You are here

Martin's PCIe Article

For anything relating to music-making on Windows computers, with lots of FAQs. Moderated by Martin Walker.

Re: Martin's PCIe Article

Postby Peter C » Tue Nov 22, 2005 10:45 am

jcschild wrote:Hey Peter,
well cant say i didnt tell you.

of the 6 manufacturers i asked they all said not now mabye later.

You did, and you may well be right.

But I still think it's equally possible that those guys have prouducts under wraps and are holding out on us because they have not decided when to jump and don't want the competition to know what they are planning...

I have predicted 2006Q1... and (fool that I am) I'm sticking with that. Come April 1st we'll all know if I am right or wrong. At that point, if I'm prved wrong, you guys get to choose - an appology or a drink - but not both.



Truth be told for the average guy a firewire interface is fine.

True, but I have my doubts about attaching low latency equipment to any shared bus. PCIe is the right answer - tons of bandwidth and QoS capability; plus I personally prefer to habe expensive stuff inside the case where it's neat and protected.

We are in a transition phase, and this lack of PCI soundcards is a hicccup. I'm quite certain the future is PCIe-shaped.


for more advanced studios a madi card etc is the answer.

Yes... but isn't the MADI interface to the DAW a PCI (opps, shouldn't that be a PCIe) card?



however i agree PCIe is here to stay. and PCI will most certainly disapear quicker than ISA.

i think the big issue with alot of them is waiting on real PCIe spec to appear. or at least the real controller not PCI to PCIe bridge crap.

i think when Vista ships and has its first service pack you may see PCIe in force. remember the real PCIe requies a controller which requires a re-write of code, alot of re-write.


Seems you have an insight on how PCIe works and it low level OS interface that I lack.

DOS runs on a PCIe mobo. I thought the interface looked the same and only hardware had to change; and that's why, for instance the PCI device addressing scheme has survived into PCIe.


Peter
Peter C
Regular
Posts: 151
Joined: Thu Jan 08, 2004 1:00 am

Re: Martin's PCIe Article

Postby Peter C » Tue Nov 22, 2005 10:54 am

Martin Walker wrote:Sorry Peter - I'm not in charge of SOS web site design, although I can certainly pass on your request to the powers that be ;)

Thanks... that's exactly what I meant when I said I was sure you would fix it.

Let me know the e-mail of the guy who will be responsible for implentiong this,and I'll forward him the spec...

But in general terms, stuff in [rant] brackets will have its monochrome backround replaced with a video of a flickering frame interspersed with shots of mopderates like Billy Graham or Stalin at the climax of one of their more entertaining (aka controversial) speeches.

Plus, if you have embedded rants, the inner rant will (provided you have a full 3D virtual reality monitor) reach out to you and throttle you (in the nicest possible way), just to make sure you are paying attention.

...

Nothing too ostentaious



Peter
Peter C
Regular
Posts: 151
Joined: Thu Jan 08, 2004 1:00 am

Re: Martin's PCIe Article

Postby spjessop » Tue Nov 22, 2005 1:54 pm

Peter C wrote:Yes... but isn't the MADI interface to the DAW a PCI (opps, shouldn't that be a PCIe) card?

It is indeed a PCI card.

I wonder what you could achieve with a FW800 digital only fireface? A FireMADI anyone? Firefaces are meant to be stackable...
spjessop
New here
Posts: 9
Joined: Wed Sep 01, 2004 12:00 am

Re: Martin's PCIe Article

Postby Martin Walker » Tue Nov 22, 2005 1:58 pm

Ah, nothing too complex then, and still suitable for anyone here who still hasn't got broadband :beamup:


Martin
User avatar
Martin Walker
Moderator
Posts: 16102
Joined: Wed Jan 13, 2010 9:44 am
Location: Cornwall, UK

Re: Martin's PCIe Article

Postby Peter C » Tue Nov 22, 2005 1:59 pm

Absolutely


Peter
Peter C
Regular
Posts: 151
Joined: Thu Jan 08, 2004 1:00 am

Re: Martin's PCIe Article

Postby Chris Diamond » Tue Nov 22, 2005 6:33 pm

Now that the online version of the December mag has been published I've had the chance to read the article.

Interesting stuff. I think Martin's conclusions in the 'Final Thoughts' section sum things up nicely. But there's by no means unanimous agreement on all points.

I particularly like the wildly varying opinions on USB devices. The RME guy "USB 2 never had and will never have a bright future as an audio interface format" (and his explanation as to why squares up with my own understanding of the two protocols) whilst the guy from ESI said "Firewire has no real practical advantages compared to a properly designed USB 2.0 audio interface". Go figure.

But one point regarding Firewire wasn't discussed in the article or (AFAIK) on this forum: to the best of my knowledge (in the PC world at least - not sure about Mac's) all Firewire chipsets are in fact PCI devices. Sure, they may be build directly on to the motherboard but they're connected to the main chipset via the PCI bus. Conversely, USB controllers are almost always built into the south bridge of the chipset.

So does this mean that if you have a Firewire based audio device you may be able to take advantage of the QoS features of PCIe with a device you own currently when PCIe based Firewire controllers beccome available?
Chris Diamond
Posts: 2
Joined: Fri Oct 17, 2003 12:00 am

Re: Martin's PCIe Article

Postby Peter C » Tue Nov 22, 2005 8:37 pm

Hi,

I was under the impression thwt USB devices are also connected to the PCI bus, like on-board Firewire.

Whatever, they will have to end up as PCIe devices in the end; so each will be directly connected to the PCIe switch because PCIe is a P2P system - there is no shared bus with multiple devices connected to it under PCIe.

But no - the firewire controller will have a driver; and unless someone specifically goes out to create a firewire divice offering QoS then there would be no QoS on a Firewire (or USB) attached device.

USB and Firewire are shared buses anyway; and do not offer QoS to their client devices (AFAIK). So a firewire attached soundcard is just a firewire attached device; and there is no obvious way to exploit thet fact theat the Firewire is a PCIe device to get a QoS gurantee for the soundcard.



Peter
Peter C
Regular
Posts: 151
Joined: Thu Jan 08, 2004 1:00 am

Re: Martin's PCIe Article

Postby Chris Diamond » Wed Nov 23, 2005 12:10 am

Peter C wrote:I was under the impression thwt USB devices are also connected to the PCI bus, like on-board Firewire.

I'm reasonably certain that this is not the case. Whilst you can of course buy USB add-in cards that operate on the PCI bus I think that most chipsets from the last couple of years include a USB root controller directly on the south bridge.

Peter C wrote:But no - the firewire controller will have a driver; and unless someone specifically goes out to create a firewire divice offering QoS then there would be no QoS on a Firewire (or USB) attached device.

USB and Firewire are shared buses anyway; and do not offer QoS to their client devices (AFAIK). So a firewire attached soundcard is just a firewire attached device; and there is no obvious way to exploit thet fact theat the Firewire is a PCIe device to get a QoS gurantee for the soundcard.

Yes, you're right of course. The Firewire protocol in it's current specification simply doesn't allow for this. But I wonder if, for example, you might be able to run a Firewire audio device with a smaller sample buffer when attached to a PCIe based Firewire controller????
Chris Diamond
Posts: 2
Joined: Fri Oct 17, 2003 12:00 am

Re: Martin's PCIe Article

Postby robinv » Wed Nov 23, 2005 12:34 am

Has anyone actually seen any PCIe cards of any description in anything ever, other than graphics? I haven't come across any myself so rumours of it becoming an international standard seem, to me, a little overstated. The things that used to take up PCI slots now either reside on the motherboard (firewire, network and audio), or are obsolete like modems. There's nothing in my computer that actually needs to be PCI - why would any manufacturer spend time developing for a format that doesn't seem to have a purpose?

I've built plenty of machines with 915 and 925 chipsets, combined PCI and PCIe and not seem any performance problems so the reason why other forums dont talk about it is because unless you are very unlucky there isn't a problem - it's the sort of thing driver writers like to get all hot and bothered about.

M-Audio's Delta range is now soooooo old that i really dont think they would complain if PCI became obsolete, it would give them an excuse to ditch them - great though they are. Emu are the only ones i worry about as their whole range is PCI based... I have noticed that Digidesign are developing PCie versions of their HD cards - who'd have thought that they would ever be ahead in the technology game?

My only worry is what to do with my Powercore and Delta 44 when i make my next upgrade jump to whati imagine will be a PCI-less systems...
User avatar
robinv
Frequent Poster
Posts: 595
Joined: Tue Aug 31, 2004 12:00 am
Molten Music Technology - Computers for doing music on
Making Music on the Microsoft Surface

Re: Martin's PCIe Article

Postby Peter C » Wed Nov 23, 2005 12:47 am

Hi,

Yes.

Fist of all PCIe is an international standard, in the sense that it is agreed to and adhered to by all the major players.

I have actually seen a PCie Gb Lan card - it had 4 Gb lan ports on it; and as each of these, on it's own had the same capacity as teh entire PCI bus it's pretty clear that it needed to PCIe.

And you can get PCIe SATA controllers. Two HD can easily saturate the PCI bus...

But to be fair, I have not seen many PCI cards (except soundcards) installed in PC in the last two years. Most stuff has been integrated into the mobo; and anything with pretnesion to bandwidth has been directly connected to the I/O chipset. That means it is, or seen must be a PCIe device.

Firewire 800 (and isn't a 88Mb/s USB coming) would have to be PCIe devices, as either of these would consume upto 60% of total PCI bus bandwidth alone.

The total inadequecy of PCI has been masked by the steady process of removing things from the PCI bus and giving them their own direct connection to the I/O ship. In theis sense the chipsets have been evolving towards PCIe informally for years - because the whole point about PCIe is that each PCIe device has its own P2P link to the PCIe switch - which is a big multiplexor ther constitutes the heart of the PCIe chipset.

"own PCIe link" = "direct connection" via agreed protocol.




Peter
Peter C
Regular
Posts: 151
Joined: Thu Jan 08, 2004 1:00 am

Re: Martin's PCIe Article

Postby Peter C » Wed Nov 23, 2005 12:53 am

Chris Diamond wrote:
Peter C wrote:I was under the impression thwt USB devices are also connected to the PCI bus, like on-board Firewire.

I'm reasonably certain that this is not the case. Whilst you can of course buy USB add-in cards that operate on the PCI bus I think that most chipsets from the last couple of years include a USB root controller directly on the south bridge.

Hmmm... that may depend on the chipset. The issue is unclear, and I suspect neither of us knows for sure, because it's quite possible to connect a device to the PCI bus inside a PCI I/O chip.

It's further clouded because the PCI addressing system has survived into PCIe (quite reasonably) so you can have devices tht are logically connecterd via the PCI bus but physically coinnected via their own link.


Peter
Peter C
Regular
Posts: 151
Joined: Thu Jan 08, 2004 1:00 am

Re: Martin's PCIe Article

Postby Tímo » Wed Nov 23, 2005 1:34 am

robinv wrote:Has anyone actually seen any PCIe cards of any description in anything ever, other than graphics? I haven't come across any myself so rumours of it becoming an international standard seem, to me, a little overstated.


My laptop has a PCI-expresss card slot instead of a PCMCIA slot, as per the full implementation of Sonoma (Intel Centrino mk2). I think there's a TV card already available for PCI-expresscard. Admittedly I don't know of any others yet.

robinv wrote:
I've built plenty of machines with 915 and 925 chipsets, combined PCI and PCIe and not seem any performance problems so the reason why other forums dont talk about it is because unless you are very unlucky there isn't a problem - it's the sort of thing driver writers like to get all hot and bothered about.


[Nuendo forum thread - NForce4 tests]

[RME - NF4/PCIe warning]

Think these are where the initial doubts spurred from.

...

There was latterly another thread [Fix coming for NForce4].

Did that fix for NF4 ever happen, BTW?
User avatar
Tímo
Regular
Posts: 121
Joined: Wed Sep 25, 2002 12:00 am
Location: Derby, England
http://Infekted.org ~ Access Virus news & community

Re: Martin's PCIe Article

Postby cc. » Wed Nov 23, 2005 8:18 am

Peter C wrote:
USB and Firewire are shared buses anyway; and do not offer QoS to their client devices (AFAIK).

I've told you this before Peter: firewire does support QoS. Google for "isochronous firewire" and read some of the links.
User avatar
cc.
Regular
Posts: 96
Joined: Tue Mar 11, 2003 1:00 am
Location: lisbon at the moment

Re: Martin's PCIe Article

Postby Peter C » Wed Nov 23, 2005 10:37 am

Hi cc,


You are quite right, thank you.

1. You had mentioned this before, but I didn't know what to google for and could find no further info.

2. "isochronous firewire" is indeed the one. I found a lot of stuff, most of it rather old, and crucially nothing that:

2.1 Described how abritration works to provide isochronous operation when multiple devices are using the firewire bus
2.2 Specified whether the drivers have to be written to support isochronous operation, (I can't believe the hardware handles it all) and if so whether this feature is actually available on Windows platforms
2.3 Told me how to go about writing a driver for (say) a soundcard so that it used the isochronous firewire feature.

3. But assuming all that works just fine, it's useless unless/until the firewire port on the PC is a PCIe device. PCI has no QoS capability.


Peter
Peter C
Regular
Posts: 151
Joined: Thu Jan 08, 2004 1:00 am

Re: Martin's PCIe Article

Postby Martin Walker » Wed Nov 23, 2005 3:10 pm

So when the first PCI Express audio interfaces appear in the 1st quarter of 2006 (that's right isn't it Peter ;)), what we're really looking for are claims of drivers that provide 'guaranteed no-click recording' capability.

If that's possible then we'll all be putting PCI Express slots much higher on our lists of 'must have' motherboard features :bouncy:


Martin
User avatar
Martin Walker
Moderator
Posts: 16102
Joined: Wed Jan 13, 2010 9:44 am
Location: Cornwall, UK

PreviousNext