Hi..
I've customer who bought Scope Professional for using with Macintosh G4 1.25G Dual CPU, 2GB of Ram. Every time when he changed the system sample rate to 96K, the system they say..
Sorry!
DSP capacity limit reached.
Cause: Require audio transfer bandwidth between DSPs too high.
It's strange because the DSP is used only 10% of 14 DSPs. This makes it impossible to record and work in 96k in his Macintosh computer.
Is there anything I can do to fix this problem?
audio transfer bandwidth between DSPS too high on Macintosh
This is not dependant of OS, mac or pc. There's some memory on each dsp, used for example for look-ahead function, short delay lines or some granular stuff. If samplerate changes to 96, also this memory usage will go *96/44.1.
By trial and error you can detect and remove the device/module causing DSP memory overload. Also re-ordening DSP load could solve this problem, if only 10/14 is in use: For example if on chip2 there's 2 devices allocated that both use the chips memory extensively, reloading DSP so the devices are loaded on different chips (w each their own memory space) can possibly solve the problem too.
This DSP memory usage is not indicated on the DSP usage chart, which only displays the processor and not memory usage. Also we have no control over what is loaded on which chip, we can only hope DSP allocation will be better organised after reloading them.
I've also been trying 96kHz recently, and I gave up cos of these errors. The only appeared when adding a module, not while working on a steady project. 10/14 is often also my max DSP usage at that samplerate.
The DSP overloads and DSP memory problems I saw usually happened on loading silly stuff like MVC, while some wicked test stuff from Adern can be added without problems. This makes no sense at all. Do I need to say I gave up on 96kHz for the time being?
By trial and error you can detect and remove the device/module causing DSP memory overload. Also re-ordening DSP load could solve this problem, if only 10/14 is in use: For example if on chip2 there's 2 devices allocated that both use the chips memory extensively, reloading DSP so the devices are loaded on different chips (w each their own memory space) can possibly solve the problem too.
This DSP memory usage is not indicated on the DSP usage chart, which only displays the processor and not memory usage. Also we have no control over what is loaded on which chip, we can only hope DSP allocation will be better organised after reloading them.
I've also been trying 96kHz recently, and I gave up cos of these errors. The only appeared when adding a module, not while working on a steady project. 10/14 is often also my max DSP usage at that samplerate.
The DSP overloads and DSP memory problems I saw usually happened on loading silly stuff like MVC, while some wicked test stuff from Adern can be added without problems. This makes no sense at all. Do I need to say I gave up on 96kHz for the time being?
more has been done with less
https://soundcloud.com/at0m-studio
https://soundcloud.com/at0m-studio
-
- Posts: 2464
- Joined: Tue Jun 05, 2001 4:00 pm
- Location: Warsaw to Szczecin, Poland
- Contact:
Since your Mac is DP then you can try the two solutions:
#QuickTime bug
#ApplePCI extension
The first one I have done in my system 'just in case' because I don't have Nuendo; the second tip I found somewhere in the net and it works great even in my SingleProcessor Mac. Also you can try to modify the cset.ini. I found this tip in the CWA database (I can't acces it now so instead of link tehre will be explanation): cset.ini can be modified by changing the 'ini' to 'txt' and if it still doesn't get editable with SimpleText then you should find a utility which should change the default application for the very file. So, locate an antry: asyncLoopTime=50 and change it to 100. If you can't find it, create one in [hw] section.
I hope the second solution helps
_________________
Sir samplaire scopernicus
<font size=-1>[ This Message was edited by: samplaire on 2005-03-05 14:37 ]</font>
<font size=-1>[ This Message was edited by: samplaire on 2005-03-05 14:38 ]</font>
#QuickTime bug
#ApplePCI extension
The first one I have done in my system 'just in case' because I don't have Nuendo; the second tip I found somewhere in the net and it works great even in my SingleProcessor Mac. Also you can try to modify the cset.ini. I found this tip in the CWA database (I can't acces it now so instead of link tehre will be explanation): cset.ini can be modified by changing the 'ini' to 'txt' and if it still doesn't get editable with SimpleText then you should find a utility which should change the default application for the very file. So, locate an antry: asyncLoopTime=50 and change it to 100. If you can't find it, create one in [hw] section.
I hope the second solution helps

_________________

<font size=-1>[ This Message was edited by: samplaire on 2005-03-05 14:37 ]</font>
<font size=-1>[ This Message was edited by: samplaire on 2005-03-05 14:38 ]</font>
while the err message clearly points to the facts at0m|c describes above, I wouldn't be surprised at all if THAT special model is capable of adding some extra trouble.
It's one single mess of construction/software crap, starting with case and heat control and (possibly not) ending at RAM requirements and external port support. For example I have an USB stick I can plug in every machine at the office, except the dual G4...
If supplied with wrong speced memory chips, it will not complain, but produce the most wiered software errors one can imagine - in 20 years of hardware experience I've never seen something like that.
It could even be that your customer's error message is false triggered by erroneus memory access (there MUST be a lable with 'Apple' on the chips - just the proper Dimms WILL NOT do it).
Unfortunately it's the last machine capable to boot OS9...
cheers, Tom
It's one single mess of construction/software crap, starting with case and heat control and (possibly not) ending at RAM requirements and external port support. For example I have an USB stick I can plug in every machine at the office, except the dual G4...

If supplied with wrong speced memory chips, it will not complain, but produce the most wiered software errors one can imagine - in 20 years of hardware experience I've never seen something like that.
It could even be that your customer's error message is false triggered by erroneus memory access (there MUST be a lable with 'Apple' on the chips - just the proper Dimms WILL NOT do it).
Unfortunately it's the last machine capable to boot OS9...
cheers, Tom
According to experience of using creamware DSP card, I found the problem that the 14 DSPs card can't work in 96k if the DSPs was loaded more than 30%. Even if in 48k, the card can't work if the DSPs was loaded more than 85%. The error will said like they don't have enough RAM to communicate between each DSPs.
After the real hard testing on the problem, I found that if the system contains with either 6 DSPs card or 3 DSPs card in any combination, this problem will not happen. Says, if I use 3 cards of 6 DSPs (6+6+6 DSPs = 18 DSPs), I can run the DSPs to its limit of 99% in 96k. But if only the 14 DSPs card was used in the system (either standalone 14 DSPs or with any 6 or 3 DSPs card), the problem will come right away. It seems that there might have problem in 14 DSPs card designer.
One point of notice, I used to see the old generation 15 DSPs card, they have more RAM on the board. But new generation 14 DSPs card doesn't have that RAM and 1 DSP was removed. This might affect the performance?
After the real hard testing on the problem, I found that if the system contains with either 6 DSPs card or 3 DSPs card in any combination, this problem will not happen. Says, if I use 3 cards of 6 DSPs (6+6+6 DSPs = 18 DSPs), I can run the DSPs to its limit of 99% in 96k. But if only the 14 DSPs card was used in the system (either standalone 14 DSPs or with any 6 or 3 DSPs card), the problem will come right away. It seems that there might have problem in 14 DSPs card designer.
One point of notice, I used to see the old generation 15 DSPs card, they have more RAM on the board. But new generation 14 DSPs card doesn't have that RAM and 1 DSP was removed. This might affect the performance?
sharc 15 is still there, but not mentioned anymore.
People kept asking for it as it doesn't show up in the DSP-Load window of SFP.
The card needs one DSP for it's own running, and it is not available for the project loads
Just to inform
<font size=-1>[ This Message was edited by: hubird on 2005-03-11 12:09 ]</font>
People kept asking for it as it doesn't show up in the DSP-Load window of SFP.
The card needs one DSP for it's own running, and it is not available for the project loads

Just to inform

<font size=-1>[ This Message was edited by: hubird on 2005-03-11 12:09 ]</font>
beerbr,
it's exactly the same hardware as before. Same memory, same chips too. Some people were moaning that they only saw 14dsp in the DSP meter, so they now announce it as a 14dsp board. The 15th is apparently used for DSP management etc.
it's exactly the same hardware as before. Same memory, same chips too. Some people were moaning that they only saw 14dsp in the DSP meter, so they now announce it as a 14dsp board. The 15th is apparently used for DSP management etc.
more has been done with less
https://soundcloud.com/at0m-studio
https://soundcloud.com/at0m-studio