9500 Bus speed question

[mac2geezer]

For the PPC gurus here: Is the 9500 bus speed controlled by the CPU card or is there another oscillator? I can't seem to find any osc cans on some of the MB's here, hence the question.

Another, perhaps related, question. I have a 9500/132 with a Sonnet 800MHz G4 card and a 10/100 ethernet card, among others. When the 9500 is connected to the LAN via the 10/100 card through a 10/100 switch and I attempt to transfer files to the 9500, after a variable length of time the 9500 freezes solid. If the 9500 is connected via the built in 10B/T port everything works great. Is it possible that the bus can't keep up with the PCI card and thus freezes the machine?

 

[Temetka]

The PCI spec is very fast.

I would gather that you have an extension conflict causing your machine to lock up during file transfers.

The bus speed of the 9500 is 50MHz IIRC. I do believe that it is controlled via on oscillator on the mobo and not on the CPU card itself.

 

[mac2geezer]

Thanks. I had a few extension conflicts that were causing other problems and Conflict Catcher found them, but it now says there are no conflicts. So I dunno. But you may well be correct; this machine has over 150 extensions.

Mactracker lists 40 Mhz as bus speed for some 9500's and since this is a 132, the slowest model, I figure the bus must be at 40. I'll look some more but sure can't find anything that looks like an oscillator can on the other 9500 MB's I have here.

The LAN transfers seem to work using the PCI card if the files are short, but try moving a folder with multiple files and freezup. A different card gives the same result.

 

[Temetka]

bus: 40-50 MHz, controlled by CPU card

Color me corrected then.

 

[Bolle]

its controlled by the CPU card for sure.

the fastest bus speed my 9500 was able to boot and run at with interleaved RAM is 62MHz - but it only did work with hardware disabled onboard cache and was kinda unstable sometimes.

 

[trag]

Yep, controlled by the CPU card. And...

The CPU card must have a Clock Buffer/Distribution chip on board to split that clock up and send it out on five or six pins on the CPU card. Those clock pins are then routed to various devices on the motherboard.

There are also three Clock ID pins on the CPU card which are grounded or not. The various combination tell the Hammerhead controller what bus speed range the CPU card falls into so that it can adjust some MB timing parameters appropriately.

Let's see, pinout for the Mac CPU card is here:

http://www.io.com/~trag/x500_CPU_Pinout

Hmmm. That Sonnet card is supposed to solve the six-slot PCI Mac arbitration problem with G4 processors, right? So that's probably not the problem.

I would consider trying a different ethernet card. The ones with the RealTek chipsets on them seem pretty good.

If you want to try something really fun, go to compuvest.com and get the USB2, FW, gigabit e-net card by SIIG, which they have for $30. I'm not guaranteeing it will work, but iIve seen reports that it does. It uses a TI chipset for USB, NEC for Firewire, and RealTek for E-net (might have the first two backwards) so all the chipsets are well supported on the Mac. The only issue might be that the first generation of PCI Macs have some bugs when dealing with multi-fucntion cards.

 

[trag]

Let's see, pinout for the Mac CPU card is here:
http://www.io.com/~trag/x500_CPU_Pinout

URL has changed to:

http://www.prismnet.com/~trag/x500_CPU_Pinout