Home » The PARIS Forums » PARIS: Main » Some PCIe questions
| Some PCIe questions [message #79321] |
Wed, 31 January 2007 10:05  |
![Deej [4] is currently offline](/theme/PARIS_Forum_Template/images/offline.png "Deej [4] is currently offline")
Deej [4]
Messages: 1292
Registered: January 2007
|
Senior Member |
|
|
This is a multi-part message in MIME format.
------=_NextPart_000_003A_01C74527.C8E28D00
Content-Type: text/plain;
charset="iso-8859-1"
Content-Transfer-Encoding: quoted-printable
So I hear the PCIe bus can handle 500MHz whereas our wimpy ass PCI =
busses can only go 133.
I also hear that the PCIe bus doesn't share any bandwidth at all with =
the PCI bus.
Taken from a post on another forum:
PCIe guarantees a certain amount of undivided=20
bandwidth to a device in the PCIe slot. It's a dedicated channel. On a=20
modern Intel Chipset like the 965 there will be some contention between=20
devices on the south bridge chip (USB2, LAN, Serial ATA disks, PCIe) but =
the bandwidth between the north and south bridge chips is quite high=20
relative to the bandwidth of the south bridge devices.
So I'm wondering what this means exactly? Does it mean that there is =
enough bandwidth to hang 7 x PCI cards in a 7 x Slot Magma chassis and =
not have to worry about IRQ sharing, especially if there is a PCIe =
graphics card in the equation?.........and of course, something like =
this gets the hampsters jumping on the wheel in my brain and my brain =
comes up with a way to potentially make a computer perform an unnatural =
act by using a PCI video card on the PCI bus since that bus doesn't =
share with anything else, thus eliminating any possibility of problems =
down the road with a graphics card.
......the stated conclusion from the other forum........
The bottom line is that a PCIe device with hard real time requirements=20
(like a multiport audio device) will have fewer problems maintaining=20
stable audio input and output.
........so there is morepotential to poperate a native system at lower =
latencies???
C'mon folks, step up to the mic and talk to me here.
;o)
------=_NextPart_000_003A_01C74527.C8E28D00
Content-Type: text/html;
charset="iso-8859-1"
Content-Transfer-Encoding: quoted-printable
<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.0 Transitional//EN">
<HTML><HEAD>
<META http-equiv=3DContent-Type content=3D"text/html; =
charset=3Diso-8859-1">
<META content=3D"MSHTML 6.00.5730.11" name=3DGENERATOR>
<STYLE></STYLE>
</HEAD>
<BODY>
<DIV><FONT face=3DArial size=3D2>So I hear the PCIe bus can handle =
500MHz whereas=20
our wimpy ass PCI busses can only go 133.</FONT></DIV>
<DIV><FONT face=3DArial size=3D2></FONT> </DIV>
<DIV><FONT face=3DArial size=3D2>I also hear that the PCIe bus doesn't =
share any=20
bandwidth at all with the PCI bus.</FONT></DIV>
<DIV><FONT face=3DArial size=3D2></FONT> </DIV>
<DIV><FONT face=3DArial size=3D2>Taken from a post on another =
forum:</FONT></DIV>
<DIV><FONT face=3DArial size=3D2></FONT> </DIV>
<DIV><FONT face=3DArial size=3D2><EM><STRONG>PCIe guarantees a certain =
amount of=20
undivided <BR>bandwidth to a device in the PCIe slot. It's a dedicated =
channel.=20
On a <BR>modern Intel Chipset like the 965 there will be some contention =
between=20
<BR>devices on the south bridge chip (USB2, LAN, Serial ATA disks, PCIe) =
but=20
<BR>the bandwidth between the north and south bridge chips is quite high =
<BR>relative to the bandwidth of the south bridge=20
devices.</STRONG></EM></FONT></DIV>
<DIV><FONT face=3DArial =
size=3D2><EM><STRONG></STRONG></EM></FONT> </DIV >
<DIV><FONT face=3DArial size=3D2>So I'm wondering what this means =
exactly? Does it=20
mean that there is enough bandwidth to hang 7 x PCI cards in a 7 x Slot =
Magma=20
chassis and not have to worry about IRQ sharing, especially if there is =
a PCIe=20
graphics card in the equation?.........and of course, something like =
this gets=20
the hampsters jumping on the wheel in my brain and my brain comes up =
with a way=20
to potentially make a computer perform an unnatural act by using a PCI =
video=20
card on the PCI bus since that bus doesn't share with anything else, =
thus=20
eliminating any possibility of problems down the road with a graphics=20
card.</FONT></DIV>
<DIV><FONT face=3DArial size=3D2></FONT> </DIV>
<DIV><FONT face=3DArial size=3D2>.....the stated conclusion from the =
other=20
forum........</FONT></DIV>
<DIV><FONT face=3DArial size=3D2></FONT> </DIV>
<DIV><FONT face=3DArial size=3D2><STRONG><EM>The bottom line is that a =
PCIe device=20
with hard real time requirements <BR>(like a multiport audio device) =
will have=20
fewer problems maintaining <BR>stable audio input and=20
output.</EM></STRONG></FONT></DIV>
<DIV><FONT face=3DArial =
size=3D2><STRONG><EM></EM></STRONG></FONT> </DIV >
<DIV><FONT face=3DArial size=3D2>.......so there is morepotential to =
poperate a=20
native system at lower latencies???</FONT></DIV>
<DIV><FONT face=3DArial size=3D2></FONT> </DIV>
<DIV><FONT face=3DArial size=3D2>C'mon folks, step up to the mic and =
talk to me=20
here.</FONT></DIV>
<DIV><FONT face=3DArial size=3D2></FONT> </DIV>
<DIV><FONT face=3DArial size=3D2>;o)</FONT></DIV>
<DIV><FONT face=3DArial size=3D2> </DIV>
<DIV><STRONG><BR></STRONG></DIV></FONT></BODY></HTML>
------=_NextPart_000_003A_01C74527.C8E28D00--
|
|
|
|
| Re: Some PCIe questions [message #79346 is a reply to message #79321] |
Wed, 31 January 2007 18:34  |
Chris Ludwig
Messages: 868
Registered: May 2006
|
Senior Member |
|
|
Hey Deej,
DJ wrote:
> So I hear the PCIe bus can handle 500MHz whereas our wimpy ass PCI
> busses can only go 133.
>
> I also hear that the PCIe bus doesn't share any bandwidth at all with
> the PCI bus.
http://www.pcisig.com/specifications/pciexpress/
>
> Taken from a post on another forum:
>
> /*PCIe guarantees a certain amount of undivided
> bandwidth to a device in the PCIe slot. It's a dedicated channel. On a
> modern Intel Chipset like the 965 there will be some contention between
> devices on the south bridge chip (USB2, LAN, Serial ATA disks, PCIe) but
> the bandwidth between the north and south bridge chips is quite high
> relative to the bandwidth of the south bridge devices.*/
> /**/
> So I'm wondering what this means exactly? Does it mean that there is
> enough bandwidth to hang 7 x PCI cards in a 7 x Slot Magma chassis and
> not have to worry about IRQ sharing, especially if there is a PCIe
> graphics card in the equation?........
Yes all current PCI chipsets from Intel and AMD do this to some extant.
The 975 and 500x series Xeons ones seem to do a better job of it than
the 965 but I'm not sure if it because the south bridge connectivity or
other factors in the chipset design. There are also devices like this
that will keep your PCI free for your wacky needs.
http://www.gefanucembedded.com/products/774
This compatible with RME cards and UADs.
> .and of course, something like this gets the hampsters jumping on the
> wheel in my brain and my brain comes up with a way to potentially make
> a computer perform an unnatural act by using a PCI video card on the
> PCI bus since that bus doesn't share with anything else, thus
> eliminating any possibility of problems down the road with a graphics
> card.
Yes the PCI buss holds many of the devices on the south bridge, i.e,
firewire, usb, and many 3rd party sata controllers on the majority of
motherboard.
IRQ sharing still have to happen there are a limited number and will at
some point share. Sharing should not be an issue if you are using
current well designed hardware. Devies still use IRQs on the PCIe buss
also. PCI-e video cards work great. The issues seen with PCI-e video
cards was only seen now discontinued chipsets, Intell 915, 925 and
Nvidia Nforce 4 and the new AMR version.
>
> .....the stated conclusion from the other forum........
>
> */The bottom line is that a PCIe device with hard real time requirements
> (like a multiport audio device) will have fewer problems maintaining
> stable audio input and output./*
> *//*
> .......so there is morepotential to poperate a native system at lower
> latencies???
It is already possible on the regular PCI and PCI-x buss on current
motherboards. Becuase of OS/code level and to some extent hardware you
will not see substanually lower latencies than you see now. What you
will see is the possibility of high I/O and other funtionality per card.
>
> C'mon folks, step up to the mic and talk to me here.
>
> ;o)
>
> *
> *
--
Chris Ludwig
ADK
chrisl@adkproaudio.com <mailto:chrisl@adkproaudio.com>
www.adkproaudio.com <http://www.adkproaudio.com/>
(859) 635-5762
|
|
|
|
| Re: Some PCIe questions [message #79349 is a reply to message #79346] |
Wed, 31 January 2007 19:11 |
Deej [4]
Messages: 1292
Registered: January 2007
|
Senior Member |
|
|
Thanks Chris,
> Yes the PCI buss holds many of the devices on the south bridge, i.e,
> firewire, usb, and many 3rd party sata controllers on the majority of
> motherboard.
So there is traffic there after all. Makes sense. I'm just wondering if,
since the PCIe bus is handling the audio/DSP I/O, if the mobo resources
might be best used with the video card on the PCI bus. Maybe the PCIe
bandwidth/architecture is so efficient that this IRQ sharing **** being an
issue will abate in the future.
There are also devices like this
> that will keep your PCI free for your wacky needs.
> http://www.gefanucembedded.com/products/774
> This compatible with RME cards and UADs.
I called these guys today. they quoted me $2075.00 for one. I also talked to
UA tech support about this chassis. It is 64 bit compatible and there are
two busses. slots 1-4 share a buss and 5-7 share a bus. this would seem to
be a no brainer for my situation. I could hang my 4 x UAD-1 cards on the
first 4 slots, my HDSP's on the last three and "in theory" be able to take
advantage of the wider PCI bandwidth. The tests that UA did with these
chassis rated the 7 slot 64 bit Magma as performing poorly. That's a shame
as these are popping up on EBay now for around $1k
> It is already possible on the regular PCI and PCI-x buss on current
> motherboards. Becuase of OS/code level and to some extent hardware you
> will not see substanually lower latencies than you see now. What you will
> see is the possibility of high I/O and other funtionality per card.
I'm wondering if my GA-K8NS Ultra 939 fits into this category of mobos. It's
extremely quick with an AMD 64 4800 x 2. RME and UA both did a mighty fine
job with those new drivers. When Cubase 4 becomes 64 bit, I'm gonna format
myself a new HD and give it a go. I'd like to be able to max out my 4G RAM
potential. I'm not sure if it will be Vista or XP 64 though. There's a lot
of buzz about Vista being audio friendly, but it looks to me like that's
going to be if you use the new Windows driver. Might be a great situation
for Sonar users, but since RME and Cubase are ASIO heavy, it would seem that
these whizbang advantages might not apply.
;o)
Deej
"Chris Ludwig" <chrisl@adkproaudio.com> wrote in message
news:45c15231$1@linux...
> Hey Deej,
>
>
>
> DJ wrote:
>
>> So I hear the PCIe bus can handle 500MHz whereas our wimpy ass PCI busses
>> can only go 133.
>> I also hear that the PCIe bus doesn't share any bandwidth at all with
>> the PCI bus.
>
>
>
> http://www.pcisig.com/specifications/pciexpress/
>
>
>
>> Taken from a post on another forum:
>> /*PCIe guarantees a certain amount of undivided
>> bandwidth to a device in the PCIe slot. It's a dedicated channel. On a
>> modern Intel Chipset like the 965 there will be some contention between
>> devices on the south bridge chip (USB2, LAN, Serial ATA disks, PCIe) but
>> the bandwidth between the north and south bridge chips is quite high
>> relative to the bandwidth of the south bridge devices.*/
>> /**/ So I'm wondering what this means exactly? Does it mean that there is
>> enough bandwidth to hang 7 x PCI cards in a 7 x Slot Magma chassis and
>> not have to worry about IRQ sharing, especially if there is a PCIe
>> graphics card in the equation?........
>
>
> Yes all current PCI chipsets from Intel and AMD do this to some extant.
> The 975 and 500x series Xeons ones seem to do a better job of it than the
> 965 but I'm not sure if it because the south bridge connectivity or other
> factors in the chipset design. There are also devices like this that will
> keep your PCI free for your wacky needs.
> http://www.gefanucembedded.com/products/774
> This compatible with RME cards and UADs.
>
>> .and of course, something like this gets the hampsters jumping on the
>> wheel in my brain and my brain comes up with a way to potentially make a
>> computer perform an unnatural act by using a PCI video card on the PCI
>> bus since that bus doesn't share with anything else, thus eliminating any
>> possibility of problems down the road with a graphics card.
>
> Yes the PCI buss holds many of the devices on the south bridge, i.e,
> firewire, usb, and many 3rd party sata controllers on the majority of
> motherboard.
>
> IRQ sharing still have to happen there are a limited number and will at
> some point share. Sharing should not be an issue if you are using current
> well designed hardware. Devies still use IRQs on the PCIe buss also. PCI-e
> video cards work great. The issues seen with PCI-e video cards was only
> seen now discontinued chipsets, Intell 915, 925 and Nvidia Nforce 4 and
> the new AMR version.
>
>> .....the stated conclusion from the other forum........
>> */The bottom line is that a PCIe device with hard real time requirements
>> (like a multiport audio device) will have fewer problems maintaining
>> stable audio input and output./*
>> *//* .......so there is morepotential to poperate a native system at
>> lower latencies???
>
> It is already possible on the regular PCI and PCI-x buss on current
> motherboards. Becuase of OS/code level and to some extent hardware you
> will not see substanually lower latencies than you see now. What you will
> see is the possibility of high I/O and other funtionality per card.
>
>> C'mon folks, step up to the mic and talk to me here.
>> ;o)
>> *
>> *
>
>
> --
> Chris Ludwig
> ADK
> chrisl@adkproaudio.com <mailto:chrisl@adkproaudio.com>
> www.adkproaudio.com <http://www.adkproaudio.com/>
> (859) 635-5762
|
|
|
|
Goto Forum:
Current Time: Wed May 01 06:24:02 PDT 2024
Total time taken to generate the page: 0.02209 seconds
|
The PARIS Forums
Search
Help
Members
Register
Login
Home
