28 Apr
2014
28 Apr
'14
6:31 a.m.
I have finally upgraded my music computer. I started with an older P4 with
2.5G memory. It was working well and able to get good latency (sub ms by
qjackctl at 48000) but didn't have the guts to do much with it :) it had a
single core.
New MB: asus z87-k
CPU: intel i5 at 3.2G
I could have gotten an i7, but same speed/cores plus
hyperthreading. Hyperthreading is anti low latency. so
the i5 gives the same performance with less power use.
Memory: 8G, all one stick, which is slower than two 4G sticks...
maybe not the best choice, but seems ok.
Video: on board, dual monitors - VGA plus DVI both 1600x900
Disk: 2TB sata about 5 years old (as I recall)
Audio interface: Delta 66
Midi interface: ensoniq pci audio card.
I chose this MB because it had three pci slots as well as 4 pcie slots
which allowed me to keep using my old audio interfaces while giving me
possibilities for upgrades.
ASUS was nice enough to list the irq layout of the pci(e) slots in the
user guide. All three PCI slots are shared with something else :P
However, one is shared with a single channel PCIe slot - good! That PCIe
slot will not be used. That is where the delta 66 went. One is shared with
three PCIe slots and "IGD" (graphics?)... skip that one. The last is
shared with the SATA controller and that is where the ensoniq went. The
internal audio has its own irq (of course) but that is turned off in bios
(as useless). The USB setup uses 3 irqs (two for USB2 and one for USB3)
and so may do well with USB audio... if I had anything other than a USB1.1
audio device to try ;) I also turned the speed "Boost" off in bios as it
can not be controlled by the OS.
So far things look very good. With almost no changes at all and
3.13.0-24-lowlatency #46-Ubuntu for kernel (fresh ubuntustudio 64bit
14.04 install. I can run the D66 at 16 frames with very few xruns and all
are spurious. Just to be sure the ensoniq was not interfering I modified
the rtirq order to:
"rtc snd_ice1712 snd usb i8042"
This did make some small difference. Next, I turned cpufreq from ondemand
to userspace so I could play with different speeds. I set cpu0 to max
(3.2G) and assumed the others would follow all being on the same chip. I
was wrong and the other cores were actually at 800 Mhz (were ondemand had
left them). All of my xruns vanished. Ondemand sucks. I obviously wasn't
pushing this too hard (it would have been too hard on the old P4). I was
running jack at 16 frames (2 periods) with the pulse to jack bridge
running at the same time playing a video full screen too. Qjackctl said
the DSP load was less than 10% so I will have to find some way of loading
it up some more.
Things to try (to load the system):
- mostly MIDI stuff.
- keyboard in
- some HW synths I have around here
- soft synths... lots of them at once.
- IDJC with:
- live mic
- skype through pulse to idjc
- some mp3 or ogg files playing (at least two at once)
- maybe mumble at the same time too or more than one skype.
I will monitor the cpu temperature with different user set CPU speeds.
So far I am very happy with what I have seen. PCI is certainly not dead,
though it is getting harder to find. I could not find any Xeon boards with
PCI slots for example... even if I had wanted to pay the extra.
qjackctl gets loaded at session start and is setup to start jack and
a2jmidid when it runs. Pulse starts then too. Pulse is set up to have all
audio interfaces turned off so that jacksink/source is always default. For
all the bad things I have heard about pulseaudio... it just works.
--
Len Ovens
www.ovenwerks.net
28 Apr
28 Apr
9:29 a.m.
So far I am very happy with what I have seen. PCI is
certainly not dead,
though it is getting harder to find. I could not find any Xeon boards with
PCI slots for example... even if I had wanted to pay the extra.
In my company we have many Dell computers with Xeon processors and PCI slots.
for examples models : Dell precision T3500, T1700, T3600
As they are "professional" computer, high price comes with it. The
latest models made by Dell still have one PCI slot.
Could be found second hand I think.
2:04 p.m.
On Mon, 28 Apr 2014 09:29:47 +0200
Raphaël Mouneyres <rmouneyres(a)gmail.com> wrote:
> So far I am very happy with what I have seen. PCI
is certainly not dead,
> though it is getting harder to find. I could not find any Xeon boards with
> PCI slots for example... even if I had wanted to pay the extra.
SuperMicro still makes them - the company I work for was until recently
selling storage servers based on them.
I recently upgraded my HW as well, and found a Gigabyte GA-990FXA-UD3 mobo
with a PCI slot - but only one, so I finally had to give up my paired
Delta 1010s and only run one. (I've migrated to an Alesis Firewire mixer
as my primary interface anyway, it doesn't affect my workflow.)
For the CPU, I put in an 8 core, 4GHz AMD FX(tm)-8350. There's 2 Corsair
Vengeance 4GB 1866MHz RAM sticks in it - this thing compiles a new kernel
in about 10 minutes! Since I'm itemizing, there's 2 120GB SSDs mirrored for
my OS disk, a pair of 1TB drives mirrored as my main storage and a 128GB
SSD as my primary recording storage, backed up to the 1TB drives.
I like the overall setup a lot - I was amazed how slick the mobo looks.
I'm not one to say hardware's sexy but this mobo looks great! Two of my
main criteria for the mobo choice were a FW port and a PCI slot. This
board's got them, but overall it was not a good experience getting Arch
to run on it. I had to essentially choose between having the eSATA and
USB 2.0 ports working or having the onboard Realtek network port mostly
work. The driver for this particular interface is not well-supported and I
had trouble keeping the interface up.
I eventually disabled the interface and dropped an Intel card I had on hand
in, and configured the mobo so that the eSATA and USB 3.0 ports work. Since
then it's been behaving, but I wish I'd done a little more research on its
Linux compatibility before I bought it.
--
======================================================================
Joe Hartley - UNIX/network Consultant - jh(a)brainiac.com
Without deviation from the norm, "progress" is not possible. - FZappa
11:09 p.m.
On Mon, 28 Apr 2014, Joe Hartley wrote:
I eventually disabled the interface and dropped an
Intel card I had on hand
in, and configured the mobo so that the eSATA and USB 3.0 ports work. Since
then it's been behaving, but I wish I'd done a little more research on its
Linux compatibility before I bought it.
One of the companies that sells computers with linux preinstalled
(mentioned on this list within the month) uses one out of this line of MB.
So that was a good sign to start with. I picked one with more PCI slots.
--
Len Ovens
www.ovenwerks.net
10:55 p.m.
On Mon, 28 Apr 2014, Raphaël Mouneyres wrote:
Thank you. I wanted at least two slots, but preferably more so I could
pick and choose what else is using the same irq. If money was no option...
I would get such a system and add a new audio card to the mix. I would
probably end up around 5K instead of < 1K.
--
Len Ovens
www.ovenwerks.net
So far I am
very happy with what I have seen. PCI is certainly not dead,
though it is getting harder to find. I could not find any Xeon boards with
PCI slots for example... even if I had wanted to pay the extra.
In my company we have many Dell computers with Xeon processors and PCI slots.
for examples models : Dell precision T3500, T1700, T3600
As they are "professional" computer, high price comes with it. The
latest models made by Dell still have one PCI slot.
Could be found second hand I think.
10:58 p.m.
On 28 April 2014 at 13:48, Len Ovens <len(a)ovenwerks.net> wrote:
Thank you. I wanted at least two slots, but preferably
more so I
could pick and choose what else is using the same irq. If money was
no option... I would get such a system and add a new audio card to
the mix. I would probably end up around 5K instead of < 1K.
I ended up getting a PCIe to PCI translator card. It works well, but
it's a little flimsy. I need two PCI slots, one for each of my Delta
1010s.
--
Kevin
29 Apr
29 Apr
2:26 a.m.
On Mon, 28 Apr 2014, Kevin Cosgrove wrote:
On 28 April 2014 at 13:48, Len Ovens
<len(a)ovenwerks.net> wrote:
I want to get one of those just to try out. I would have to modify the
case so I could access the s/pdif port which I use for extra inputs.
However, it sounds like they work just fine for you anyway.
--
Len Ovens
www.ovenwerks.net
Thank you. I wanted at least two slots, but
preferably more so I
could pick and choose what else is using the same irq. If money was
no option... I would get such a system and add a new audio card to
the mix. I would probably end up around 5K instead of < 1K.
I ended up getting a PCIe to PCI translator card. It works well, but
it's a little flimsy. I need two PCI slots, one for each of my Delta
1010s. 
28 Apr
28 Apr
9:52 a.m.
Len Ovens wrote:
Memory: 8G, all one stick, which is slower than two 4G
sticks...
maybe not the best choice, but seems ok.
I doubt that the difference is measurable, let alone noticeable.
In any case, this allows easier upgrading to 16 GB. :)
Disk: 2TB sata about 5 years old (as I recall)
Nowadays, not using an SSD tends to be the bottleneck.
(But this is probably not the case for streaming audio.)
I chose this MB because it had three pci slots
The PCIe/PCI bridges built into chipsets are reliable, but motherboards
that have a separate chip for this tend to use cheap crap.
Your ASM1083 is infamous. Revision 1 was unusably buggy; rev. 3 mostly
works (except for virtualization). Which one do you have (lspci)?
ASUS was nice enough to list the irq layout of the
pci(e) slots in the
user guide.
"Nice" isn't enough, it should be _correct_ (which was not the case for
my last Asus manual).
Anyway, PCIe devices that support MSI use a completely separate
interrupt line. (USB3 ports, graphics, and built-in audio should be
harmless then.)
You need to check /proc/interrupts to be sure.
Regards,
Clemens
1:17 p.m.
On Monday 28 April 2014 07:11:52 Clemens Ladisch did opine:
Cheers, Gene
--
"There are four boxes to be used in defense of liberty:
soap, ballot, jury, and ammo. Please use in that order."
-Ed Howdershelt (Author)
Genes Web page <http://geneslinuxbox.net:6309/gene>
US V Castleman, SCOTUS, Mar 2014 is grounds for Impeaching SCOTUS
Len Ovens wrote:
Ye,s its measurable as it disallows the alternate chip access mode, cutting
the bandwidth in and out of memory in half.
Memory: 8G, all one stick, which is slower than
two 4G sticks...
maybe not the best choice, but seems ok.
I doubt that the difference is measurable, let alone noticeable.
In any case, this allows easier upgrading to 16 GB. :) Disk: 2TB sata about 5 years old (as I recall)
Nowadays, not using an SSD tends to be the bottleneck.
(But this is probably not the case for streaming audio.)
I chose this MB because it had three pci slots
The PCIe/PCI bridges built into chipsets are reliable, but motherboards
that have a separate chip for this tend to use cheap crap.
Your ASM1083 is infamous. Revision 1 was unusably buggy; rev. 3 mostly
works (except for virtualization). Which one do you have (lspci)?
ASUS was nice enough to list the irq layout of
the pci(e) slots in the
user guide.
"Nice" isn't enough, it should be _correct_ (which was not the case for
my last Asus manual).
Anyway, PCIe devices that support MSI use a completely separate
interrupt line. (USB3 ports, graphics, and built-in audio should be
harmless then.)
You need to check /proc/interrupts to be sure.
Regards,
Clemens
_______________________________________________
Linux-audio-user mailing list
Linux-audio-user(a)lists.linuxaudio.org
http://lists.linuxaudio.org/listinfo/linux-audio-user
7:47 p.m.
On 28 April 2014 at 9:52, Clemens Ladisch <clemens(a)ladisch.de> wrote:
Len Ovens wrote:
I use 7200rpm SATA drives with Ardour and routinely run 24 track, 24
bit, 48kHz without any drive speed issues. I have 16GB of RAM, which
I'm sure is helping.
--
Kevin
Disk: 2TB sata about 5 years old (as I recall)
Nowadays, not using an SSD tends to be the bottleneck.
(But this is probably not the case for streaming audio.)
11:04 p.m.
On Mon, 28 Apr 2014, Clemens Ladisch wrote:
I have to be careful where I spend my money... so one thing at a time.
What I have is already a very big step up from what I had.
PCI bridge: Intel Corporation 8 Series/C220 Series Chipset Family PCI
Express Root Port #3 (rev d5)
00:1c.3 PCI bridge: Intel Corporation 82801 PCI Bridge (rev d5)
I am not sure which is which. The first may be the express port.
Most of my MB manuals so far have had nothing and so I have had to try
various slots and see what I get. In any case cat /proc/interupts tells me
I have two clear irqs for my two PCI cards.
Len Ovens wrote:
That was my thought too.
Memory: 8G, all one stick, which is slower than
two 4G sticks...
maybe not the best choice, but seems ok.
I doubt that the difference is measurable, let alone noticeable.
In any case, this allows easier upgrading to 16 GB. :) Disk: 2TB sata
about 5 years old (as I recall)
Nowadays, not using an SSD tends to be the bottleneck.
(But this is probably not the case for streaming audio.) I chose this
MB because it had three pci slots
The PCIe/PCI bridges built into chipsets are reliable, but motherboards
that have a separate chip for this tend to use cheap crap.
Your ASM1083 is infamous. Revision 1 was unusably buggy; rev. 3 mostly
works (except for virtualization). Which one do you have (lspci)? ASUS was nice
enough to list the irq layout of the pci(e) slots in the
user guide.
"Nice" isn't enough, it should be _correct_ (which was not the case for
my last Asus manual). Anyway, PCIe devices that support MSI use a completely
separate
interrupt line. (USB3 ports, graphics, and built-in audio should be
harmless then.)
I have no PCIe devices at this time.
--
Len Ovens
www.ovenwerks.net
30 Apr
30 Apr
10:23 a.m.
Len Ovens wrote:
PCI bridge: Intel Corporation 8 Series/C220 Series
Chipset Family PCI Express Root Port #3 (rev d5)
00:1c.3 PCI bridge: Intel Corporation 82801 PCI Bridge (rev d5)
I am not sure which is which. The first may be the express port.
The ASM1083 (which the manual says you have) should be listed as "ASM1083".
What is the output of "lspci -t"?
On Mon, 28 Apr 2014, Clemens Ladisch wrote:
All your motherboard devices are likely to be PCIe.
Regards,
Clemens
Anyway, PCIe devices that support MSI use a
completely separate
interrupt line. (USB3 ports, graphics, and built-in audio should be
harmless then.)
I have no PCIe devices at this time.
1 May
1 May
1:49 a.m.
On Wed, 30 Apr 2014, Clemens Ladisch wrote:
Not what I meant :) (I was meaning only that I had nothing in the slots)
So it sounds like you are saying that the PCIe devices are assigned an irq
to be compatible with the PCI standard, but most PCIe devices don't use
the irq having another way to get the OS attention. Looking at
/proc/interrupts, I see that ahci (which from lspci seems to be the SATA
controller) shows irq 42 even though the book says it is sharing an irq
with PCI slot 1 which has an ens1370 at irq 19. So I would guess that this
irq 42 is a "soft" irq? (and why do I have an ISA bridge?) So that would
follow that irq 41 to 45 would be onboard soft irqs... They are called
PCI-MSI-edge.
I would guess that if I installed a PCIe card that was PCI HW with a
PCIe->PCI bridge on the card to update it, that it would still use the old
irq and could clash with one of my other PCI audio cards. I can't see me
getting anything other than an ide, audio or firewire card. (I have a
number of ide drives around, but may instead put them in the server and
mount over network)
This is starting to make sense. I have seen a number of people have xrun
trouble as soon as they got new a MB that were faster than trouble free
slower/older HW, but my experience has been totally the opposite. The only
thing that has given any xruns so far as I can see is cpu speed changes
when using ondemand (I already don't have hyperthreading to worry about).
I did choose my MB/CPU based on what I felt would be good for audio, not
for a fast desktop (though I have gained there too). For example, I chose
the i5 cpu over the i7 because I didn't want hyprthreading, not because it
was cheaper.
--
Len Ovens
www.ovenwerks.net
Len Ovens wrote:
len@music:~$ lspci -t
-[0000:00]-+-00.0
+-01.0-[01]--
+-02.0
+-14.0
+-16.0
+-1a.0
+-1c.0-[02]--
+-1c.2-[03]----00.0
+-1c.3-[04-05]----00.0-[05]--+-00.0
| \-02.0
+-1d.0
+-1f.0
+-1f.2
\-1f.3
00:1c.3 would be the PCI bridge then as my two audio cards are listed as
05:00.0 and 05:02.0. I would guess that the third PCI slot would be 01. I
see that the bridge is listed as [04-05] I am guessing that means 04 is
the PCIe side and 05 is the PCI side? 04 is:
04:00.0 PCI bridge: ASMedia Technology Inc. ASM1083/1085 PCIe to PCI
Bridge (rev 03)
PCI bridge: Intel Corporation 8 Series/C220
Series Chipset Family PCI Express Root Port #3 (rev d5)
00:1c.3 PCI bridge: Intel Corporation 82801 PCI Bridge (rev d5)
I am not sure which is which. The first may be the express port.
The ASM1083 (which the manual says you have) should be listed as "ASM1083".
What is the output of "lspci -t"? On Mon, 28 Apr
2014, Clemens Ladisch wrote:
All your motherboard devices are likely to be PCIe. Anyway, PCIe devices that support MSI use a
completely separate
interrupt line. (USB3 ports, graphics, and built-in audio should be
harmless then.)
I have no PCIe devices at this time.
9:53 a.m.
Len Ovens wrote:
On Wed, 30 Apr 2014, Clemens Ladisch wrote:
Device 00:1c.3 is one of the PCIe ports of the southbridge.
Device 04:00 is the ASM1083 PCIe/PCI bridge.
Devices 05:xx are the chips on the PCI cards.
Bus 00: is the "root complex", i.e., CPU/northbridge/southbridge.
Bus 04: is the PCIe connection between the southbridge and the ASM1083.
Bus 05: is the PCI bus.
"[04-05]" means that both buses are reachable through this bridge.
Len Ovens wrote:
len@music:~$ lspci -t
-[0000:00]-+-00.0
+-01.0-[01]--
+-02.0
+-14.0
+-16.0
+-1a.0
+-1c.0-[02]--
+-1c.2-[03]----00.0
+-1c.3-[04-05]----00.0-[05]--+-00.0
| \-02.0
+-1d.0
+-1f.0
+-1f.2
\-1f.3
00:1c.3 would be the PCI bridge then as my two audio cards are listed
as 05:00.0 and 05:02.0. I would guess that the third PCI slot would be
01. I see that the bridge is listed as [04-05] I am guessing that
means 04 is the PCIe side and 05 is the PCI side? PCI bridge: Intel Corporation 8 Series/C220
Series Chipset Family PCI Express Root Port #3 (rev d5)
00:1c.3 PCI bridge: Intel Corporation 82801 PCI Bridge (rev d5)
I am not sure which is which. The first may be the express port.
The ASM1083 (which the manual says you have) should be listed as "ASM1083".
What is the output of "lspci -t"? So it sounds like you are saying that the PCIe devices
are assigned
an irq to be compatible with the PCI standard, but most PCIe devices
don't use the irq having another way to get the OS attention.
There are different kinds of interrupts. 0-15 are ISA-compatible
interrupts; the next four or eight can be used for PCI interrupts.
(The interrupt controller (IOAPIC) has only a limited number of
interrupt lines.)
Message-signaled interrupts (MSIs) are implemented with a different
mechanism and do not need to go through the IOAPIC; you can have any
number of them, and they are never shared between devices.
Looking at /proc/interrupts, I see that ahci (which
from lspci seems
to be the SATA controller) shows irq 42 even though the book says it
is sharing an irq with PCI slot 1 which has an ens1370 at irq 19. So
I would guess that this irq 42 is a "soft" irq?
Yes, it's an MSI.
(and why do I have an ISA bridge?)
Because you have ISA devices. (Intel likes to call them "legacy", but
they are still there.)
I would guess that if I installed a PCIe card that was
PCI HW with
a PCIe->PCI bridge on the card to update it, that it would still use
the old irq and could clash with one of my other PCI audio cards.
Yes.
Regards,
Clemens
11:25 p.m.
On Mon, 28 Apr 2014, Ivan K wrote:
Isn't M-Audio going to have to eventually make
versions of
their sound cards for PCI-e ? What are they waiting for?
There are very few cards for pcie. M-audio appears to have decided USB/FW
is the safe way to go. Their line of PCI cards seems to depend on what is
left in stock, there have been no delta 1010s or delta 44s for a few years
now. The are still 1010LTs and D66s, but I think once they are gone they
will make no more "internal" cards. The reality is that even with an
internal card, a BOB or cable is needed. With USB or FW the BOB becomes
the whole IF and there is less digital noise to deal with.
My opinion is that the "cheap" option will be USB... with (possible)
problems :P and then FW which means adding a FW card anyway, but better
stability and expansion. PCIe cards will be the high end costing more than
the rest of the machine for most of us. (1K plus for 8 i/o at line level
with no mic pre)
Linux support for new HW seems to be on the wane too. Used to be all audio
was addon cards... and the internal audio when it did show up (AC97) was
poor quality. So there were a lot of "ordinary" users who still added
cards. The audio quality of the HDA chip set has got to the point most
people can't tell anything better, and those who can have HDMI out in
digital format anyway. so there is now a lot less push to support
everything out there. The audio crowd is a pretty small segment of the
Linux user crowd anymore. On the up side, there are some more IF makers
who are putting out linux drivers.
--
Len Ovens
www.ovenwerks.net
3657
days inactive
3660
days old
linux-audio-user@lists.linuxaudio.org
15 comments
7 participants
participants (7)
-
Clemens Ladisch -
Gene Heskett -
Ivan K -
Joe Hartley -
Kevin Cosgrove -
Len Ovens -
Raphaël Mouneyres