set the bus ratio to 2:1 and fiddled with it for an hour or so. It seemed solid.

    We downloaded the OSX MSC, Version 3.03 from Daystar, to be able to set the cache ratio and disable the motherboard cache in X. We used Xpostfacto 3.0a17 to install X 10.3. No installation problems, booted up. What a sight.

    Finder operations were sluggish though. Then we went about installing the MSC software for X. The instructions in the box indicated that it would be best to remove Stuffit kext and any OWC cache control kext. We found two and removed the Stuffit framework.

    MSC for X installs two extensions and a start-up package. We tried everything, short of human sacrifice, we could think of, but it will freeze at the point the start-up package loads. It then suggests, in several language, that the computer needs to be restarted. Over and over and over.

    You can get a desktop by removing one of the packets in the XLR8 start-up, but it will crash the same way when you attempted to start the MSC program.

    Powerlogix CPU Director 1.5f1 works to set the cache ratio. We have it set on 5:2. It runs solid in OS 9.2.2. But will lock-up intermittently in X. After a significant period of use. We have been mindful of the CPU temp. Does not appear to be excessive. We are subjecting it to a 48 hour ram test, because it does appear to be a “demand” related lock up. CPU Director will not disable the motherboard cache. Could that be the problem?

    In 10.3 the 9600 will:
    recognize and use the modem port, but not the printer
    recognize and burn CD's
    No problem with ADB, mouse/keyboard (USB mouse too)
    Up load and play iTunes
    AOL is seriously slow, but works
    iPhoto fully functional with USB port
    Firewire works at 400

    You just have to reboot every hour or so.

    To stop the lock-ups, we tried deceleration of the CPU (way back - actually lasts longer at the faster speeds), ram deinterleaving, PCI pulls, re-ordering, removing other start-up items and extensions. We are at a loss on recognizing what to remove (or add) here.

    We have contacted Daystar and expect to hear something this week. In the meantime, anyone have any ideas on what to try? Has anyone gotten the MSC software to load in X 10.3 after using XPostfacto?

  2. #2
    Join Date
    Jan 2001
    Location
    Mobius Strip
    Posts
    13,045

    Default

    I would suspect the RAM is the heart of the problems. It is next to impossible to go forward if RAM is suspect at all. Panther makes more use of memory than ever before. G4 pushes it faster. The system will cache as much as possible, even going so far as to compress windows and idle processes, not just deferred writes, multiple copies for each volume directory.

    But... you feel the RAM is new, and testing it further. In OS X, running "memtest" is probably the best such test there is. Also, disabling or removing startupitems and extensions in OS X is tricky, Apple changed the startupitems methods and locations, too. Might just need to run 10.3.3, XPF, and PL 1.4.1 (assuming 1.5.1 is alpha or beta and not public yet).

    how much RAM is in there? 512? 1024?
    USB cards and hubs are always suspect.

    A non-destructive test of drive media surface would be in order also, though less likely.

    With your ATA controller, no need to limit OS X to just 8GB. Will make it easier if there is more free space to burn CD or DVDs (unless you use Toast of course). I don't know what is safe minimum but it is easy to run out of free space, even when there is 1GB reported available.

    Stuffit 8.03?
    PowerLogix: I have 1.41, but you are using 1.51? Is that right?

    Only other thing is repairing permissions, and XPF 3.0a17, along with Combined Update 10.3.3 or 10.3.2... I don't think XPF is a problem, but I only focus on Beige forum over on XPF forums.

    The longer my system is up and running, the better it feels. Some background optimizing and defrag any time a file is written out that was fragmented.

    Let us know what works if you figure it out, okay?

  3. #3
    Join Date
    Jan 2004
    Posts
    237

    Default Layers of problems

    Thanks for the response, TZ.

    I must say, getting 10.3 to run well on this 9600 has turned out to be a nice challenge. It’ll teach you a thing or two about being systematic in approach and thorough in methods. Initally, it appeared that nothing we tried was going to help, but, in retrospect, there were a series of things that we “corrected” that, together, made a substantial difference in stability.

    To be totally transparent, I do not possess the knowledge to speak accurately of the cause and effect relationships, except in the form of primitive metaphor - but . . . having engaged the demon, we may have some meaningful observations for those brave souls who would attempt this transformation.

    First, we did not undertake this without attempting to understand some of the problems (and solutions) of those who had gone before. Xlr8yourmac.com. CPU reports. Important resource.

    Second, with our unit and the Sonnet IDE 133 card, we were unable to boot from the ATAPI CD drive by holding down the “C” key upon restart. It's set to master on Bus 1. ATA Drive on bus 2 as master. Wound up putting back the original SCSI CD-ROM. To my understanding, none of the burning features of the OS X system are going to work on SCSI - there are third party programs that may. But when you are facing the possibility of numerous crashes and lock-ups, booting from the CD is going to be important.

    Third, start out with a reduced set of variables. Although the motivation here is to new capabilites to the machine, if you make all of those changes all at once and then have problems with stability, you don’t know which change or changes are the source of the problems. Start with simple stability and work towards capability.

    Now, as to observations of specific components:

    XLR8 Carrier G4 450 CPU

    Problem 1:

    We are unable to boot OSX 10.3 with Mach Speed Control (MSC) for OS X installed. It will make it to the point where the start up package loads and then go to re-boot request screen. Even if you finagle around and get it booted (we could by removing the start up package from the start-up folder in systems), we got the same screen upon attempting to run the MSC program. The reason we persisted at this effort was to be able to disable the motherboard cache. Powerlogix’s CPU Director is now installed and working. It will set the cache ratio and disable speculative processing, but leaves the mobo cache operational. That leaves a wildcard for stability. We spent hours removing Kexts , other start-up packages, etc. to eliminate the conflict with the MSC software, but to no avail. Booting in verbose and with panic text enabled was not terribly enlightening as we do not yet speak in tongues (of Unix). Looked to be a conflict with the Darwin kernel rather than another package or kext. This could be a conflict caused by the Xpostfacto programs installation of “required” software to enable this class of computer to boot with Panther. Daystar tech support has not responded. Probably won’t.

    Problem 2:

    Although we got the notion that the G4 450 Carrier no longer has the problems it initially had with 6 slot PCI macs, with cards in the lower 3 slots, in our case, the problem remains. This CPU is not the Mpe or SSE model. It is recently new, though. We are unable to put the Sonnet Tempo IDE 133 card in any of the lower three slots without causing frequent lock-ups. More so of booting onto the IDE Drive. My imagination may be limited here, but I am unable to envision a meaningful upgrade of this type of mac without adding USB, Firewire and IDE. Have to have a video card. Something is going to go in a lower slot. The best stability we got was by putting the video on the bottom. Tempo on top and the USB and Firewire 2nd and 3rd. This configuration did slow data transfer from the External firewire drive to the ATA drive. I presume because both are now confined to the same PCI bus.


    Ram

    One layer of lock-up problems was the ram.

    We were unable to detect any ram errors in the chips that we had. We had to test it in OS 9 because we don’t have an X program for that yet. We’ll try to find the one you mentioned. Total test time of 48 hours. I was thinking maybe we had a fuzzball in a slot or something. So we tore the unit down to clean. Even cleaned out the power supply and took the opportunity to change out a noisy fan. We did notice that we had ram from two different sources in the unit. They looked the same from the top, you had to pull them out to really tell. We gained an increase in stability and speed by interleaving ram from the same source and pulling those that we could not tell were they came from. This showed up by the unit being able to tolerate higher cache ratios without apparent demand based lock ups. At first, it appeared that X could not tolerate the same cache ratio that OS 9 could handle - 2:1. It would lock at that speed. But after the above and some very thorough and careful cleaning of the ram slots it proved to be able to sustain that ratio. Wound up with about 1 gig of ram. De-interleaving the ram was not necessary and noticeably slowed performance at the same CPU speed and cache speed settings. Bandwidth, I presume. X is picky about ram. Very picky. I would not try it without new chips. Best you can afford.

    Video card

    We have a Radeon Mac Edition. The one ATI no longer makes. After addressing the ram, the finder would more or less operate without freezing. We could make it freeze by playing a CD in iTunes and running the Visualizer. It would run for about 15 or 20 minutes and then lock. That got us sniffing around the video set-up. We thought we were “doing right” by going to the ATI website and downloading the latest driver for this card. Turns out there is a Rom upgrade for the Mac Edition for X. Had to be put in the AppleNVRD file. Removing that upgraded Rom actually helped stabilize, at least in this context. Video performance does seem a bit choppier. Not a lot.

    We’re still testing and will report again.

    We’ll try and get some data transfer rates and more on the USB and Firewire as we add that into the soup.

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

    Cool Panther on 9600

    I think there is a Wizzard in the house! And a sleuth at that! ;-)

    Cache and crashing on startup was one of the problems with PL Cache Director 1.3 - once it would 'see' that there was a crash previously, it would disable L2 on subsequent restarts until YOU cleared the log.

    Talk about picky... must have a fuzz ball detector in there as well.

    Seems like in this case, newer isn't better, it's just new but not up to the quality of the SSE or MPe card? Not good.

    I hope you have an account so you can check Ryan's forum on OWC or offer feedback after this to others there struggling with similar configurations.

    Kaye put OS X (10.2?) on PTP and there are some other 9xxx based systems (I never learned all the names, Kansas etc of that series), not sure about Panther, but tend to be long threads - trying anyone's patience. those who hang in there, can succeed in the end.

  5. #5
    Join Date
    Jan 2004
    Posts
    237

    Default

    Last night we poked some old VCR tapes through Firewire into iMove just to see if it would choke. Nope. Video playback was a bit choppy though. Didn't fine tune the programe though. Haven’t been able to get it to crash in 48 hours of every form of exertion we could mandate. USB printing works great. Mouse too.

    Still, I was thinking about trying out an SSE card for this 9600 but when I went to Daystar, it is now shown as discontinued . . . :>(

    But you know, after all the effort to get this CPU to work, John, my son, announces that G4 7400 chip is not fast enough for what he needs. G4 seems zippy enough to me. I must be getting old. I think he wants to burn a (uniquely identifiable, mind you) hole in the space-time continuum or some such. I might be able to get him to settle for igniting all flammable objects within a 5 foot radius of the CPU on boot. I take that back - a 10 foot radius. So hot, ya have to turn it on with a stick.

    He’s been poking around the Powerlogix website and printed up some IBM specs on the 750GX G3 chip. I bet that’s the next project, if PL ever ships it. Told him to check his planned apps for AltiVec.

    It is remarkable to think that he will be able to, and wants to, update a computer I bought new about ten years ago and he grew up spilling his Kool-aid into. That poor thing has seen every possible form of abuse. I don’t think it’s been tied on to the car bumper and driven around the block . . . yet. Still going strong with only recent hard disk change. It is a testament to the mac.

    At any rate, the major problems are solved on this 9600. We’re Overclocking it to a 52 Mgh bus. 468Mhz on the CPU and the backside at 2:1. Not sure we’ll keep it there, but we’re trying to get it to act up. I believe it would run at an easy gallop all day above 50 Mhz. Finder commands in Panther are very promptly obeyed. The ATA 120 is testing at around 30Mbs for 1000k read/writes. While the old scsi 6GB grunts to get to 7Mbs. Incidently, the installer on OS X 10.3 would not allow me to install on the 120 without partitioning to 8 Gig. This is an ATA drive connected to a Sonnet Tempo 133. Puzzel.

    Things are going well enough at this point that I'm not sure we'll need to do it, but we thought about removing the R31 resistor next to the A1 ram slot to disabled the motherboard cache. We're not really sure how to do that and I don't relish the idea of ruining the mobo we've cleaned so well. Me using a soldering iron on that thing would be like trying to fix fine furnature with a sledge hammer.

    There is so much we tried that didn’t work - it’s way too much to post. If anyone has a question, I’ll be checking this post from time to time.

    After the trial by ordeal, I can say with some confidence that this can be done, no doubt about that, but IF and, and only if, you have the time, money and desire.


    Til later.

  6. #6
    Join Date
    Jan 2001
    Location
    Mobius Strip
    Posts
    13,045

    Default

    On the subject of 8GB "rule" affecting things with Panther Install CD
    the installer on OS X 10.3 would not allow me to install on the 120 without partitioning to 8 Gig. This is an ATA drive connected to a Sonnet Tempo 133. Puzzel.
    Apple probably didn't test or care that OldWorld systems and 10.3.0 was rushed out the door, would treat REAL SCSI controller and drives as "ATA" and yet, if you used Drive Setup 2.07/2.1 to format in OS 9, you might be safe. I saw this with ATTO and Sonnet Trio. I got around it on ACard by using Drive Setup.

    I've been trying to build a custom install CD or DVD with updated extensions to SCSI and the rest. I've got Install disks 1-3 on the DVD working, but not the extensions.

    You can install, clone to another drive, clone back etc. to get around it of course. 10.3.2 seems to see SCSI as SCSI and ATTO had some better drivers by January or so, what they had back in September gave me the "what were you guys thinking?"

    /rant on/

    Apple has this way of changing things and not communicating. And vendorrs learn that Apple makes last minute changes that invalidate all their earlier work, so they learn to not put pencil to paper - it is more work to fix and rewrite than to just start fresh ONCE YOU KNOW WHAT THE NEW RULES AND GAME PLAN IS. Or reverse engineer or look at the kernel documentation. And Panther is built on new kext and kernel, new compiler, which necessitated updating and new drivers for anything 3rd party. It takes time, also, to master new code and new compiler routines. Anything written for a G5, to run well and take advantage of it, has to understand the intricacies of load and feth routines and how to design or redesign their program. Working under Apple for the Mac platform is not easy I suspect, even if the small market is enough to sustain a business, and it might, but when it constantly changes as it has for the last three years...?

    /rant off/

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

    Default

    Wizzard,

    First, as TZ sez, you are a wizzard in the house. I've just been lurking but am totally impressed with your attention to the last detail and thorough troubleshooting. Great work.

    Some miscellaneous ramblings. I have a PowerTower Pro with 6 PCI slots and did many of the same things with it that you are doing. I did try the XLR8 G4-450 with SSE and it was a complete dud, essentially because of the software at that time. I know that Daystar worked on that but I'm guessing that they have given up on it. Some at the time I tried it were successful with the SSE, but only at the cost of deinterleaving RAM and maybe other things I don't recall. But I was not willing to deinterleave since I had eight 128MB RAM sticks (PTP has 8 RAM slots) and I did not want to halve my RAM. So I junked the SSE.

    One of the Gurus moderators, Trag, modded my PTP mobo with the Kansas 9600 ROM chips, so that my PTP would have native speculative processing in ROM. That was a definite plus to me but the mod does have a minus, a pretty well known slight decrease in RAM thruput which has always afflicted the Kansas 9600 and one Kansas 8600 model later machines. If I recall correctly, it was first noticed in a bit slower video thruput.

    What did totally solve the 6 slot problem for me was the Sonnet G4-800 upgrade. So when I was done with that and jumping thru hoops to get both OS9.1 and 10.2.x (I think it was) installed, I had a good 6 slot machine and spec processing. The PTP is in deep storage now. No time for it. k

  8. #8
    Join Date
    Apr 2000
    Posts
    693

    Default

    Yo, kaye beat me to it, but I was going to say... Reading your posts and following your troubleshooting logic almost gave me chills. It was like reading kaye's original benching of LVD devices that earned him the Sioux name "Recursively-Tests-SCSI-Settings."

    Way to go, Wizzard, and welcome to a place where your tenacity and hankering for horsepower will always be respected. As a fellow six-slot owner, I know the feeling of ripping through PCI cards until something is isolated, endless reboots, and suspect RAM.


  9. #9
    Join Date
    Jan 2004
    Posts
    237

    Default

    Thanks, Fellas. I consider it high praise indeed, coming from the Gurus.

    I’ve had this MacGurus button in my browser bar for a long time. Apparently it wasn’t wasted.

    John heard from Powerlogix yesterday that the “750FX 900 G3 zif” was “not supported” in the Carrier card. He was asking about the 750GX 1 Ghz. “Not supported” can mean “won’t work.” But, it can also mean “we don’t want to answer questions about you’re using it in that manner.”

    If they mean the latter, I can understand that, because you are sticking another mfgs product in the works. If they are saying that the 900 won’t work, on a Carrier card, that would be false, I think. At least from the information I’ve seen.

    The real question for John is whether he wants to lay out $379 worth of mowed lawns to find out if the GX will work. Could get the Sonnet Crescendo - a G4 @ 800 - nearly as fast and very much supported. As Kaye indicated - not going to have the PCI slot problems. It’s academic at this point because they don’t have the chip to ship.

    His call.

    The subject 9600 is doing well enough that I re-tethered it to it's complex of midi cables and interfaces upon which I run Performer 6 and playback through a absurd rig of filters, synthesizers, expanders, etc., rack mounted, accumulated through the years - don’t even want to know how much I’ve got it all that, but I could not start over. That has to be done in 9.2.2 because Performer 6 is a classic app and 2) 10.3 does apparently does not recognize the modem port as a valid midi input. Either that or I've missed something.


    Again, I apprecate the kind thoughts.

  10. #10
    Join Date
    Dec 2003
    Posts
    13

    Default

    Quote: "10.3 does apparently does not recognize the modem port as a valid midi input"

    It won't in X; you'll need to use a USB midi interface, or a USB or Firewire device with midi interface integrated-like some of the M-Audio audio/midi interfaces and controllers. I got a Griffin G3Port for my B&W, and then a G4Port for my G4; my old serial midi interfaces work with these in X, but a basic USB interface isn't much more than one of these. The Serial>USB adapters(Keyspan, Belkin) won't work in X, either-no working drivers- but they do work in 9.x on machines without serial ports.

  11. #11
    Join Date
    Jan 2004
    Posts
    237

    Default

    Thanks HJS,

    I've seen the M-Audio stuff on-line. Don't know much about how well it works. My tendancy would be to go for firewire because of the higher transfer rates, but this machine may be capable of USB 2 now, haven't really investigated that yet. But I know now that the serial ports won't work.

    I can upgrade Performer to a X verson of Digital Performer - but that would mandate one of these interfaces at the same time. I'm headed there. 10.3 is such a pleasure. Things work there gracefully, that seem to struggle in OS9. DVD player, for example. I find myself tireing of 9. I've got Aqua on the brain.

    I did want to also mention what a great tool XPostFacto has been in this ordeal. I know on there website there are lots of people with lots of troubles getting it to work, but I found that it performed flawlessly. I didn't. But it did. I swear you could install OSX on a peanut butter sandwich with it. But then, as always, you're left with the problem of getting it to work after it's installed. That is the rub.

    TZ, I have not been able to find a driver that will allow me to install X on a partition larger than 8GB in the 9600. I formatted the disk in OS10.3.3 and it still wouldn't. I've gotten the notion to stick the disk in the DigAud and install 10.3, that would bypass the problem. I can then put it back in, repair permissions and let XPostFacto install it's patches and so forth and then see if that will work. I'll post the results.

    Later

  12. #12
    Join Date
    Jan 2004
    Posts
    237

    Default

    It does appear that it is a viable method of getting around the 8GB “requirement to format and install in a machine that correctly recognizes the type of disk. Then move it over. I guess there is something about the 9600 that registers an IDE drive (SATA as well, I notice) such that the 8GB partition is mandated. I tried drivers installed in 9.1 9.22, 10.3.3 and Intech to no avail.

    There appeared to be a host of permissions to correct before booting, if this method is used. I had to have a drive with X on it in the receiving machine so that permission could be repaired. Without that step, XpostFacto is unable boot the newly installed X. Haven’t fully checked everything out yet.

    There was no problem installing iLife on the 9600. Certain features will not work because of the limitations of the hardware (no DVD drive yet) and the processor is not fast enough for some of the features of Garage Band, but I don’t have a midi input device that it would recognize at the moment, as was previously mentioned.

    I did note that iDVD was not going to write to firewire or USB bussed drives - apparently Apple set that up so that it requires a “Superdrive” installed on IDE bus. I'm not too happy with the Sonnet Tempo cards performance. Ricks is sending me a IDE to SATA bridge. We'll see what I can do with that.


    Will post later.

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

    Default

    SATA drives ARE ATA and do have the limitation. ATA drives on some Sonnet and most ACard controllers, when formatted with Drive Setup 2.07+ should work. 10.3 has a bug so that SCSI is seen and treated as ATA.

Posting Permissions

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