Matt Van Mater wrote:
>>* Antonios Anastasiadis <[EMAIL PROTECTED]> [2005-05-21 11:27]:
>>> Are all the xl-based cards crap without exceptions?
>
>>yes.
>
> While I don't doubt Henning knows much much more than I do about such
> things, this answer doesn't exactly satisfy me. Poor performance on
> xl nics has been discussed many times on misc, but I have never seen
> much beyond 'they suck, use xyz'.
Short version: 3Com has a reputation for quality in some people's mind
which is basically undeserved. Here's a list of most of the 3Com cards
I've worked with over the years:
3c501: technically, probably worst ethernet card ever made, though as
an early player, probably forgivable. I think the $500 price tag fell
off one of the ones I have here a few years back...but that's what we
used to sell 'em for.
3c503: Weird. Freaking. Card. (actually, works ok, but strangest
config of any NIC I've used. Jumpers set IO port and shared RAM, driver
sets IRQ. Really.)
3c505: early attempt at a "server" grade NIC -- on board 80186
processor, lots of RAM, DMA. Worked ok, as long as your machine had no
more than 16M of RAM, otherwise things got...exciting. Never saw a
performance difference, in spite of all the technology and price tag.
3c507: Oh, gawd. "Just plug it in, watch it work!" I'm sure it did
for someone, sure didn't for our customers who were buying it. "What
IRQ is it at?" "Dunno" "What IO port?" "Dunno". Sheesh.
3c509: Worked. Easy software config.
3c509B: Worked Really Well, at least if you turn of PnP mode. OpenBSD
has more trouble than Windows with the PnP mode, but it certainly wasn't
trouble free on Windows.
3c515: 100Mbps ISA NIC, capable of...10Mbps performance. But it does
make the 100Mbps light light up on your switch. Now, I'd not expect it
to saturate the pipe...but it is a serious under-performer.
3c590: Look, ma! 100Mbps PCI! (see the link light?)
3c905: Ok, the 590 was a piece of junk, let's try again. We'll make
it really hard to tell the difference between the 3c590 and the 3c905,
too -- the poor saps will have to dig around with a magnifying glass to
see which card this is.
3c905b: If at first you don't succeed...
3c905c: ...try, try again. BTW: the revisions of the 3c905 are ..
annoying.
3c920, 3c555, 3c{lots}: ok, people are tired of us slapping letters on
the end of a mid-grade card, let's make more confusing numbers. Oh,
each requires an updated driver for Windows. Some are ok, some are
abominations that were put in laptops and delivered performance that
would make a decent ISA card hide in shame...but they DO light up that
100mbps link light, and we know that's all that matters, the users will
never notice the 100kbps performance, or blame it on the laptop.
Not included on this list: 3Com EISA cards who's numbers I've forgotten
and don't care enough to go look up (worked. One appeared to have the
3c509's chip (not the B version). 100Mbps card didn't blow me over with
its performance), and the laptop PCMCIA cards ("worked").
I'm one of the people that used to sing 3Com's praise, until one day I
was reading through the FAQ, saw a cheap shot at 3Com, and started
thinking about my experience and realized my praise was completely
unwarranted. At a certain point, one starts to realize that all the
"bizzare, abnormal cases" were not that bizzare or abnormal, as they
keep happening. The only "best in class" NIC I can think of that they
ever made was the 3c509B, and as the 'B' shows..took 'em a couple tries.
> From my experience, xl performance
> on FreeBSD, NetBSD, Loonix, and Windows* isn't all that terrible and
> I've been at a loss to explain why it is so disproportionately bad on
> OpenBSD.
well, our experience differs.
My experience with the 3c905 is that it usually works ok, but if things
are wonky on my network, they get changed, and problems go away. That's
on a lot of OSs, mostly Netware and Windows.
The MAJORITY of time, yes sure, they work fine. When they go wonky,
they irritate the heck out of you.
> If someone told me that 3com wanted an NDA before giving the
> inside scoop on their chipsets and that is why the OpenBSD driver is
> so poor I could understand, but I've never seen anything like that.
well, we have no docs. The OpenBSD driver probably DOES have problems
above and beyond those of the cards, but you can clean the pig up a
lot...but it's still a pig (no offense intended to pigs). 3Com wrote
the Netware and Windows drivers...and they didn't get it "perfect",
either. In fact, the 3Com Windows drivers win a non-unique prize for
being annoying ("Hi, we'll load this stupid diagnostic that kicks you
off the network, crashing windows, if you try to run it. So don't run
it. Just kiss a chunk of RAM bye-bye!")
...
> I have quite a few older Dell GXwhatevers in my lab that come with
> integrated xl nics and I hate to have to replace the nics in them.
> but that's just my ranting...
As do I...GX100, GX1, etc.
Hey, use 'em. If they do the job for you, great. If they puke on you,
you have been warned. If you want a sure-thing, that's probably not the
NIC to chose, but I'd give you 80%-90% odds that it will work
sufficiently well. Almost every device out there has someone who has
had no problems with it, and someone who has had nothing but problems.
Most people never need to fill a 100Mbps cable, and won't notice an
occassional bobbled packet or a little bit high of processor load. I
use 'em as lab computers myself, usually with the on-board NIC. On the
other hand...you can put three quad-port cards in a GX100, and five
quad-port cards in a GX1:
dc19 at pci7 dev 7 function 0 "DEC 21142/3" rev 0x41: irq 10, address
00:60:f5:08:54:27
lxtphy11 at dc19 phy 1: LXT971 10/100 media interface, rev. 1
:)
Reality check: if you are running a firewall with a cable modem on one
side, you don't need to worry much. Unless the thing bobbles the duplex
or link speed. AGAIN.
Nick.
