Conference mvblab::alphaserver_4100

    A01 revs should work, but you should make sure that a fix has been
    implemented: cut of PIN 13 on E3 (23-239Y3-00) AS described in
    Waiver # KZPDA-001 (which I don't have).
    
    Also sounds like something else is wrong if you are killing boards just
    by putting them in a particular slot.  Not sure what is could be if the
    Saddle has already been replaced.
    
    The other caveat is only one KZPDA per IOD (PCI segment) is supported.
    
    Todd

    Questions:
    
      * What version(s) of SRM console firmware were on the system? i.e.
        for both the initially-shipped system and for the replacement PCI
        motherboard?
    
      * Can you provide an SRM console SHOW CONFIG display? Most
        importantly, what else is in IOD1's PCI bus? Any secondary VGAs?
    
      * Just to confirm, you are saying you've seen two KZPDAs die as the
        result of placing them in PCI1's bus 0 slot 3, each on a different
        PCI motherboard? And a KZPAA works fine when in the same slot?
    
      * Have you tried the KZPDA in PCI1 alone; i.e. with no other devices?
    
    No ideas at this point. Would've suspected the Saddle myself. It should
    be noted that spares are only Stage I tested and are not Stage II
    tested, so it is possible you have a bad PCI motherboard spare. Still, 
    to have seen the *same* symptoms re: slot bus 0 3 with two different 
    PCI motherboards would seem to indicate that the PCI motherboard is 
    definitely not the problem here.

Thanks for quick replies....


Re .2
=====
I can see a cut pin on the chip in question, but I am wondering if 
Manufacturing have not cut the correct one!



		Chamfer ------>	  3  2  1 20 19
				4       o      18

				5              17

				6              16

				7              15

				8              14
				  9 10 11 12 13

The above is my assumption (right or wrong) of the pin numbers on the chip, 
with pin 1 being defined by the "o" in the encapsulation, as viewed from 
above. The chamfered corner is between pins 3 & 4 as drawn.
If the above is correct then all my Rev A KZPDAs have pin 14 cut, not 13!

Can anyone confirm this?
========================

All 4 spares are from the same batch, origin Kanata, & are new build (not from 
a repair centre) so they could all be wrong!




Re .3
=====
SRM was initially 3.0-10. Spare Saddle (PCI motherboard) was shipped at 2.0-3 
and I have now upgraded to 3.0-10.



SHOW CONFIG is:- (typed in long-hand so format may be out)

BUS0	IOD0	(PCI0)
Slot	Option			Type		Rev		Name
1	PCEB			4828086		0005		PCEB0
2	S3 trio64/Trio62	88115333	0000		VGA0

BUS1	PCEB0	(EISA Bridge connected to IOD0 slot 1)
Slot	Option			Type		Rev		Name
			<nothing>

BUS0	IOD1	(PCI1)
Slot	Option			Type		Rev		Name
1	NCR53C810		11000		0002		NCR0
2	DECHIP21140-AA		91011		0012		Tulip0
Slot 3 is where the KZPDA should be!



I've actually got 4 dead spare KZPDAs as a result of this (2 "blown" 
in each motherboard, PCI1 Bus 0 Slot 3), plus the dead original. The KZPAA 
however, and the DE500 for that matter, works in all slots. (Note all the 
spare KZPDAs have the modification as above ... so maybe they are all 
incorrect?)




Tried KZPDA initially in PCI1 slot 2 (nothing else in PCI1) and seemed to be
OK. The console came up OK and sized the device correctly. Move it up one to 
slot 3, and dead.




My opinion/plan now:-
1) I am waiting for 2 more KZPDAs to arrive .... I will check them for rev 
level etc.. If they are Rev A then I feel I need confirmation of the 
"pin cut" before I try one in the system. If they are Rev B then I will risk 
one immediately.

2) If 1) fails, then:- Replace the horse module (the IOD1/IOD0 bridge on 
which I think the PCI drivers reside), the saddle (PCI motherboard) and the 
KZPDA all in one go.



	Any more thoughts?

				Steve.

    Steve,
    
    In the course of researching whether or not the correct pin was
    cut for the Rev B01 5424739 I found conflicting information as
    to which pin should be cut.  An engineering memo that could have
    been the technical source for the waiver that described the pin
    cut had E3 pin 14 as the pin to cut.  An ECO that mentioned the
    pin cut but actually documented the Rev B module had E3 pin 13
    as the  pin to cut.  Referring to the print set for the KZPDA
    the correct pin to cut is E3 pin 14.  
    
    The Rev A01 module should have a pin cut on E3 pin 14 and the PAL 
    part number should be 23-239Y3-00.  The PAL marking might
    be truncated to 239Y3.
    
    The Rev B01 module should have no pin cut and the E3 PAL should 
    be part number 23-309y3-00.  The PAL marking might be truncated to 
    309Y3.
    
    If the incorrect pin is cut there should be no adverse effect 
    on the module or the system.  Pin 13 is the signal ~PCI_INTA
    and pin 14 is ~PCI_INTB.
    
    Your pin counting assumptions are correct.  Of course we 
    would like to get this issue resolved.
    
    
    Regards,
    
    Dennis H.

Re .4
	Thanks for the detailed update, I can therefore confirm that my
KZPDA modules at revision A01 are built correctly, with pin 14 cut on the 
correct PAL.

	I have now got the system working ... the B3040-AA (horse) appears 
to have been the culprit. This was "blowing" the KZPDAs when they were put 
into IOD1 (PCI1) bus 0 slot 3 only!

	I do not sufficiently understand the PCI to explain this, or why other
devices in that slot were undamaged. (The KZPDA is a PCI-PCI bridge device, 
the others were not, probably relevent somehow!) The PCI motherboard B3050-AA 
was undamaged, the original being back in the working system.

		Again, thanks for your inputs. As this is a new product from 
the MCS perspective we are only now starting to see any volume of faults. 
Thus we are on the learning curve for the "quirks" of the products failures.


				Steve.