RE: AMCC S5933 and Master Abort

[Neal Palmer <neal@dinigroup.com>]

For this errata, there are a bunch of work arounds that weren't mentioned
in the errata sheet:

1) Don't let NT get idle.  When all NT processes become idle, it sends out
special cycles (and I'm pretty sure that it is the only time that it does
generate them).  So make a program that does basically:

int main() {
  while (1) 
	sleep(0);
}

It will keep the processor busy doing context switches when no other
program wants to use the CPU.

2) The problem occurrs when a master abort occurrs on PCI (special cycles
always end in a master abort).  And only when the s5933 is the next bus
master on the bus.  You could gate GNT from the s5933 every time that
FRAME is asserted, until you see the next TRDY/STOP.

3) "no factory respin".  Oops, they need to update their errata sheet.
The s5935 IS A RESPIN of the part.  Functionally identical except that
this bug has been removed, and IO timing might be a little bit different.
The s5935 should really be named a new version of the s5933...

4) Contact the App Engineers in San Diego to get a better list of
work-arounds.  I think that there are 2 or 3 that I'm missing.

--Neal

On Tue, 2 May 2000, Tom Warren wrote:

> Is this the one?
> 
> - Tom
> 
> >I was also having problems using a PCI board with the AMCC S5933QE in a
> >newer computer when using bus master transfers. It turns out that there is
> a
> >"bug" in the S5933QE that is activated by Special Cycles on the PCI Bus
> where the IDSEL signal is active during the Special Cycle. This is
> documented in http://www.amcc.com/pdfs/SumQE4.pdf. Apparently the chipsets
> in the newer computers are aggravating this bug. AMCC has a new part, the
> >S5935QF that is a drop-in replacement for most applications. It took care
> of
> >the problem for us.
> >
> >Regards,
> >George Reasoner
> >Unisys Corporation
> >
> >-----Original Message-----
> >From: Allen, Douglas [mailto:Douglas_Allen@perkinelmer.com]
> >Sent: Tuesday, April 11, 2000 2:23 PM
> >To: ''
> >Subject: PCI 2.2 question
> >
> >
> >A couple of years ago, I designed a PCI board using the AMCC 5933QE chip.
> >Recently, I found that
> >I was receiving data errors when running the board on some newer computers.
> >So far, the only common
> >thread I see is that all of the computers on which it fails are specified
> as
> >PCI V2.2. Are there known
> >issues when using a PCI 2.1 board in a PCI 2.2 bus? Since I have not been
> >designing PCI cards of late, 
> >I was wondering if there was something I did not take into account with the
> >new spec.
> >
> >Any help would be greatly appreciated.
> >
> >Thanks.
> >
> >Mark Allen
> >Douglas_Allen@perkinelmer.com
> 
> -----Original Message-----
> From: Lame Brooks-G14738 [mailto:brooks.lame@motorola.com]
> Sent: Tuesday, May 02, 2000 2:13 PM
> To: PCISIG List (E-mail)
> Cc: 'Thomas Heller'
> Subject: RE: AMCC S5933 and Master Abort
> 
> 
> I only vaguely recall something about newer host chipsets, Windows NT, and
> unexpected PCI Special Cycles causing unexpected problems with peripherals.
> Can you get the change documents for the AMCC S5933?  Perhaps someone else
> on the list knows more?   -- BrooksL
> 
> > -----Original Message-----
> > From: Thomas Heller [mailto:thomas.heller@ion-tof.com]
> > Sent: Tuesday, 02 May, 2000 11:01
> > To: Lame Brooks-G14738
> > Subject: Re: AMCC S5933 and Master Abort
> > 
> > 
> > > Thomas, if I recall correctly, someone said this is a known 
> > issue with the
> > > AMCC when transfers are interrupted with PCI Special 
> > Cycles?  Did you
> > > contact AMCC yet?  -- BrooksL
> > > 
> > AMCC support (provided through the distributor here in Germany)
> > has not been very helpfull. Their suggestions included:
> > - try another slot for the board
> > - try the S5935 instead of the 5933
> > - ???
> > 
> > Do you have more information about this issue?
> > 
> > Thanks and kind regards
> > 
> > Thomas
> > 
> 

-- Neal Palmer

The Dini Group
1010 Pearl St #6
La Jolla, CA 92037
(858) 454-3419 x16
(858) 454-1728 (Fax)