Strongly recommend you read VPD ECR ASAP

[Paul Kalapathy <paulk@chromatic.com>]

Sorry for the acronym run-on :-)

I think that like many people, I have gotten lax about reading ECRs on
the assumption that PCI was a stable bus design.  Most of the recent
ECRs have been beneficial or innocuous.  I just got around to reading
the VPD ECR, and it does not, in my opinion, fall into that category.

The VPD ECR will require all PCI designs to include an EEPROM or be
non-compliant.  There does not appear to be any real benefit to this
cost adder.  I suggest that you read the ECR and comment to the SIG if
cost is a concern for your products.  Unfortunately, the comment
period ends on Monday, 6/30/97.

For what it's worth, I've included our comments on the ECR below.
	-Paul

--------------------

Dear SIG members,

Chromatic Research objects to the "VPD required" ECR to the PCI 2.1
specification on several grounds.

First, there are many successful systems shipping today without any
VPD support at all.  It is clear that systems can be effectively
configured using other mechanisms.  Mandating an unnecessary mechanism
that increases product cost is not in anyone's interests.

Second, in systems where specific configuration information is
required, Microsoft has already mandated the use of the variable
Subsystem Vendor ID field.  This will ensure that virtually all
systems shipping by 1998 have the capability to discriminate various
hardware versions based on the same silicon.  VPD is not needed for
this.

Third, product-specific variable data that needs to be updated and
survive power cycles can already be stored in product-specific ways
using system non-volatile resources (e.g., disk drives) that do
not increment the product cost.  VPD is not needed for this.

Fourth, the time-frame for compliance is far too short.  The ECR
requires devices shipped by June 1998 to comply with the ECR.
Products that are already in silicon and shipping will still be
shipping in volume at that date.  The vendors will either incur costs
of redesign (which may not be possible in the existing pinout) or will
lose business due to their non-compliance.  To render them
non-compliant ex post facto is to impose unacceptable costs on
hardware vendors.

Everything that can be done under the new VPD ECR can already be done
using other mechanisms.  It benefits a few companies and imposes costs
across the industry.  The ECR is redundant and costly.  It should not
be approved.

	Paul Kalapathy			paulk@chromatic.com
	Director, HW Design		408-752-9411
	Chromatic Research, Inc.





Ú¤
”