Re: Intel 845PE and 'idebus=66'?

From: John-Paul Stewart (jpstewart_at_binaryfoundry.ca)
Date: 05/20/04 

Date: Thu, 20 May 2004 11:13:33 -0400

Aaron Fox wrote:
> Greetings.
>
> I have a question I hope the hardware gurus can answer. I've been
> reading several different guides on tweaking the maximum amount
> of performance out of UATA100/133 hard drives.
>
> I see a common suggestion that I should append:
> "idebus=66"
> to lilo.conf, on the basis that the Linux 'IDE Bus' runs by default
> at 33 MHz.
>
> Now, after looking at the Intel specs for the chipset in my laptop (Intel
> 845PE,which runs the ICH4 I/O chipset), I see that the IDE controller
> (along with the USB, PCI/PCMCIA, and peripheral controllers) all run
> under the same controller chip (the ICH4), which runs at 33 MHz.
>
> My questions are:
> 1) Is passing the "idebus=66" actually attempting to double the
> clock speed of the PCI bus controller? If so, this seems to me a
> Really Bad Thing in a laptop????
>
> 2) Or is passing the 'idebus=66' actually related to the burst
> transfer speed of the hard drive itself (i.e. the 66 Mb/s or 100
> Mb/S of UDMA Mode 4/Mode 5). If so, should I be appending
> 'idebus=100' to reflect the actual burst speed of the drive?
>
> I guess what it comes down to is that I don't fundamentally understand
> what parameters are being affected by passing 'idebus=66; to the kernel at
> boot time. Can anyone enlighten me on the hardware side?

Passing the "idebus=66" parameter will make the kernel think the IDE
controller is _on_ a 66MHz bus (i.e., the kernel will think it talks to
the IDE controller over 66MHz PCI). It will _not_ attempt to change the
speed of the hardware, only the kernel's idea of what that speed is. It
has _nothing_ to do with 66MB/sec (or higher) transfer rates over the
IDE bus. So unless you're using, say, a 40MHz VESA local bus (common in
the 486 days), you shouldn't need to change the default since nearly
every current system has the IDE interface on a 33MHz PCI bus. As
you've already looked up the specs for your system, we know the default
is correct for you.

Moreover, the message printed by the kernel says:

"ide: Assuming 33MHz system bus speed for PIO modes; override with
idebus=xx"

Notice how it explicitly says "for PIO modes". The message tells you it
has nothing to do with DMA---the timings are only needed for slower PIO
modes. So the "idebus=" parameter is only relevant for unusual hardware
with something other than a 33MHz bus clock when using PIO (i.e.,
non-DMA) modes.

Relevant Pages

  • Re: IDE/DMA-related kernel panic in 2.4.20-6, but 2.2.16 is fine
    ... and the IDE is on a bus running at 1/3 that speed. ... your hardware is of a vintage such that it doesn't surprise me that IDE ... Telling the kernel it is at 33MHz will just ...
    (comp.os.linux.hardware)
  • DMA hangs IDE bus (cdrecord or driver?)
    ... The primary IDE bus is unused, ... 0000:00:02.0 PCI bridge: ATI Technologies Inc RS480 PCI-X Root Port ... Initializing USB Mass Storage driver... ...
    (Debian-User)
  • Re: ANSI C question about volatile
    ... The *intent* of the C Standard is clear: the hardware has some ... Only one "bus width" is supported, ... the register numbers might change in the process). ... programmer might use the "bis" or "bic" instructions on the VAX: ...
    (comp.lang.c)
  • Re: [PATCH 29/30] W1: Documentation/w1/masters/ds2490 update
    ... DS9490is a USB W1 bus master device ... +- While the ds2490 supports a hardware search the code doesn't take ... a write buffer and a read buffer as arguments. ... the bulk read will return an error and the hardware will ...
    (Linux-Kernel)
  • [PATCH 32/33 replaces 34/35] Documentation/w1/masters/ds2490 update
    ... DS9490is a USB W1 bus master device ... +- The hardware will detect when devices are attached to the bus on the ... a write buffer and a read buffer as arguments. ... the bulk read will return an error and the hardware will ...
    (Linux-Kernel)
Loading