16 Bit PCMCIA USB Adapter?

622 Aufrufe
Direkt zur ersten ungelesenen Nachricht

Jason Bertram

ungelesen,
22.06.1999, 03:00:0022.06.99
an
Does anyone out there know the name of a company that manufacutres a 16-bit
PCMCIA to USB adapter card? Have tried the usual sources (PCConn,
MicroWarehouse, etc...) and have searched the web to no avail. All anyone
seems to be producing are CardBus adapters.

Is there a technical reason for this, or are the manufacturers just being
lazy? :)

Thanks In Advance,
Jason Bertram

Virgil Smith

ungelesen,
23.06.1999, 03:00:0023.06.99
an

Yeah! Do the math off of 8MHz <ISA> clock rates with a 16-bit data bus
translating into 12Mb <USB> throughputs. It doesn't work out
particularly favorably.

For a more thurough analysis search this group and
comp.os.ms-windows.programmer.nt.kernel-mode at www.deja.com for
PCMCIA and USB <and posts from me if that turns up too much to sift
through>.


-----------== Posted via Newsfeeds.Com, Uncensored Usenet News ==----------
http://www.newsfeeds.com The Largest Usenet Servers in the World!
------== Over 73,000 Newsgroups - Including Dedicated Binaries Servers ==-----

Ian Stirling

ungelesen,
29.06.1999, 03:00:0029.06.99
an
Virgil Smith <Vir...@Nomadics.com> wrote:
>>Does anyone out there know the name of a company that manufacutres a 16-bit
>>PCMCIA to USB adapter card? Have tried the usual sources (PCConn,
>>MicroWarehouse, etc...) and have searched the web to no avail. All anyone
>>seems to be producing are CardBus adapters.
>>
>>Is there a technical reason for this, or are the manufacturers just being
>>lazy? :)

>Yeah! Do the math off of 8MHz <ISA> clock rates with a 16-bit data bus
>translating into 12Mb <USB> throughputs. It doesn't work out
>particularly favorably.

That's megabits.
USB is 12megabits/sec not megabytes.


Virgil Smith

ungelesen,
29.06.1999, 03:00:0029.06.99
an

I KNOW THATS WHAT I SAID.
It still does not work out particularly favorably.

BTW: Everyone I have ever dealt with has adopted Mb as megabits and MB
as megabytes. Its not really set in stone anywhere that I know of,
and I'm sorry I didn't spell it out in a public forum, but the point
remains.

David Hinds

ungelesen,
30.06.1999, 03:00:0030.06.99
an
Virgil Smith (Vir...@Nomadics.com) wrote:
: >Virgil Smith <Vir...@Nomadics.com> wrote:
: >
: >>Yeah! Do the math off of 8MHz <ISA> clock rates with a 16-bit data bus

: >>translating into 12Mb <USB> throughputs. It doesn't work out
: >>particularly favorably.
: >
: >That's megabits.
: >USB is 12megabits/sec not megabytes.
: >
:
: I KNOW THATS WHAT I SAID.
: It still does not work out particularly favorably.

I guess that depends on your definition of "particularly". 16-bit
PCMCIA devices can sustain better than 12 Mb/sec throughput. Not to
mention that many vendors seem to think it is reasonable to sell
16-bit 100baseT cards, where the mismatch is far worse.

-- Dave Hinds

Virgil Smith

ungelesen,
30.06.1999, 03:00:0030.06.99
an
>I guess that depends on your definition of "particularly". 16-bit
>PCMCIA devices can sustain better than 12 Mb/sec throughput. Not to
>mention that many vendors seem to think it is reasonable to sell
>16-bit 100baseT cards, where the mismatch is far worse.

Well OK people do screwy things, but a 16-bit PCMCIA to USB adapter
<figured as having to cope with PCMCIA being supported through ISA>
gives 8MHHz * 16bits = 128Mb/s (megabits/second). Now any ISA
read/write cycle takes at least two "clocks" so that gives 8MHHz *
16bits / 2 = 64Mb/s. Divide that by the USB data rate of 12Mb/s gives
roughly 5.

So ISA/16-bit PCMCIA has a 5 times throughput advantage on USB.

So one could accomplish this by taking as much advantage as can be
managed via DMA and making the card itself generate as much of the USB
protocol overhead as possible, but I personally wouldn't bother. This
is especially true since 16-bit PCMCIA DMA support is so totally flaky
that I <and probably most others> just wouldn't bother with it. Thus
you'd need about 20% CPU utilization to make this work.

David Hinds

ungelesen,
30.06.1999, 03:00:0030.06.99
an
Virgil Smith (Vir...@Nomadics.com) wrote:
: >I guess that depends on your definition of "particularly". 16-bit

: >PCMCIA devices can sustain better than 12 Mb/sec throughput. Not to
: >mention that many vendors seem to think it is reasonable to sell
: >16-bit 100baseT cards, where the mismatch is far worse.
:
: Well OK people do screwy things, but a 16-bit PCMCIA to USB adapter
: <figured as having to cope with PCMCIA being supported through ISA>
: gives 8MHHz * 16bits = 128Mb/s (megabits/second). Now any ISA
: read/write cycle takes at least two "clocks" so that gives 8MHHz *
: 16bits / 2 = 64Mb/s. Divide that by the USB data rate of 12Mb/s gives
: roughly 5.
:
: So ISA/16-bit PCMCIA has a 5 times throughput advantage on USB.

In practice, I think it is more like 1-2 times. DMA is not an issue:
ISA bus DMA is too slow to be useful, even if PCMCIA had a better way
of supporting it.

Even if a PCMCIA-to-USB card soaked up 100% CPU, that isn't really an
argument against it. If you've got a laptop without USB, and want to
use USB devices, you don't have many options. As another example,
there are plenty of great uses for 16-bit PCMCIA SCSI adapters, with
10 MB/sec burst capability, that can sustain 1/5 of that over the
PCMCIA bus if they are lucky.

My guess is that the existence of Cardbus-to-USB cards probably
reflects the fact that PCI-to-USB chips are readily available and make
a Cardbus card trivial. I thought I saw ads for 16-bit USB cards a
couple years ago, when USB was new; maybe they weren't cost effective
to produce in small numbers.

-- Dave

Virgil Smith

ungelesen,
30.06.1999, 03:00:0030.06.99
an
>: So ISA/16-bit PCMCIA has a 5 times throughput advantage on USB.
>
>In practice, I think it is more like 1-2 times. DMA is not an issue:
>ISA bus DMA is too slow to be useful, even if PCMCIA had a better way
>of supporting it.
>
>Even if a PCMCIA-to-USB card soaked up 100% CPU, that isn't really an
>argument against it. If you've got a laptop without USB, and want to
>use USB devices, you don't have many options. As another example,
>there are plenty of great uses for 16-bit PCMCIA SCSI adapters, with
>10 MB/sec burst capability, that can sustain 1/5 of that over the
>PCMCIA bus if they are lucky.

It's not an argument against the physical possibility of it, just a
reason that someone wanting to develop, sell, and support it is
unlikely. If you make it, you usually have to support it, and if it
is that much of a system hog, you're going to catch a lot of support
flak.

>My guess is that the existence of Cardbus-to-USB cards probably
>reflects the fact that PCI-to-USB chips are readily available and make
>a Cardbus card trivial. I thought I saw ads for 16-bit USB cards a
>couple years ago, when USB was new; maybe they weren't cost effective
>to produce in small numbers.

Agreed this is the easy out that reduces the demand for PCMCIA to USB
cards.

The original question asked if there was a technical reason for this
lack. I gave an overly brief response that would have served much
better if I'd said something like....

1. Cardbus-to-USB cards are probably much easier than PCMCIA to USB
cards because of the similarity between CardBus and PCI, and the
availability of PCI-to-USB chipsets.

2. If you do the math 16-bit PCMCIA throughput does not compare
favorably enough to USB throughput to make such an adapter a
reasonably system-friendly and/or continuously bandwidth reliable
product.

3. 16-bit PCMCIA is already an out-dated technology for laptops and is
being replaced with CardBus. Also most new laptop models include USB
off the shelf, and more and more are adding this. So making a PCMCIA
to USB adapter is a short term worth project <from a marketing sense>.
Since CardBus already has a good lead start on USB and can handle USB
easily and without major performance issues it makes much more sense
to make CardBus to USB adapters, and PCMCIA adapters probably just
aren't worth the effort. There may be people now who would <really>
like them because they don't want to upgrade their laptops, but the
number will dwindle too quickly to make such an effort generate a good
return.

4. BTW: If you haven't read the USB specification do so and you should
quickly notice that host controllers are very very far from trivial.
If you don't notice this you've obviously never tried to develop
scheduling software or hardware.

Allen antworten
Dem Autor antworten
Weiterleiten
0 neue Nachrichten