Re: new root and base disks available!
Paul Slootman wrote:
> On Tue 12 May 1998, Wes Bauske wrote:
> > Alexander Jolk wrote:
> >
> > > Loic Prylli <lprylli@graville.fdn.fr> writes:
> > >
> > > > Paul Slootman <paul@wau.mis.ah.nl> writes:
> > > > > Besides that, the kernel on the rescue disk I used didn't autodetect where I had
> > > > > connected the network on my multia (which was coax); luckily I had a
> > > >
> > > > That again is a problem of the kernel, and though we should address it
> > > > either by documentation (maybe a command-line option to the kernel can
> > > > help there) or by generating custom kernel, it is not linked to the
>
> Does the tulip driver accept any commandline parameters?
It does as a module.
>
>
> > Tulip.o does a much better job handling non-DEC Enet cards.
> > de4x5 basically works only on real DEC HW. If you use
> > de4x5, it won't work on my systems and lots of other people's.
> > Tulip is preferred by most people running 21x4x on Alpha,
> > particularly those using 100B-TX.
> > I believe most people using Linux Alpha are running on non-DEC
> > systems. Note MB is a DEC design but, it doesn't usually have
> > a built in Enet which leaves selection up to the buyer.
>
> That may be, but this was a rescue disk for the udb-noname, which only
> comes with the onboard network...
>
You mean it has no PCI slots to put a better Enet card in?? Prettysure UDB's have at least one slot
available.
> offtopic: is it really true that most alpha's are non-dec? Here in the
> Netherlands most linux users with an alpha get it from one supplier, who
> only sells digital stuff at quite low prices (I think; my XLT-300 with
> 1GB disk, cdrom, matrox mystique, 96MB ram was about US$1250).
>
If you listen to the RH alpha list, you'll find almost all users either have164LX/164SX or the cheap
UDB boxes. There are a few oddballs too
but basically the UDB is getting less common. I honestly see no use
for any of the older DEC boxes. My Intel stuff will out-perform
them for most things. Only the 500MHZ+ are able to keep
significantly ahead of them. You can buy a 164SX MB/CPU for $800
here(USA) and build your own system for about the same price you
quote. Think a 533MHZ 164SX is quicker for the money. Of course
if you just like to be different, that OK by me.
I just want fast FP performance and all I look at are the high end models.
Really want to get a 21264... Should be the fastest workstation on the
planet when it gets here, at least for a few months.
> > Your comment that Tulip doesn't do selection as well as de4x5 is
> > incorrect. When you have to cope with a dozen different card designs,
> > the problem is much more difficult than a single manufacturer's
> > products. In general, tulip's weakness is due to the many varied
> > ways one can set up a 21x4x chip on it's programmable pins.
> > 16 ways for each chip set (21040/21140/21142/21143/etc.).
> > And, you have to hope the manufacturer put the correct
> > pin settings in the EEPROM! Quite a few don't.
>
> All the evidence I have is that the de4x5 does selection correctly,
> while the tulip doesn't. So, as far as I'm concerned, tulip does not do
> selection as well as de4x5 :-)
Read what I said above....Try a non-DEC Enet card with de4x5. High probability it won't
work. Last I heard, about 60% of all 100Mb Enet cards were
tulip based. DEC has sold lots of tulip chips to other vendors.
I can buy 100Mb-tx clone cards for $30.
If you want some insight as to how all this works, look at how each
DD setups up it's programmable pins. I read DEC's Tulip HW
reference to fix my non-MII cards in my PPro's. Also, get the tulip
DOS NIC programming kit if you want examples of how to set those
pins.
>
>
> You'd expect that digital had done the Right Thing; apparently not as
> far as the tulip driver is concerned.
>
>
Not really. Most of DEC's stuff are first generation designs.Almost all new tulip adapters use MII.
(Media Independent Interface)
That's the thing that talks to the wire (something called a PHYS chip).
I have one DEC 10Mb tulip card, a 21040 which I pulled out because
all I run are 100Mb-TX. Didn't bother to look at it since it's old/slow.
The old designs use the non-MII entries that describe each
possible type of connection. (10Mb/10MbFD/100Mb/100MbFD)
Also, some vendors used non-conforming entries in their EEPROMS.
By that I mean they ignored what DEC recommends in their HW
reference manual to put in the EEPROM. Tulip has to be modified
to handle each non-conforming adapter. If the adapter vendor follows
DEC's recommendations, it will work with tulip with no changes.
That means the DEC adapters don't follow their own standards.
> > If you use MII , tulip works great. When you have cards
>
> Please explain MII? This is the first I've heard of that...
>
It's all related to N-WAY auto negotiation. Most new hubs/switches use it to determine capabilities
of the other side of
the wire. Even the non-MII stuff does it with the help of the
tulip DD, but it works slick with MII cause the DD doesn't
care what's going on.Wes
--
To UNSUBSCRIBE, email to debian-alpha-request@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
Reply to: