'Legacy-Free' PCs Appearing Everywhere

gjt writes "Finally. The death of of the ISA slot is near. Red Herring is running a story on the Legacy Free PC. Plug all of your mice, keyboards, joysticks, modems, etc. into the Universal Serial Bus. Compaq is releasing a computer called the Vista which will do just that. Yes, Apple did that over a year ago with the iMac and PCI based G3 and G4. Of course, if you're like me, you'd want to build your own box. Asus makes legacy free "PC 99" compliant motherboards. I wonder if this means more IRQ numbers. And what's the state of USB and Firewire support in Linux?" Suddenly USB is everywhere. Will it take hold? A lot of PC manufacturers sure seem to think so.

Comments

Status of Linux USB

[kroah]

USB is up and running just fine in the 2.3.x series of kernels.
There is even a backport of the USB stack into 2.2.12 right here
Also check out the USB HowTO for getting started.

And the main Linux-USB page is www.linux-usb.org

Of Course it will take hold

[ecampbel]

PC's are aiming for the mainstream, and most consumers will not miss support for the legacy hardware. As long as the computer has a modem, and capabilities for peripherals such as printers and joysticks, the computer is appealing to the masses. USB supports all these technologies. For more high end customers, FireWire will complete the package. Apple realized this a long time ago, and the rest of the world is following suit.

Re:Of Course it will take hold

[Cuthalion]

Not really. Firewire doesn't need Intel; it has the entire consumer electronics industry behind it. Plus Compaq, NEC, and Apple.

Intel makes the most common PC chipsets - at least for intel based systems. It usually takes some time before Via et al catch up.

If USB is integrated on the motherboard, and firewire requires a $20+ card, USB has a large advantage.

Re:Of Course it will take hold

[Salamander]

>Intel makes the most common PC chipsets

How convenient for Intel, that they can hide the USB2 implementation cost in higher chipset prices (hint: more than the $0.25/device Apple et al are charging for 1394). Even if USB2 were technically superior to 1394 - which it's not - I'd be leery of adopting a "standard" controlled by a single manufacturer over one approved by IEEE. That goes double if the manufacturer is Intel.

Maybe VIA or someone will integrate 1394 on their motherboards instead of USB2. That would be great.

Re:Of Course it will take hold

[Cuthalion]

How convenient for Intel, that they can hide the USB2 implementation cost in higher chipset prices (hint: more than the $0.25/device Apple et al are charging for 1394).

Am I mistaken in thinking that the $0.25/device is merely for the licensing? The cost of implementation is certainly more than that for both formats?

Re:Of Course it will take hold

[Salamander]

>Am I mistaken in thinking that the $0.25/device is merely for the licensing? The cost of implementation is certainly more than that for both formats? Yes, that is just the cost of licensing, and yes, the cost of implementation is more than that for both formats. However, the cost to implement presents an interesting tradeoff.

• USB is based on a model of hosts polling devices. This allows for cheaper device implementations, because devices don't need to initiate transactions, but makes the host-side implementation more complex. It's a good fit when you have several dumb devices and the polling latency doesn't matter - mice, keyboards, stuff like that.
• 1394 is more peer-to-peer, with devices able to initiate transactions independently. This makes device implementations more costly and host implementations less so. It's a good fit when you only have one or a few devices, when the polling latency matters, or when the device is connected to more than one machine.

My point, besides the obvious complementary nature of the two approaches, is that Intel has a much higher market share in motherboard chipsets than in consumer electronics. Of course they're going to push the host-centric solution, because they can bury a lot more profits for themselves in the cost of a motherboard.

USB and Firewire rock

[friedo]

What makes them great is that at long last we have a virtually limitless number of slots for devices. 127 on a USB bus, 63 on a Firewire bus. Only problem is, it's not getting implimented correctly. For example, vendors are making USB hard drives and removable media drives. USB was NOT designed for this. It's certainly faster than serial connections or good ol' ADB, but it was designed for low bandwidth peripherals, i.e. modems, keyboards, mouses, printers, cameras. I'd love to see more Firewire devices (especially sotrage) out there...bye bye SCSI!

Re:USB and Firewire rock

[Cuthalion]

so if you're a masochist, you'll be able to hook that USB 2 hard drive up to your current USB ports, and blaze along at a breathtaking 1.5 megabytes per second. I can see the advertising now; "Get the power of a hard drive with the speed of CD-ROM!"

CD-ROMs aren't that slow anymore. You're talking 10x. Real nice CD-ROMs (such as kenwood trueX) go 43-70x (5-10 megabytes/s)

Does this mean no more jumpers?

[KBrown]

I don't know you, but with only 15 IRQs I prefer to buy ISA cards which still use jumpers so I can control which IRQs I want to use.

And this is because my BIOS is so "SMART" that it does not want to use the IRQ 12 for a PCI card even if I explicitly specified not to use any serial mouse neither it's IRQ (12). The same happens with the second serial port and IRQ 3.

So if I want to use any of these IRQs I have to use a card with jumpers and set them to either IRQ 3 or 12 or 7 (the parport) and this is the only thing I can do if I begin running out of IRQs.

What will happen when I am no longer able to purchase a MB with ISA slots where I can use jumpers to choose the IRQs I like for my cards?

What will I do to all the ISA cards with jumpers I have purchased during the years?

What will I do if Linux USB support and Fire Wire support are not ready yet when that moment arrives?

Will some day Alan Cox and Linus say yes to Devfs so I can know with more accuracy where to find my USB, SCSI and Fire Wire devices without the need to have 10 000+ /dev nodes for every possible piece of hardware I can plug in my machine?

Re:Does this mean no more jumpers?

[scrytch]

> Will some day Alan Cox and Linus say yes to Devfs so I can know with more accuracy where to find my USB, SCSI and Fire Wire devices without the need to have 10 000+ /dev nodes for every possible piece of hardware I can plug in my machine?

What on *earth* do you need all of them present for? Detect device, MAKEDEV. Repeat. Perfectly automatable, and it's pretty much exactly what Solaris already does. I like the idea of devfs, I don't like the idea of having to tar it up to maintain state though.

Legacy is good

[Jon Peterson]

There's nothing wrong with FDD and Serial. I know whole companies that are not upgrading their Macs simply because they require floppy drives. Floppy drives are useful. They are a universal standard for transportable data. A text file on a floppy can easily be read on Sparc, Intel, Mac, SGI and most other hardware you care to mention. Sneakernet is a useful out-of-band comms system for use when the network is down or not there.

Serial is similar. Serial is still the default way of connecting to a massive range of hardware appliances, from robots to burglar alarms, to telecoms hardware. Having just designed a large server farm, I can testify to the usefulness of Serial as a fall-back remote access channel.

Removing floppy drives from computers because they have USB and Ethernet is about as smart as removing the staircase from a 20 story building because it's got plenty of lifts in.

Re:Legacy is good

[dufke]

Removing floppy drives from computers because they have USB and Ethernet is about as smart as removing the staircase from a 20 story building because it's got plenty of lifts in.

True. But I would rephrase it: "Removing removable media..."

The trouble with floppies is that they are WAY too small for a lot of modern applications. Example: I'm an amature photographer. The average file size of a scanned negative is 20Mb. For ONE picture. Zip's are in use by the pro photographers I know, but 100Mb (250Mb for new) is rapidly becoming very small. CD-RW is too slow...

Castlewood ORB anyone? Is there any Linux support?

-

Re:Keep ISA.... DUMP PS/2!!!!

[Yarn]

I hot swap PS2 devices frequently.

AFAIK the only difference betweeen an AT keyboard connector and a PS2 keyboard is the connector.

The PS2 mouse is moderately different tho.

Re:USB keyboards are EVIL

[Yarn]

I heard that despite the USB style connector 'USB' keyboards are still PS2 underneath and the USB controller on the mobo emulates a PS2 controller.

Anyone know if I was hallucinating?

It could well mean no more cards ....

[taniwha]

and as a result no more IRQs, or IO ports or.... hooray!

Just boxes on your desktop hooked together with USB or firewire .... that hot plug .... usb/firewire disks, cameras, (firewire) video cards, net connections, sound cards (usb/firewire speakers really), kbds, joysticks, ..... a brave new world - I can't wait 'till ISA is dead!

Last time I checked firewire's pretty close in speed to a backplane bus anyway - for a low end box why not get rid of all that empty space, slots etc - if you do that and only have external connections then you can build smaller form factor cases, get away from standard sized MBs. With no holes for cards you can do cheaper FCC (meaning cheaper MBs and cases). etc etc

With the push to much cheaper PCs this sort of thing is going to happen - even if it saves $10 on the production price to someone who's making 1M boxes/year that's $10M.

You *want* that IRQ

[Bill Currie]

Without IRQs, the os has to sit there polling the hardware to see if it's finished: not good for performance. Rather than getting rid of IRQs, what you really want to do is increase the number of available IRQs (NOTE: I may be off track here, as well designed IRQ sharing might not be a bad thing).

As to your other points: in general, I would agree, but there are times when at least serial ports are useful (though not to the average user). The rest? What for? I rather like ethernet printers:).

Re:You *want* that IRQ

[Millennium]

Yes, you're right. IRQ's are good, but they have to be done right ro be good. What do I mean? I mean auto-configuring IRQ's, such that there are no conflicts but the user never has to worry about configuring them.

By the way, Macs have done that for years. I see no advantage to hand-configuring them; more than a handful of PC magazines have even commented that Apple did it right. I wonder why the PC industry has yet to catch up.

Re:You *want* that IRQ

[Bill Currie]

I know for a fact that sharing IRQs on an ISA bus does not (usually?) work well at the hardware level! I once tried to do this when I had four serial ports (two on irq 4 and two on irq 3, ie the standard settings for 4 ports) and even though my serial port driver supported shared IRQs, it did not work well: I kept losing interrupts on the second port on that line. Adittedly, it may have been bugs in my code (for those interested, have a look at serio on my web page), but as it seemed intermittent, I think it was more likely the hardware. I don't know if this was a limitation of ISA or cheap boards.

Re:You *want* that IRQ

[Bill Currie]

Thanks for that, it's nice to know that IRQ sharing is possible on ISA (and that it probably wasn't my code's fault:). Hmm pretty easy to implenet a tri-state output IRQ: use a tri-state buffer with it's input tied to the IRQ level needed and have the board's local IRQ line drive the output enable of the buffer. That should do the job very nicely while allowing the IRQ level to be programmable (the buffer input can be tied to a register output rather than Vcc and Gnd or Vdd or Vss (did I get those right? it's been a while since I did any electronics, especially CMOS)).

M$/Intel are pushing a legacy free world ...

[taniwha]

They have been working up to this for a couple of years now and I beleive the big push will be in the coming year. This is more aimed at building ISA-less systems with all the legacy devices gone. This means no ISA interrupt controller, FD controller, KBD controller, serial, parallel etc etc, no VGA compatable frame buffer

PCI/AGP and USB are the big winners here and I expect will be the main stream for the next few years. Low end machines will probably do without PCI slots (but will have a PCI bus between the chips on the motherboard).

big change

[jetson123]

Most of the peripherals we now take for granted (sound, 3D graphics, SCSI, etc.) started out as expansion boards. Closing the mainstream PC architecture will destroy the market for such expansion boards, even if a few high end machines keep a bus. On the other hand, PCI isn't really all that fast anymroe anyway, so something had to change.

I think a priori, this is not a good change as far as hardware is concerned. But something good may yet come from it: the bottleneck that going to USB and FireWire for expansion causes may finally propel the PC industry towards a more distributed and parallel architecture. In a USB/FireWire-only world, a novel piece of 3D graphics simply has to include its own general purpose processor that handles communications back to the PC.

Whatever its effect on hardware, this should be great news for Linux and other non-Microsoft operating systems. It looks to me that drivers for USB and FireWire-based devices ought to port much more easily between different operating systems. Many of them can actually even run in user mode. Configuration and resource allocation should also get simpler.

One problem

[scumdamn]

Most of these "legacy free" PCs run off of Intel's crippled 810 or 810e chipset. As far as I know, there's no X driver for the 810. So if you're cool with no Xwindows on these legacy free boxes, that's fine, but I'd first make sure they had either a BX chipset or a non-Intel chipset.

Oooooh! Hot swappable!

[slothbait]

I have to laugh when I see "hot-swappable" listed as a compelling reason to "upgrade" to USB. Yeah? My COM ports have always been hot-swappable. I switch back in forth between my track ball and mouse all the time and I'm sure as *hell* not going to reboot for that. PS/2 was a bad idea.

("I've got an idea! Why don't we make the keyboard plug look *exactly* like the mouse plug! Now *that's* ease of use...." "Great! Then we'll make it so that your system crashes whenever accidentally unplug a periphereal. They'll *love* that!")

ISA should definately die, and USB should probably take over as well, but software on the PC side is not to the point where we can make the USB plunge yet. Of course, it may take a drastic action such as this to *force* the software to come in line, but I pitty the poor users caught in the middle.

("Why can't I use my keyboard/mouse in safe mode? My display settings are messed up and I can't fix them! I'm stuck!" "This is a known issue with MS Windows, and will be addressed in the next service pack. Please wait patiently for your patch".)

*shudder*
--Lenny

Legacy Hardware

[wowbagger]

I'm all for ditching the legacy hardware as soon as possible, but...

Why should a manufacturer eliminate the ISA slots in a computer? If you want to avoid using legacy systems, simply do so, but don't deny me the option!

Here are, as I see it, the problems with doing away with the ISA slots as things stand today:

1. PCI only allows you to have about 4-5 slots without adding a PCI-PCI bridge chip. My two PCI machines have no PCI slots left after:
   1. Video card
   2. 3D card (or two)
   3. Network card
   4. SCSI card
   5. Sound card
   
   In fact, my game machine MUST have an ISA sound card, since the dual V2's, video card, NIC and SCSI take up all 5 PCI slots. If it didn't have ISA, it wouldn't have sound!
   
2. There are a lot of legacy devices out there that are either not going to be available on other busses or shall be so expensive as to not be worth considering. GPIB cards, EPROM burners, certain DSP development boards all come to mind (gosh, what do I do for a living?)
   
3. It is a DAMN site easier to design and prototype an ISA card than a PCI card! Kiss the garage hobbyist goodbye when ISA dies.
   

Now, I do want to address a couple of items I've seen mentioned in this thread about IRQs:

1. First, getting rid of ISA doesn't measurably increase the number of IRQs an x86 machine has. When Intel designed the PC implementation of PCI, they (IMNSHO) screwed up by not putting in a dedicated interrupt controller for the PCI bus. So, even when ISA is dead, 15 IRQs will be the law of the land. Now, by sharing an IRQ among all your USB devices, and another IRQ among all your Firewire devices, and getting rid of COM1-4, LPT1-3, and the mouse, you might make a few more devices available, but unless you cut off all back-compatiblity, you cannot get rid of the keyboard IRQ or COM1.
   
2. The fact that USB "doesn't use an interrupt" and therefor will require polling is false. USB uses an interrupt, and when a device changes state, it generates a message that causes a USB interrupt. So, you don't poll your USB keyboard/mouse/whatever. Ditto for Firewire.
   
3. USB is too slow for disk drives: That would depend upon what you are using. Would I want to hook up a true hard drive to USB? Of course not! But would I hook a Jaz, Orb, or other removable up? The speeds on these devices are not that large compared to USB, especially USB 2.0 (400MBit/sec).
   

The only thing about USB/Firewire/I2O etc. that worries me is the "You want drivers? Yer runnin' Winders ain'tcha?" mindset most HW venders have. As an embedded systems designer, I am CONSTANTLY telling these morons "No, I am NOT running Windows, I am running a real time OS, and I need the programming specs for that! No, I CANNOT use the BIOS you provide, I am running in protected mode and your BIOS only works in real mode. No, I am NOT running Windows, weren't you listening the first twelve times I told you that?"

However, things are getting better with more HW vendors supporting Linux (therefor releasing source that I can adapt as needed to my needs).

And before you ask, while I am considering using Linux in several projects I am designing, there are other places where it just doesn't make sense, and therefor I have to adapt drivers, not install the RPM. Let's not get into the mindset of "You want drivers? Yer running Linux ain'tcha?" ;^)

PS: Rob, why don't you put a "Spellcheck" button on the post page? It would sure help us all out!

Re:Legacy Hardware

[Tiroth]

I would get rid of ISA in a second. Here's why:

-ISA cards don't share IRQs. That means that even with a constant number of IRQs, replacing an ISA card with a PCI generally means more free IRQs, and less conflicts.

-ISA cards (as we've all noticed) don't do PnP worth crap. It's also not always possible to tell what the IRQ/DMA/IO settings are for a given card...they often don't listen to the BIOS when in PnP mode.

-ISA cards are harder to troubleshoot on a system. Believe me, I service computers. I hate ISA cards.

-less ISA slots means more room for PCI slots. Even if these are bridged, it's still a good thing. More slots mean more support for _current_, as opposed to old (legacy) hardware.

-finally, moving from ISA provides more encouragement for designers to take advantage of the more capable bus.

As far as the maturity of Firewire and USB, their time is coming, but I'm skeptical of their ability to replace basic hardware like mice and keyboards. I'm certainly not going to go buy half a dozen keyboards and mice to get the same functionality I currently have, and end up with less USB ports to boot.

Re:Legacy Hardware

[Anonymous Shepherd]

Well, in a non-legacy solution you'd be ridding yourself of three PCI video cards for one AGP or one PCI solution.

Your sound could very well be offloaded onto USB, as could your modem, if you use modems.

SCSI can't yet be replaced by FireWire, I agree, since I use SCSI peripherals. But it's*almost* on parity, and much simpler. I do have FireWire, btw, because NT doesn't support USB!

We may see network get shifted onto FireWire as well.

So the only PCI solutions would be:

SCSI
Network

Video AGP
Sound USB

Now that might not make you happy...


-AS

Also cards where no USB/FW/PCI version exists

[jd]

I have many cards which exist only in ISA form. Perhaps the best-loved is my Roland LAPC-1 - easily still the best soundcard for sampled sounds and MIDI, for all that it's ancient.

Personaly, I wouldn't be surprised if Microsoft pushed USB for all it's worth, to keep Linux out of competition for just a little bit longer, ensuring the standards and specs change just enough each version to break the Linux drivers.

Is it -really- the best way to go, to burn ALL the bridges, when in ogre territory?

Re:Also cards where no USB/FW/PCI version exists

[IntlHarvester]

Personaly, I wouldn't be surprised if Microsoft pushed USB for all it's worth, to keep Linux out of competition for just a little bit longer

Except that Linux has better USB support than Windows NT, which will have absolutely no support until next year some time. The biggest victim of MS/Intel's "PC99" push has been Microsoft, because it's limited the adoption of WinNT workstation (which is twice the price of 98). Hope that ends the conspiracy theory.
--

USB and FireWire not competing

[Roy Ward]

USB and FireWire (IEEE-1394) aren't competing standards - at least they weren't designed to be. USB is designed for low bandwidth devices, FireWire for high bandwidth, with not a lot of overlap. USB is not offically a standard at all, in that the definition is still under the control of Intel. I'm not saying that USB is bad - but it is optimised for low bandwidth, right down to having cheap cables.

Then Intel gets this idea that they can have the whole pie to themselves (talk about a bob each way - they are part of the IEEE-1394 consortium too), so they bring on this USB 2.0 vaporware - the _specification_ due in a maybe a few months specifies something that maybe will (under ideal conditions) have a similar speed to that available with FireWire _now_.

There is an excellent article on this at:

http://www.MacKiDo.com/Hardware/USB2.html

Read that article if you think that USB 2 will be as good as FireWire.

If USB 2.0 succeeds, it will be on Intel marketing power, not technical merit.

OTOH, that hasn't stopped inferior techology before. I'm writing this on a G3 macintosh with EIDE drives, when EIDE ousted SCSI _not_ because it was in any real way better, but because the fact that it is the 'standard' in PCs made the drives a lot cheaper. I hope we don't also get the 'nearly good enough' technology of USB 2.0 in a similar way.

Roy Ward

Maybe once it doesn't cost an arm, leg, and teste.

[Wakko Warner]

Sure, it's "free" when you buy a new PC, but a cheap, no-frills PCI Firewire board is $150. At that price, I imagine it'll be a while until we see support for it in Linux. Who's gonna go out and buy a card, and then a $1500 DV camera (or a $700 Firewire hard drive) to start hacking in support in Linux?

- A.P.
--

"One World, one Web, one Program" - Microsoft promotional ad

Re:Voodoo 2 is legacy

[Pascal Q. Porcupine]

Not to mention that although an SLI V2 setup has a decent fillrate, it completely lacks any sort of rendering quality and has none of the rendering features which have been in OpenGL for ages and are going to start showing up soon, such as stencil effects (VERY useful for shadows and CSG-based modelling, among other things). Do yourself a favor and get a TNT2 or G400 and free up two of your PCI slots and make your 3D less kludgy. :) (Yeah, I know, 3dfx cards are the only ones which can do 'decent' hardware 3D under Linux right now. It's changing quickly, and in the meantime, with a TNT or TNT 2 you can always run nVidia's unstable GLX driver, which really works quite well.)
---
"'Is not a quine' is not a quine" is a quine.

Re:Isa is *slow*

[Cuthalion]

The PCI soundcards that you mentioned are OBJETIVELY no better to the Human Ear, since a good 16-bit soundcard can produce two sounds that are separated by so little that the ear treats them as the same (either in duration, tone, or volume).

1 channel of CD quality audio = 150 kBytes/s.
8 channel of CD quality audio = 150 kbits/s

Yeah, you could mix the audio in software. What if you now want to put different DSP effects on the channels? (3d sound?) You want to do that on the main processor too. Sound cards are progressing the same way video cards have. They're not as high-profile, because a lot of people don't care, however. A PCI sound card does make some sense.

Show me the typist that can type over 1000 cpm, so I guess that 9600 baud serial keyboard is gonna last a while. The reason we shouldn't scrap all legacy hardware is there is a good portion of it that exeeds it's design specifiations to this day.

Bandwidth are not the only reasons to use a USB or PCI device. They are also allow for more flexible/easy configuration. ISA sucks for plug-n-play - even ISAPNP isn't great. USB keyboards let you put a bunch of them on the same system (if you're so inclined), and are more orthogonal (if you're using USB for other stuff as well).

Relax, people will still make SOME motherboards with ISA slots as long as there are people like you to buy them. They will be (within a few years, I imagine) be more expensive (more logic, fewer produced), but you'll be able to get one if you refuse to use "bleeding-edge" technologies.

Firewire a niche product?

[IntlHarvester]

Wait until Sony starts including iLink (FireWire) on all of it's consumer electronic products - not just the Playstation, but TVs, DVD players, recievers. (Most Sony computers come with iLink too.)

Perhaps then the applications will be more obvious, and we'll start to see greater adoption.
--

Re:Isa is *slow*

[Cuthalion]

I believe 3com's "Internet Gaming Modem" fits the bill.

Also, external modems are pretty decent, although the UART on your standard motherboard may not be able to make the most of a compressed 56K connection.

Flaimbait?

[Anonymous Shepherd]

Why am I flamebait?

-AS

Re:More IRQs, and less BIOS code, and more partiti

[Bill Currie]

There's actually room in most MBRs for 16 partitions (possibly more, I can't remember), but thats starting to cramp the code space a little. and then all the partition tools would have to be updated (do you beleive MS would update fdisk and dos to support 16 partitions?). Hmm, autodetecting a 16 partition MBR wouldn't be to hard: grovel around in the code section looking for where it loades the register (cx IIRC) with the number of partitions to scan and you're off. Hmm MBR signature checking, yummy (not).

Re:Firewire is a proprietary, royalty required

[Salamander]

>Apple demands one dollar per firewire chip in royalties.

At least it's an honest, up-front charge. The "Intel tax" on USB2 is likely to be much higher, but hidden in things like increased motherboard/chipset prices. I'll gladly pay a buck for the difference between an IEEE standard and yet another Intel pseudo-standard, and I think most educated consumers would as well.

Re:USB

[Salamander]

>I'd like to see some progression on IEEE-1355.. I dunno offhand what the status of work on 1355 is, but it's a far more exciting technology than either USB or IEEE-1394.

It sounds interesting, but I'm not convinced yet. How many nodes does it support on a simple loop/bus/whatever without added-cost routers/switches. How hot-pluggable is it? Does it support isochronous transfers?

I have the feeling that 1355 may be a truly great thing...for another niche. It may be a mistake to push it as an alternative or replacement for 1394, just as it is/was a mistake to push USB - a perfectly wonderful thing in its own niche - the same way.