Boca BEN1PI PCI Ethernet card - STRIKE FOUR!

Boca BEN1PI PCI Ethernet card - STRIKE FOUR!

Post by Dave Pla » Sat, 17 Jun 1995 04:00:00



AARGH!  I've reached the point of complete boilover.  Please pardon me
while I vent my frustrations, and issue a warning.

Back in September, I bought a Boca BEN1PI PCI Ethernet card (a "Lance"
type, based on one of the AMD chips).  I have been having trouble with
it ever since - it's sensitive to certain data patterns, and discards
packets with a "FCS error / packet overrun" when they contain certain
data patterns that tickle the problem.  The card "mostly works", but
locks up solidly when attempting to transfer certain files.

This is a design problem with this particular card.  Boca ignored one of
the notes in Appendix A of the AMD 79c970 chip data sheet, and omitted a
fairly critical .1 uF power-supply bypass capacitor across a couple of
power-supply pins that AMD warns can be quite noisy.  This omission
probably saved $.15 on the cost of the card... and has caused problems
for many users.

Boca has an ECO (engineering change order) to "repair" the card... they
solder a surface-mount capacitor to one pin, and run a small wire over
to the other.

I sent my card back for repair in February.  They didn't repair it -
they simply sent me a replacement card, which did not have the ECO
installed and exhibited exactly the same problem.

I sent the card back again in March, including a note indicating that
installation of the ECO was essential, and making sure that this
requirement was entered into their RMA service-tracking computer.  The
card was sent back to me with the ECO installed, and it appeared at
first that the problem was fixed.  However, a couple of months of
extensive use demonstrated that the ECO was not 100% reliable - certain
files could still trigger the "FCS error / overrun" error and lock up
the file transfer.

I had heard on the Net that Boca had redesigned the card to correct the
defect, and was shipping corrected cards.  I called their tech-support
line again last month (another 30-minute hold, on a non-toll-free line),
discussed the problem, and received another RMA number.  I informed the
tech that I wanted this card replaced by a redesigned card which did not
have this problem.  I was assured that this would be done, and that this
need was entered into the computer.  I also included this request (well,
a bold-faced 12-point-type demand) in a note I included with the card
when I mailed it back.

I got a package from Boca today.  New card, still in the shrink-wrapped
box.  Guess what?

Right.  They sent me another of the ORIGINAL model (no on-board bypass
capacitor), WITHOUT the ECO installed.  It exhibits EXACTLY the same
problem that the first two cards did - it is MUCH worse than the one I
returned.

Worse yet, it has a visible defect.  One of four surface-mount resistors
feeding the ThinNet transceiver isn't soldered down right - it's
sticking up in the air, with one end* in space.  As a result, the
ThinNet interface is only partially functional - it completely misses
many packets.  [It's amazing to me that it works at all.]  The automated
pick&place machine screwed up, and Boca's alleged quality control
process completely missed this defect.

I am REALLY annoyed.  I'm quite sick of dealing with Boca about this.
I'm going to take the card back to the store from which I bought it, and
demand a full refund - I have no confidence that any replacement card
would be any better.

My warning:  AVOID this Ethernet card at all costs.  It is not reliable,
and Boca seems unable or unwilling to do a professional job of
correcting the problem.
--

      USNAIL: The 3DO Company, Systems Software group
              600 Galveston Drive
              Redwood City, CA  94063

 
 
 

Boca BEN1PI PCI Ethernet card - STRIKE FOUR!

Post by Jos V » Sat, 17 Jun 1995 04:00:00



>Back in September, I bought a Boca BEN1PI PCI Ethernet card (a "Lance"
>type, based on one of the AMD chips).  I have been having trouble with
>it ever since - it's sensitive to certain data patterns, and discards
>packets with a "FCS error / packet overrun" when they contain certain
>data patterns that tickle the problem.  The card "mostly works", but
>locks up solidly when attempting to transfer certain files.

Yep.  Try, for example, to start a remote xv 3.10 with the X11 DISPLAY
set to a system with a Boca card.  In most cases, the xv picture
never appears (no problems with xv 3.01, however, the card just doesn't
like the new xv fish :-().

In previous versions of the card I had the same problem with *any*
remote Netscape.  Now most of those sessions work, but it's still
very unreliable.

Quote:>I had heard on the Net that Boca had redesigned the card to correct the
>defect, and was shipping corrected cards.  [ ... ]

I have two questions related to this:
-  Are the newest cards *really* OK, or are there still some problems.
-  How can I recognize the newest cards (see also below)?  I heart
   something about a driver diskette labeled 2.15, but I want to identify
   the card itself, not the floppy.  Is there a unique number or other
   recognizable issue on the cards?

Quote:>I got a package from Boca today.  New card, still in the shrink-wrapped
>box.  Guess what?

A Dutch Boca distributor got a new bunch of cards a few weeks ago.
These cards work better than the old ones, but still have severe problems.
Then I heart that newer (the newest?) cards could be recognized by
a driver floppy labeled 2.15.  But: I got cards with a driver diskette
version 1.1.  So they seem to have sent a bunch of hundreds of old(er)
cards to my distributor just a few weeks ago.  Arrgghh...

--

--    X/OS Experts in Open Systems BV   |   Phone: +31 20 6938364
--    Amsterdam, The Netherlands        |     Fax: +31 20 6948204

 
 
 

Boca BEN1PI PCI Ethernet card - STRIKE FOUR!

Post by Dave Pla » Sat, 17 Jun 1995 04:00:00



>I have two questions related to this:
>-  Are the newest cards *really* OK, or are there still some problems.
>-  How can I recognize the newest cards (see also below)?  I heart
>   something about a driver diskette labeled 2.15, but I want to identify
>   the card itself, not the floppy.  Is there a unique number or other
>   recognizable issue on the cards?

At this point, I have no first-hand evidence that there _are_ redesigned
cards.  There may very well be some, but the card I got appears to have
the same specific design deficiency of the older cards.  This card has,
on the back, the legends "PN 4186", "PCA-1-0-94V-0", and "0495".

This last suggests to me that Boca was still manufacturing cards without
the correct bypass chips as late as April of this year.

Quote:>A Dutch Boca distributor got a new bunch of cards a few weeks ago.
>These cards work better than the old ones, but still have severe problems.
>Then I heart that newer (the newest?) cards could be recognized by
>a driver floppy labeled 2.15.  But: I got cards with a driver diskette
>version 1.1.  So they seem to have sent a bunch of hundreds of old(er)
>cards to my distributor just a few weeks ago.  Arrgghh...

Double aargh.

I just reviewed Appendix B in the 79c970 data sheet.  It recommends a
_minimum_ of 8 high-frequency bypassing caps (.1 uF multilayer ceramic)
around the chip, to keep ground and power bounce from causing unreliable
operation.  Looking at the card I have here, I see a total of 5 such
caps on the entire card - only a couple of them are close enough to the
chip to do a decent job of high-frequency bypassing.

If you hold the card with chips facing up, and the PCI pins pointed
towards you, the lower-left corner of the '970 has the positioning
dimple.  The upper-left corner of the chip is the "analog corner", where
the analog power and ground pins are.  Appendix B calls these "the most
cricical pins in the layout of a PCnet-PCI card".  There are 4 analog
power pins, and 2 analog ground pins... all of them are supposed to be
connected, and properly bypassed.

On the left side of the chip, the uppermost pin is AVSS1 (analog ground
#1).  On the top edge, the fourth pin from the left is AVDD3 (analog
power #3).  The Appendix specifically recommends having a .1 uF bypass
cap tied directly to these two pins (_not_ going through the common
ground plane).  On the card I have, there is no such cap - the nearest
bypass cap is about half-an-inch away over by the crystal, and is tied
to the ground plane.

The Appendix also specifically recommends "low-frequency bulk
capacitors" (by which I assume that they mean multi-uF tantalum or
aluminum electrolytics) as well as high-frequency bypass caps.  I see
only two bulk capacitors (one 10 uF in the AVSS2/AVDD2 filter circuit,
and one 4.7 uF up above the chip which also appears to be part of a
filter circuit).  The Appendix recommends "at least one low-frequency
bulk (e.g. 22 uF) bypass capactor... connected directly to the power and
ground planes."  There is no capacitor on the board which matches this
description.

It appears that Boca ignored _several_ of AMD's recommendations,
regarding the number, size, placement, and wiring of the power supply
bypass capacitors.  I don't know whether this was the fault of the
layout designer, or whether management put pressure on the designers to
eliminate "unnecessary" components in order to keep the build cost to a
minimum.

My hunch is that this board design is flawed beyond the ability of a
"quick fix" to correct reliably.  The card needs to be redesigned.

--

      USNAIL: The 3DO Company, Systems Software group
              600 Galveston Drive
              Redwood City, CA  94063

 
 
 

Boca BEN1PI PCI Ethernet card - STRIKE FOUR!

Post by Dave Pla » Sun, 18 Jun 1995 04:00:00


Quote:>I was wondering, we have this card and a 486DX-4 100 Motherboard
>(ASUS SP3).  I was wondering if this affects 486 machines also?  I've heard
>that it does affect the high speed pentiums.

The data-pattern-sensitivity problem definitely affects 486 machines
(mine is a 496DX2-66 in an ASUS SP3G).

The "card won't work at all" problem has only been reported, as far as I
know, with high-speed Pentium machines.  This _may_ be a problem with
the AMD chip rather than with the Boca PC-board layout (I've heard that
AMD has spun a new version of the chip, the 79C970A, to resolve Pentium
incompatibility problems).

Quote:>Should I be worried as this machine is acting as the schools drive x:
>(Basically, where teachers can save all their files to (SAMBA) from windows).

To quote Dirty Harry Callahan, "How lucky do you feel today?"  If I were
you, I'd look for another solution.

Quote:>I bought three of them, one for our NT server (I haven't experienced any
>lockups with it, (pentium 90 - 32 megs RAM, Asus P54TP4Burst), One for
>our 486DX4100-server (I got it replaced with an SMC etherpower) and one more
>on our Linux Box)..  Does this problem only affect linux?

The problem definitely does not affect only Linux.  I was able to
reproduce it under DOS, using Boca's own packet driver and the PC/TCP
software stack.  The same file, FTP'ed from a Sun, would freeze up both
DOS and Linux (fyi it was /usr/bin/gawk from one of Slackware's
distributions).

It really and truly is a hardware problem.

I returned my card today, and (after a long debate with the store
manager) got most of my money back.  Boca apparently isn't much better to
its dealers than to its customers - she said that all Boca would do for
the store was "repair" the card, or exchange it for another one "not in
a sealed box", leaving the store unable to sell it as new.  She finally
agreed to treat my problem as a return (subject to the same 15%
deduction usually taken as a restocking charge) rather than as a
warranty issue, as I'd already pursued the warranty path and gotten no
satisfaction.  

--

      USNAIL: The 3DO Company, Systems Software group
              600 Galveston Drive
              Redwood City, CA  94063

 
 
 

Boca BEN1PI PCI Ethernet card - STRIKE FOUR!

Post by Jos V » Sun, 18 Jun 1995 04:00:00



>I was wondering, we have this card and a 486DX-4 100 Motherboard
>(ASUS SP3).  I was wondering if this affects 486 machines also?  I've heard
>that it does affect the high speed pentiums.

The problems also occur on 486 systems.  I've ASUS SP3G boards with
486DX2/66 processors, all having problems with the Boca card.

Quote:> [...]  Does this problem only affect linux?

No.  In fact, Boca admits the problems, and they probably wouldn't do that
when only Linux users had problems ....  They only don't seem to be able
to solve the problems adequately.  When I told my distributor about the
problems the first time, he said something about "new drivers" (but he
didn't have a new Linux-driver :-)).  So there may be some patched MS-*
drivers that hack around (some of) the problems.

A more realistic view is that MS-* users are used to things that don't
work properly, so they just don't note the problems of the card.  :-)

--

--    X/OS Experts in Open Systems BV   |   Phone: +31 20 6938364
--    Amsterdam, The Netherlands        |     Fax: +31 20 6948204

 
 
 

Boca BEN1PI PCI Ethernet card - STRIKE FOUR!

Post by Eric Wampn » Sun, 18 Jun 1995 04:00:00


: The "card won't work at all" problem has only been reported, as far as I
: know, with high-speed Pentium machines.  This _may_ be a problem with
: the AMD chip rather than with the Boca PC-board layout (I've heard that
: AMD has spun a new version of the chip, the 79C970A, to resolve Pentium
: incompatibility problems).

Well, I am not that familiar with the issue, but I have a Zeos with the
974 (scsi + network) on the motherboard, 90mhz pentium, and have had
zero problems with it. YMMV

eric
--
---


   ***G'Kar: "Weep for the chicken, Na'Toth. Weep for us all"***

 
 
 

Boca BEN1PI PCI Ethernet card - STRIKE FOUR!

Post by George Daswa » Sun, 18 Jun 1995 04:00:00


: This is a design problem with this particular card.  Boca ignored one of
: the notes in Appendix A of the AMD 79c970 chip data sheet, and omitted a
: fairly critical .1 uF power-supply bypass capacitor across a couple of
: power-supply pins that AMD warns can be quite noisy.  This omission
: probably saved $.15 on the cost of the card... and has caused problems
: for many users.

I was wondering, we have this card and a 486DX-4 100 Motherboard
(ASUS SP3).  I was wondering if this affects 486 machines also?  I've heard
that it does affect the high speed pentiums.

Should I be worried as this machine is acting as the schools drive x:
(Basically, where teachers can save all their files to (SAMBA) from windows).

I bought three of them, one for our NT server (I haven't experienced any
lockups with it, (pentium 90 - 32 megs RAM, Asus P54TP4Burst), One for
our 486DX4100-server (I got it replaced with an SMC etherpower) and one more
on our Linux Box)..  Does this problem only affect linux?

George Daswani

 
 
 

Boca BEN1PI PCI Ethernet card - STRIKE FOUR!

Post by Jos V » Tue, 20 Jun 1995 04:00:00


The latest news from my distributor:

The set of 200 BEN1PI cards they got from Boca a few weeks ago contained
85 (!) of the oldest cards (the serie that works very bad), and 115 newer
(but not: newest) cards (the serie that works bad).  We've got some of the
latter set, which still don't work ok.

This proves (again) that Boca is not serious about their customers.
And for this reason it seems almost impossible to get the newest model
(*if* that model works ok, which is still not known for sure).

Triple Argh...

--

--    X/OS Experts in Open Systems BV   |   Phone: +31 20 6938364
--    Amsterdam, The Netherlands        |     Fax: +31 20 6948204

 
 
 

1. BOCA PCI: BEN1PI now safe?

As one of the former "victims" of Boca cards (I had 4 of them :-(),
I think I have to follow-up on this question.

First, Boca cards were *also* failing on 486 systems.  The problems
that occurred at my site were data-sensitive (!).  About half a year
ago, I did send them back to my local supplier and I got 4 new ones,
with a different desig.
The problems occurred less frequent, but they still were there.
Some months ago, I asked my supplier if they could send me working
cards.  They said that I had the latest version and I have returned
the cards and got my money back.

Donald Becker has a WWW-page with information about the Boca-porblems.
Unfortunately, I don't have a reference here.

I now have ExpertBoard PCI cards, also with the Lance-chip, which
work fine (that is, I don't see any problems), although the kernel
logs some problems now and then (not frequently).

No.  There are also cards based on the DEC 21040 chip, like the SMC
EtherPower and DEC's own cards.  But, they are probably about twice
as expensive as the relatively cheap Lance-cards...
And there are other cards.

Using LILO (the boot prompt), specify "ether=irq,ioaddr,eth0" (e.g.,
say "linux ether=12,0x280,eth0" using IRQ 12 with I/O addr 0x280).
Or use an
        append = "ether=irq,ioaddr,eth0"
line in your lilo.conf.

--

--    X/OS Experts in Open Systems BV   |   Phone: +31 20 6938364
--    Amsterdam, The Netherlands        |     Fax: +31 20 6948204

2. remote queue time out

3. BOCA PCI Ethernet card

4. im have a problem with cs4236-KQ

5. Boca phone number? PCI Ethernet card

6. two different internet addresses

7. BOCA PCI Ethernet CARD woes

8. Filename for File descriptor

9. Boca PCI Ethernet card

10. Boca PCI Ethernet card problems - more details?

11. Boca Ethernet PCI card problems

12. Opinions wanted about Boca PCI ethernet card

13. Boca PCI ethernet card woes : Solved - Maybe ?