RE: Power Managment

[Eric Rehm <eric@equator.com>]

David,

Thanks for the update.  A few comments...or "encouragements" to the PCI SIG and Microsoft:

With repsect to Win95 (or should I say, Win9x?) I find the information in the PC '97 Hardware Design Guide to be even less informative than the WinHEC notes for PCI.

*  FYI, there is no "power management paper" on the PCI-SIG web page in the additional documents section.  

* Windows Driver Model support for PCI will probably be non-existent for Win95. Therefore, for Win95, PC '97 Chapter 2 ("OnNow and WDM for PC 97") really only applies, in the 1997 time frame, to USB and IEEE 1394.  (Admittedly, this is useful information for the next release of Windows NT, where (of course) WDM will apply to PCI.)

  However, where does this leave Win9x?  The VPOWERD VDD?  APM V1.2?

* Turning then to PC '97 Chapter 10 ("PCI"), we find a couple underwhelming paragraphs:

	-  "Compliance with PCI Bus Power Management Interface Spec" is *required*;
	-  "PCI bus implementations must support all four device and bus power states,
	   as well as the standard wakeup mechanism";
	-   PCI devices must "at minimum" "support the D0 and D3 power states"

  D0 and D3 can easily be defined as "On" and "Off".  I interpret this to mean that little or nothing is required of the PCI hardware, and (depending on the device class, O/S, and driver model), that this is save/restore context problem for the device driver.

* The "Power Management Reference Specifications" for audio, display, and communications device classes  do not adequately address the coming gaggle of multifunctions DSP-based multimedia cards (e.g., Chromatic, Nvdia).  These Cards/devices/DSPs have a lot of state (and may have an RTOS that needs to be rebooted too.)  At a minimum, Device Classes should be added for Multimedia (video codecs) and Input devices (joysticks). 

/eric rehm




----------
From: 	David O'Shea
Sent: 	Tuesday, September 24, 1996 9:54 PM
To: 	Eric Rehm; Mailing List Recipients
Subject: 	Re: Power Managment


First, the WinHec '96 notes are somewhat out of date in terms of what 
Microsoft is expecting.  A better (more current) and accurate view is
found in the Microsoft PC '97 Design Guide 1.0.  (This is available from
Microsoft's Web site www.microsoft.com   - It in the developers section).

At 04:45 PM 9/24/96 -0500, Eric Rehm wrote:
>What is the status of APM V1.2 (or power management in general) and PCI?
>
> The WinHEC '96 notes allude to the use of CLKRUN# on portables (and that
> PC's should use this too).  However, the PCI SIG "Specifications" page no
> longer lists the "PCI Mobile Design Guide" as one of the current documents.

That is a mistake that the SIG is supposed to be fixing.  They are also
supposed to be doing another printing of the Mobile Design Guide so that
people can buy them again.  As far as the CLKRUN# signals usage goes, it
is basically still just for portables.  This is somewhat more clearly 
presented in the PC '97 Design Guide from Microsoft.

>
> Further, the WinHEC '96 notes mention the need for the PCI SIG to add
> definitions for power managment.
>

The power management functions that Microsoft is talking about are the
still not finalized power management functions being driven by Bruce Young
of Intel.  This is a current SIG working group project that is trying to
define the power states for PCI device and the additional hardware and
signalling issues and signal pin needed to implement power management in
PCI.  The power management paper is available on the PCI-SIG web page in
the additional documents section.   This is still in the white paper++ 
stage.   It is not a PCI SIG finalized item yet.   (Despite this, Microsoft
has specified this unfinished hardware methodology as the PCI power
management H/W solution that they want to see in PC '97.)

Regards,

David O'Shea
daveo@corollary.com
Principle S/W Engineer
Corollary Inc.
(714)250-4040


-