f



SCSI adapter PCI bandwidth usage

I have an Adaptec 19160 SCSI controller with a SCSI Ultra160 HDD (LVD) and a
SCSI CD writer (SCSI-3)
I'd like to know if the SCSI controller "reserve" the PCI bandwidth (set at
160 for the controller and the HDD and 20 for the CD writer). Which would
means that even if the HDD is not reading or writing anything, the
controller still use the total PCI bandwidth (160) of the SCSI controller.
I have a doubt about that because I had some drop outs while recording audio
with a sound card with the SCSI controller and the HDD set at 160. I lowered
the value to 80 for both of them and there was no more audio drop outs
during recording. That make me think that the drop outs were directly
related with PCI bus congestion. Since I already have several other PCI
devices (firewire, USB, LAN, sound card) in my system, I really need to save
as much PCI bandwidth as possible.
Lowering the value to 80 (instead of 160) for my HDD and the controller
doesn't affect the audio recording (it's still fast enough for that
purpose). But for other applications, I still loose some HDD speed
(according to HDD benchmarks, mainly for the burst speed)
I also would like to know if disabling write caching (or not) of the HDD has
any effects on the PCI bandwidth (use more or less?)

Thanks for any advice and comments!


0
Cyrille
4/30/2006 7:00:15 PM
comp.periphs.scsi 2473 articles. 0 followers. l_x2828 (153) is leader. Post Follow

13 Replies
404 Views

Similar Articles

[PageSpeed] 13

On Mon, 1 May 2006 03:00:15 +0800, "Cyrille Briegel" <wanhua69@ms56.hinet.net>
wrote:

>I have an Adaptec 19160 SCSI controller with a SCSI Ultra160 HDD (LVD) and a
>SCSI CD writer (SCSI-3)
>I'd like to know if the SCSI controller "reserve" the PCI bandwidth (set at
>160 for the controller and the HDD and 20 for the CD writer). Which would
>means that even if the HDD is not reading or writing anything, the
>controller still use the total PCI bandwidth (160) of the SCSI controller.
>I have a doubt about that because I had some drop outs while recording audio
>with a sound card with the SCSI controller and the HDD set at 160. I lowered
>the value to 80 for both of them and there was no more audio drop outs
>during recording. That make me think that the drop outs were directly
>related with PCI bus congestion. Since I already have several other PCI
>devices (firewire, USB, LAN, sound card) in my system, I really need to save
>as much PCI bandwidth as possible.
>Lowering the value to 80 (instead of 160) for my HDD and the controller
>doesn't affect the audio recording (it's still fast enough for that
>purpose). But for other applications, I still loose some HDD speed
>(according to HDD benchmarks, mainly for the burst speed)
>I also would like to know if disabling write caching (or not) of the HDD has
>any effects on the PCI bandwidth (use more or less?)
>
>Thanks for any advice and comments!

There is no mechanism for controlling PCI bus bandwidth to individual agents.
0
daytripper
4/30/2006 8:47:15 PM
On Mon, 1 May 2006 03:00:15 +0800, "Cyrille Briegel"
<wanhua69@ms56.hinet.net> wrote:

>I have an Adaptec 19160 SCSI controller with a SCSI Ultra160 HDD (LVD) and a
>SCSI CD writer (SCSI-3)
>I'd like to know if the SCSI controller "reserve" the PCI bandwidth (set at
>160 for the controller and the HDD and 20 for the CD writer). Which would
>means that even if the HDD is not reading or writing anything, the
>controller still use the total PCI bandwidth (160) of the SCSI controller.
>I have a doubt about that because I had some drop outs while recording audio
>with a sound card with the SCSI controller and the HDD set at 160. I lowered
>the value to 80 for both of them and there was no more audio drop outs
>during recording. That make me think that the drop outs were directly
>related with PCI bus congestion. Since I already have several other PCI
>devices (firewire, USB, LAN, sound card) in my system, I really need to save
>as much PCI bandwidth as possible.
>Lowering the value to 80 (instead of 160) for my HDD and the controller
>doesn't affect the audio recording (it's still fast enough for that
>purpose). But for other applications, I still loose some HDD speed
>(according to HDD benchmarks, mainly for the burst speed)
>I also would like to know if disabling write caching (or not) of the HDD has
>any effects on the PCI bandwidth (use more or less?)
>
>Thanks for any advice and comments!
>

You have liberally mixed SCSI bus speed with PCI bus speed. The two
are only slightly related. If your PCI bus can't keep up, it is
probably that the computer just doens't have the 'horsepower' to do
the job. 

Although there are some priorities on the PCI bus, this is not
generally configurable by the user.

Regardless if both the drives and the CDR drive are on the scsi bus,
then PCI has nothgin to do with it...
0
PeterD
5/1/2006 12:40:42 AM
My system is an AMD Athlon64 3500+ (runs at 2.2 Ghz) with 1GB corsair DDR400
(dual channel) memory (CAS 2.5). The MB also has 2 PCI buses (bus 0 and bus
1). The "horsepower" should be sufficient enough I think.
I'm sorry, I got confused with SCSI bus and PCI bus. But because the SCSI
controller is on the PCI bus, the data going in and out the SCSI bus is also
going necessarily through the PCI bus which has a limited bandwidth (about
133 MB/s I think).
I know only very little about SCSI, so correct me if I'm wrong!

Thanks for your information!
Cyrille

"PeterD" <peter2@hipson.net> wrote in message
news:a4ma52tub105015bcuua3p17sj942kvo5k@4ax.com...
> On Mon, 1 May 2006 03:00:15 +0800, "Cyrille Briegel"
> <wanhua69@ms56.hinet.net> wrote:
>
> >I have an Adaptec 19160 SCSI controller with a SCSI Ultra160 HDD (LVD)
and a
> >SCSI CD writer (SCSI-3)
> >I'd like to know if the SCSI controller "reserve" the PCI bandwidth (set
at
> >160 for the controller and the HDD and 20 for the CD writer). Which would
> >means that even if the HDD is not reading or writing anything, the
> >controller still use the total PCI bandwidth (160) of the SCSI
controller.
> >I have a doubt about that because I had some drop outs while recording
audio
> >with a sound card with the SCSI controller and the HDD set at 160. I
lowered
> >the value to 80 for both of them and there was no more audio drop outs
> >during recording. That make me think that the drop outs were directly
> >related with PCI bus congestion. Since I already have several other PCI
> >devices (firewire, USB, LAN, sound card) in my system, I really need to
save
> >as much PCI bandwidth as possible.
> >Lowering the value to 80 (instead of 160) for my HDD and the controller
> >doesn't affect the audio recording (it's still fast enough for that
> >purpose). But for other applications, I still loose some HDD speed
> >(according to HDD benchmarks, mainly for the burst speed)
> >I also would like to know if disabling write caching (or not) of the HDD
has
> >any effects on the PCI bandwidth (use more or less?)
> >
> >Thanks for any advice and comments!
> >
>
> You have liberally mixed SCSI bus speed with PCI bus speed. The two
> are only slightly related. If your PCI bus can't keep up, it is
> probably that the computer just doens't have the 'horsepower' to do
> the job.
>
> Although there are some priorities on the PCI bus, this is not
> generally configurable by the user.
>
> Regardless if both the drives and the CDR drive are on the scsi bus,
> then PCI has nothgin to do with it...


0
Cyrille
5/1/2006 5:48:15 AM
I understand, but if I limit the sync transfer rate in the SCSI bios (via
scsiselect) let's say from160 MB/s to 80 MB/s, then it will free up some PCI
bandwidth for other PCI devices (like the sound card). That what I figured
out with my audio recording tests. If I set it back to 160 MB/s for both the
SCSI controller and the SCSI HDD, I will get audio drop outs again. I agree
that I can't control the PCI bandwidth with the other PCI devices but it
seems like it's possible to do it at least through the SCSI controller by
limiting the sync transfer rate of SCSI devices.

That's what I thought logically, but since I know very little about SCSI,
correct me if I'm wrong.

Thanks for your comment!
Cyrille


"daytripper" <day_trippr@REMOVEyahoo.com> wrote in message
news:uh8a52psmb1q7i0mgpmkjp242ffsb3r0ka@4ax.com...
> On Mon, 1 May 2006 03:00:15 +0800, "Cyrille Briegel"
<wanhua69@ms56.hinet.net>
> wrote:
>
> >I have an Adaptec 19160 SCSI controller with a SCSI Ultra160 HDD (LVD)
and a
> >SCSI CD writer (SCSI-3)
> >I'd like to know if the SCSI controller "reserve" the PCI bandwidth (set
at
> >160 for the controller and the HDD and 20 for the CD writer). Which would
> >means that even if the HDD is not reading or writing anything, the
> >controller still use the total PCI bandwidth (160) of the SCSI
controller.
> >I have a doubt about that because I had some drop outs while recording
audio
> >with a sound card with the SCSI controller and the HDD set at 160. I
lowered
> >the value to 80 for both of them and there was no more audio drop outs
> >during recording. That make me think that the drop outs were directly
> >related with PCI bus congestion. Since I already have several other PCI
> >devices (firewire, USB, LAN, sound card) in my system, I really need to
save
> >as much PCI bandwidth as possible.
> >Lowering the value to 80 (instead of 160) for my HDD and the controller
> >doesn't affect the audio recording (it's still fast enough for that
> >purpose). But for other applications, I still loose some HDD speed
> >(according to HDD benchmarks, mainly for the burst speed)
> >I also would like to know if disabling write caching (or not) of the HDD
has
> >any effects on the PCI bandwidth (use more or less?)
> >
> >Thanks for any advice and comments!
>
> There is no mechanism for controlling PCI bus bandwidth to individual
agents.


0
Cyrille
5/1/2006 6:15:09 AM
"PeterD" <peter2@hipson.net> wrote in message news:a4ma52tub105015bcuua3p17sj942kvo5k@4ax.com
> On Mon, 1 May 2006 03:00:15 +0800, "Cyrille Briegel" <wanhua69@ms56.hinet.net> wrote:
> 
> > I have an Adaptec 19160 SCSI controller with a SCSI Ultra160 HDD (LVD) and a
> > SCSI CD writer (SCSI-3)
> > I'd like to know if the SCSI controller "reserve" the PCI bandwidth (set at
> > 160 for the controller and the HDD and 20 for the CD writer). Which would
> > means that even if the HDD is not reading or writing anything, the
> > controller still use the total PCI bandwidth (160) of the SCSI controller.
> > I have a doubt about that because I had some drop outs while recording audio
> > with a sound card with the SCSI controller and the HDD set at 160. I lowered
> > the value to 80 for both of them and there was no more audio drop outs
> > during recording. That make me think that the drop outs were directly
> > related with PCI bus congestion. Since I already have several other PCI
> > devices (firewire, USB, LAN, sound card) in my system, I really need to save
> > as much PCI bandwidth as possible.
> > Lowering the value to 80 (instead of 160) for my HDD and the controller
> > doesn't affect the audio recording (it's still fast enough for that
> > purpose). But for other applications, I still loose some HDD speed
> > (according to HDD benchmarks, mainly for the burst speed)
> > I also would like to know if disabling write caching (or not) of the HDD has
> > any effects on the PCI bandwidth (use more or less?)
> > 
> > Thanks for any advice and comments!
> > 
> 
> You have liberally mixed SCSI bus speed with PCI bus speed. 

> The two are only slightly related. 

Nope.

> If your PCI bus can't keep up, it is probably that the computer just 
> doens't have the 'horsepower' to do the job.

Utter nonsense.

> 
> Although there are some priorities on the PCI bus, this is not
> generally configurable by the user.

> Regardless if both the drives and the CDR drive are on the scsi bus,
> then PCI has nothgin to do with it...

That's not even a sentence.
0
Folkert
5/1/2006 2:24:25 PM
The SCSI adapter does not reserve bandwidth but there could be a lot of disk activity when you don't expect it, e.g. when WinXP frequently reads and writes to the page file. During this time the SCSI adapter, rated at 160 MB/s will saturate the PCI bus, rated at 133 MB/s. Turning off write cache will help the problem because the adapter will wait for a write operation to complete before doing the next read, but this will probably hurt performance more than limiting speed to 80 MB/s.

Cyrille Briegel wrote:
> 
> I have an Adaptec 19160 SCSI controller with a SCSI Ultra160 HDD (LVD) and a
> SCSI CD writer (SCSI-3)
> I'd like to know if the SCSI controller "reserve" the PCI bandwidth (set at
> 160 for the controller and the HDD and 20 for the CD writer). Which would
> means that even if the HDD is not reading or writing anything, the
> controller still use the total PCI bandwidth (160) of the SCSI controller.
> I have a doubt about that because I had some drop outs while recording audio
> with a sound card with the SCSI controller and the HDD set at 160. I lowered
> the value to 80 for both of them and there was no more audio drop outs
> during recording. That make me think that the drop outs were directly
> related with PCI bus congestion. Since I already have several other PCI
> devices (firewire, USB, LAN, sound card) in my system, I really need to save
> as much PCI bandwidth as possible.
> Lowering the value to 80 (instead of 160) for my HDD and the controller
> doesn't affect the audio recording (it's still fast enough for that
> purpose). But for other applications, I still loose some HDD speed
> (according to HDD benchmarks, mainly for the burst speed)
> I also would like to know if disabling write caching (or not) of the HDD has
> any effects on the PCI bandwidth (use more or less?)
> 
> Thanks for any advice and comments!

-- 
                   Mike Walsh
            West Palm Beach, Florida, U.S.A.
0
Mike
5/2/2006 1:43:10 PM
What a load of bullshit. Pagefile IO is random and negligible.

"Mike Walsh" <spamscks@netrox.net> wrote in message news:445761EE.193AC3A5@netrox.net...
>
> The SCSI adapter does not reserve bandwidth but there could be a lot of disk activity when you
don't expect it, e.g. when WinXP frequently reads and writes to the page file. During this time the
SCSI adapter, rated at 160 MB/s will saturate the PCI bus, rated at 133 MB/s. Turning off write
cache will help the problem because the adapter will wait for a write operation to complete before
doing the next read, but this will probably hurt performance more than limiting speed to 80 MB/s.
>


0
Eric
5/2/2006 3:12:10 PM
Actually, I have 3 HDD (2 IDE and 1 SCSI)
I exclusively use the SCSI disk for audio recordings.
The page file is split over the 3 disks. The bigger part are on the 2 IDE
disks. (according to Microsoft, this is the best way to handle page file on
XP. XP always uses first the page file on the disk that have the less
activity. That means when there's heavy writing and reading on the SCSI
disk, XP will use the 2 other disks first for paging)
That's the theory. Now, it doesn't means it's working as well as they say it
does.
I was thinking maybe I should only split the page file over the 2 IDE disks
to make sure no paging is performed on the SCSI disk while recording audio.
Also, for audio recording, it is usually recommended to disable write
caching because the writing is sequential and continuous (the buffer can't
handle properly so much data anyway for that long period of time when write
caching is enable). I have disabled the write cache on the SCSI disk only. I
know by doing so, I slow down the disk performance for other apps but audio
recording only needs the disk to write steadily. The speed is fast enough (I
get about 9 MB/s during sequential writing with write caching disabled). At
that rate I could in theory record more than 32 mono tracks @ 96khz - 24 bit
depth simultaneously. I never record more than 8 tracks at a time.

Thanks for the information. I will try removing the page file from the SCSI
disk and perform further tests with that setting. But I think it's safer to
keep the transfer rate of the SCSI disk and controller limited to 80 MB/s.

Cyrille


"Mike Walsh" <spamscks@netrox.net> wrote in message
news:445761EE.193AC3A5@netrox.net...
>
> The SCSI adapter does not reserve bandwidth but there could be a lot of
disk activity when you don't expect it, e.g. when WinXP frequently reads and
writes to the page file. During this time the SCSI adapter, rated at 160
MB/s will saturate the PCI bus, rated at 133 MB/s. Turning off write cache
will help the problem because the adapter will wait for a write operation to
complete before doing the next read, but this will probably hurt performance
more than limiting speed to 80 MB/s.
>
> Cyrille Briegel wrote:
> >
> > I have an Adaptec 19160 SCSI controller with a SCSI Ultra160 HDD (LVD)
and a
> > SCSI CD writer (SCSI-3)
> > I'd like to know if the SCSI controller "reserve" the PCI bandwidth (set
at
> > 160 for the controller and the HDD and 20 for the CD writer). Which
would
> > means that even if the HDD is not reading or writing anything, the
> > controller still use the total PCI bandwidth (160) of the SCSI
controller.
> > I have a doubt about that because I had some drop outs while recording
audio
> > with a sound card with the SCSI controller and the HDD set at 160. I
lowered
> > the value to 80 for both of them and there was no more audio drop outs
> > during recording. That make me think that the drop outs were directly
> > related with PCI bus congestion. Since I already have several other PCI
> > devices (firewire, USB, LAN, sound card) in my system, I really need to
save
> > as much PCI bandwidth as possible.
> > Lowering the value to 80 (instead of 160) for my HDD and the controller
> > doesn't affect the audio recording (it's still fast enough for that
> > purpose). But for other applications, I still loose some HDD speed
> > (according to HDD benchmarks, mainly for the burst speed)
> > I also would like to know if disabling write caching (or not) of the HDD
has
> > any effects on the PCI bandwidth (use more or less?)
> >
> > Thanks for any advice and comments!
>
> -- 
>                    Mike Walsh
>             West Palm Beach, Florida, U.S.A.


0
Cyrille
5/2/2006 4:52:55 PM
My main home PC has WinNT with three hard drives and it will in fact use the disk with the least activity for most of the paging. My laptop PC with WinXP and one drive will have bursts of disk activity (which I assume is page file activity) at inopportune times which will affect performance. I find it quite annoying that WinNT based OSs have a lot of page file activity even when there is plenty of free memory.

Cyrille Briegel wrote:
> 
> Actually, I have 3 HDD (2 IDE and 1 SCSI)
> I exclusively use the SCSI disk for audio recordings.
> The page file is split over the 3 disks. The bigger part are on the 2 IDE
> disks. (according to Microsoft, this is the best way to handle page file on
> XP. XP always uses first the page file on the disk that have the less
> activity. That means when there's heavy writing and reading on the SCSI
> disk, XP will use the 2 other disks first for paging)
> That's the theory. Now, it doesn't means it's working as well as they say it
> does.

-- 
                   Mike Walsh
            West Palm Beach, Florida, U.S.A.
0
Mike
5/2/2006 7:39:06 PM
In message <445761EE.193AC3A5@netrox.net>, Mike Walsh 
<spamscks@netrox.net> writes
>
>The SCSI adapter does not reserve bandwidth but there could be a lot of 
>disk activity when you don't expect it, e.g. when WinXP frequently 
>reads and writes to the page file. During this time the SCSI adapter, 
>rated at 160 MB/s will saturate the PCI bus, rated at 133 MB/s. Turning 
>off write cache will help the problem because the adapter will wait for 
>a write operation to complete before doing the next read, but this will 
>probably hurt performance more than limiting speed to 80 MB/s.
>
....

"Turning off write cache will help the problem"?
Sounds like the problem is insufficient memory and consequent use of 
swap space!

-- 
Jeremy Boden
0
Jeremy
5/3/2006 12:16:03 PM
"Jeremy Boden" <jeremy@jboden.demon.co.uk> wrote in message news:iaIHjVBD8JWEFwPj@jboden.demon.co.uk
> In message <445761EE.193AC3A5@netrox.net>, Mike Walsh <spamscks@netrox.net> writes
> > 
> > The SCSI adapter does not reserve bandwidth but there could be a lot of
> > disk activity when you don't expect it, e.g. when WinXP frequently
> > reads and writes to the page file. During this time the SCSI adapter,
> > rated at 160 MB/s will saturate the PCI bus, rated at 133 MB/s. Turning
> > off write cache will help the problem because the adapter will wait for
> > a write operation to complete before doing the next read, but this will
> > probably hurt performance more than limiting speed to 80 MB/s.
> > 
> ...
> 
> "Turning off write cache will help the problem"?
> Sounds like the problem is insufficient memory and consequent use of
> swap space!

Yes Jeremy, that's what he said.
0
Folkert
5/3/2006 12:50:03 PM
Could you explain more?

I have 1 GB of ram and page file is never in use during recording. I also
never run apps other than the recording software while recording. The
background services running are only the necessary ones for XP to run
properly (no antivirus scan or so). The task manager indicates that there's
plenty of extra physical memory available when I record (unless if I use
lots of virtual instruments, but that not the case at that moment).
Before, I used to have another system (same SCSI controller and disk, same
sound card and graphic card, but older MB, CPU (PIII 800 Ghz) and RAM (512
MB). I never experienced that problem even with half the ram and 3 times
less CPU power that I have now.
I just try to understand the reason. The only explaination I think of up to
now is that my new MB has more PCI devices than I used to have with my older
system (firewire, 2 USB 2.0, 2 Lan controller and maybe some other stuff I'm
not even aware of). I've already deactivated one of the 2 Lan controller in
the BIOS and other unnecessary ports or devices I never use.
That's why it makes me think of PCI bus congestion problem.
By the way my MB is K8N Neo2 platinum.

Thanks for the feedback!
Cyrille

> "Turning off write cache will help the problem"?
> Sounds like the problem is insufficient memory and consequent use of
> swap space!
>
> -- 
> Jeremy Boden


0
Cyrille
5/3/2006 5:39:15 PM
In message <e3apsi$7ns$1@netnews.hinet.net>, Cyrille Briegel 
<wanhua69@ms56.hinet.net> writes
>Could you explain more?
>
>I have 1 GB of ram and page file is never in use during recording. I also
>never run apps other than the recording software while recording. The
>background services running are only the necessary ones for XP to run
>properly (no antivirus scan or so). The task manager indicates that there's
>plenty of extra physical memory available when I record (unless if I use
>lots of virtual instruments, but that not the case at that moment).
>Before, I used to have another system (same SCSI controller and disk, same
>sound card and graphic card, but older MB, CPU (PIII 800 Ghz) and RAM (512
>MB). I never experienced that problem even with half the ram and 3 times
>less CPU power that I have now.
>I just try to understand the reason. The only explaination I think of up to
>now is that my new MB has more PCI devices than I used to have with my older
>system (firewire, 2 USB 2.0, 2 Lan controller and maybe some other stuff I'm
>not even aware of). I've already deactivated one of the 2 Lan controller in
>the BIOS and other unnecessary ports or devices I never use.
>That's why it makes me think of PCI bus congestion problem.
>By the way my MB is K8N Neo2 platinum.
>
>Thanks for the feedback!
>Cyrille
>
>> "Turning off write cache will help the problem"?
>> Sounds like the problem is insufficient memory and consequent use of
>> swap space!
>>
My comment was intended to show that trying to cause *more* disk I/O (by 
getting rid of the write cache) could never be a sensible solution  - 
indeed it reduces the performance of your whole SCSI subsystem.

What about the effect of the built-in soundcard?
What about the SATA & firewire connections?

-- 
Jeremy Boden
0
Jeremy
5/3/2006 7:54:12 PM
Reply: