Scope PCI-card effectivity

A place to talk about whatever Scope music/gear related stuff you want.

Moderators: valis, garyb

Inferman
Posts: 133
Joined: Thu Apr 17, 2008 10:32 pm

Scope PCI-card effectivity

Post by Inferman »

Hi Scope community.

I've decided to buy some DSP Booster for my Creamware Project but I found an article (in russian language), where they say that PCI can't provide such bandwidth to allow whole 14 DSP-card works in full power.

So I have a doubt about it. Have You any information about technical specification of Creamware DSP cards?

I have read on soniccore.com that Z-link is necessary for two Creamware cards because PCI bus cannot manage with such information stream ... but anyway this two cards are in PCI-slots so what's the difference if it would be connected via PCI bus?
User avatar
garyb
Moderator
Posts: 23364
Joined: Sun Apr 15, 2001 4:00 pm
Location: ghetto by the sea

Re: Scope PCI-card effectivity

Post by garyb »

the pci bus can handle the info from 3 14dsp cards, depending on the motherboard and chipset design. this means that most intel chipset motherboards, and many amd boards work just fine.
Inferman
Posts: 133
Joined: Thu Apr 17, 2008 10:32 pm

Re: Scope PCI-card effectivity

Post by Inferman »

Yeah, but what part plays MB and chipset design? As I understand PCI specifications are the same, no matter what chipset is used PCI have its 33 or 66 MHz and 32 or 64 bits, it can vary a little bit because of quality of MB compounds (capacitors, transistors etc.) but not too far I think to handle or do not handle with CW cards. So I think chip and MB series are not too important as theoretical and practical bandwidth of PCI slots and real data stream producing by CW cards.

So the article says that exactly PCI bandwidth is not enough to allow 14 DSP card works fine. Its like insert nVidia GTX 280 into PCI-Express 1.0 slot - it will works but not optimised.

Notice that the new Xite planning to release with eSATA connectors because USB and FireWire are probably too weak with its 480 or even 800 Mbits/sec.
User avatar
Neutron
Posts: 2274
Joined: Sun Apr 29, 2001 4:00 pm
Location: Great white north eh
Contact:

Re: Scope PCI-card effectivity

Post by Neutron »

you can put 2 cards no problem. the DSP communicate with each other by a separate STDM cable, and the PCI bus only uses one of the cards for data. "card0 = "master" of your scope system

you will saturate the PCI bus if you run a lot of delays and stuff. there used to be a motherboard testing thread here called "masteverb test" or something like that

as long as your motherboard is not really stupid with IRQ conflicts (yes they still happen, just hidden these days)
User avatar
garyb
Moderator
Posts: 23364
Joined: Sun Apr 15, 2001 4:00 pm
Location: ghetto by the sea

Re: Scope PCI-card effectivity

Post by garyb »

Inferman wrote:Yeah, but what part plays MB and chipset design? As I understand PCI specifications are the same, no matter what chipset is used PCI have its 33 or 66 MHz and 32 or 64 bits, it can vary a little bit because of quality of MB compounds (capacitors, transistors etc.) but not too far I think to handle or do not handle with CW cards. So I think chip and MB series are not too important as theoretical and practical bandwidth of PCI slots and real data stream producing by CW cards.

So the article says that exactly PCI bandwidth is not enough to allow 14 DSP card works fine. Its like insert nVidia GTX 280 into PCI-Express 1.0 slot - it will works but not optimised.

Notice that the new Xite planning to release with eSATA connectors because USB and FireWire are probably too weak with its 480 or even 800 Mbits/sec.
chipset and mobo design are VERY important. there is the theoretical standard and then there is the implementation and the two are WHOLEY different. i have a 15dsp and two 4 dsp cards and i regulary use them near capacity. the number of dsps and compressors have little effect on bandwidth as most audio processing is done on the card with no need to communicate with the rest of the computer whatsoever. in fact, if the computer crashes, the card continues to operate normally until power is removed. asio channels and reverbs/delays are the bandwidth eaters, and there is plenty of reasonable bandwidth for that in a well designed mobo.

esata connectors are for HDs. the XITE uses pci-e and cardbus.
Inferman
Posts: 133
Joined: Thu Apr 17, 2008 10:32 pm

Re: Scope PCI-card effectivity

Post by Inferman »

garyb wrote:esata connectors are for HDs. the XITE uses pci-e and cardbus.
Yea, I've little bit mistaken, anyway external PCI-e 1.1 bandwidth is equal with eSATA (even a bit more) and PCI-e 2.0 is twice more.
Inferman
Posts: 133
Joined: Thu Apr 17, 2008 10:32 pm

Re: Scope PCI-card effectivity

Post by Inferman »

Neutron, what is the Masterverb test?

I suppose it defines by how many MasterVerb instanses I can use with my CW card?

What standards 4xDSP, 6xDSP and 14xDSP cards have?
User avatar
Neutron
Posts: 2274
Joined: Sun Apr 29, 2001 4:00 pm
Location: Great white north eh
Contact:

Re: Scope PCI-card effectivity

Post by Neutron »

Inferman
Posts: 133
Joined: Thu Apr 17, 2008 10:32 pm

Re: Scope PCI-card effectivity

Post by Inferman »

MSI P45 Platinum
2xCorsair 1066 MHz
P Duo 6750

Only 1xScope Project 4.0 (6 DSP)

9 Masterverbs.

Is it good :) ?
User avatar
garyb
Moderator
Posts: 23364
Joined: Sun Apr 15, 2001 4:00 pm
Location: ghetto by the sea

Re: Scope PCI-card effectivity

Post by garyb »

not incredible, but not bad. :)

can a Scope Project load many more reverbs?
Inferman
Posts: 133
Joined: Thu Apr 17, 2008 10:32 pm

Re: Scope PCI-card effectivity

Post by Inferman »

I think I tested wrong.

I opened MVerbs in mixer and as I saw it must be done in clear Routing Window so when I closed any other windows and opened Masterverbs only the count increased to 10 instances.

Anyway I still interesting about bandwidth problem. Sadly, there is no any CW cards technical specification on soniccre.com.
User avatar
Fede
Posts: 300
Joined: Sat May 05, 2001 4:00 pm
Location: Genoa, Italy

Re: Scope PCI-card effectivity

Post by Fede »

The pci bandwidth problem summarized:

- When pc memory is intensively needed by scope plugins and/or many audio tracks are used the bus may get saturated. This condition is worsened by a bad motherboard chipset (also the whole system performance is affected by a poor chipset in any daw)

- Scope plugins using pc memory are delays, reverbs, sample players. Reverbs being the killers most demanding ones.

- To be affected by the problem you are to use more than 10 reverbs together. I've heard of some scope reverb plugins from Sonic Timeworks that are so pci demanding to saturate the bus with fewer instances than Masterverb, being those super-extra-detailed-hi-quality ones (someone here may confirm).

- The same limitations affected ProTools, TC PowerCore and all the other dsp platforms that used pci cards before the advent of PCIe. The future is of course PCIe (see XITE stuff)


Practically, it is more likely to run out of dsp power than of pci bandwidth on a 6 dsp card. Does a mix need all those reverbs? I personally haven't ever encountered such a situation in real work :)

Fede
Inferman
Posts: 133
Joined: Thu Apr 17, 2008 10:32 pm

Re: Scope PCI-card effectivity

Post by Inferman »

"Practically, it is more likely to run out of dsp power than of pci bandwidth on a 6 dsp card. Does a mix need all those reverbs? I personally haven't ever encountered such a situation in real work"

Fede, You are right about reverbs but if I only think to use some CW synths I need some more power than 6 DSP (for example 8 voice of synth syn-chrome or another 3rd party plugin kills whole 6 DSP power at once).
User avatar
Fede
Posts: 300
Joined: Sat May 05, 2001 4:00 pm
Location: Genoa, Italy

Re: Scope PCI-card effectivity

Post by Fede »

Yes of course. You'll want to have more dsps for those synths and effects too.

To be clear: the supposed waste, referring to the article you quoted, is that, even with enough dsps for running 20 or more reverbs, you'll end up using as many as the motherboard chipset will permit. However, the dsp power left is not lost! it *will* be available for synths or other effects.
I don't see the Nvidia comparison much appropriated: graphic cards use all the power of the gpu in a sort of "one way" usage (fps). That suggests that in audio world things are similar, but it is not quite true unless the processing power is used only for reverbs.

Btw Z-link has nothing to do with this, it is a proprietary digital connection to send/receive audio channels among creamware products.

Fede
Inferman
Posts: 133
Joined: Thu Apr 17, 2008 10:32 pm

Re: Scope PCI-card effectivity

Post by Inferman »

Fede wrote: Btw Z-link has nothing to do with this, it is a proprietary digital connection to send/receive audio channels among creamware products.
Fede
Fede, that is what Im talking about.
So we have the following: CW developers says that Z-link connect CW cards because PCI slots cannot provide such data-stream to manage it. So my first CW card for example send to the second some data via Z-link because PCI-slot is too weak for this, is it correct? But where this data goes next? It connect with MB, CPU and other devices via PCI-slots (because CW cards inserted there). So If CW card 1 sent to CW card 2 "X"-Data and CW card 2 produce some "Y"-Data it seems PCI bus have to mange with X+Y Data (or X and Y separately by each PCI-slot). But if they say PCI cannot provide such data-stream how can it manage it? So in general why do I need Z-link with "faster" data-connect with other CW card if in result it will send to "slower" PCI-slot ? That's is my question.
User avatar
garyb
Moderator
Posts: 23364
Joined: Sun Apr 15, 2001 4:00 pm
Location: ghetto by the sea

Re: Scope PCI-card effectivity

Post by garyb »

no, that's not how it is. not at all.

z-link is strictly for connecting an A16 ultra ad/da converter to a card. it is an alternative to adat light pipe.

cards are connected together by an s/tdm cable. the cards communicate with each other via s/tdm and with the applications through pci.
Inferman
Posts: 133
Joined: Thu Apr 17, 2008 10:32 pm

Re: Scope PCI-card effectivity

Post by Inferman »

I'm sorry, I mean s\tdm cables, not Z-link )) Thanx for correction, I'm always confused in names.

Anyway its not change the things ...
User avatar
garyb
Moderator
Posts: 23364
Joined: Sun Apr 15, 2001 4:00 pm
Location: ghetto by the sea

Re: Scope PCI-card effectivity

Post by garyb »

what's the problem?

all the dsps can be used even with 3 14dsp cards. pci works wonderfully. pci-e is potentially better.

more than traffic problems connecting cards through pci, there are timing issues, hence s/tdm. the pci connection wouldn't allow the cards to access each other directly, there would be controllers in between creating timing issues in a real time environment(as i understand it).

this is such a non issue....
Inferman
Posts: 133
Joined: Thu Apr 17, 2008 10:32 pm

Re: Scope PCI-card effectivity

Post by Inferman »

The problem is - there is suspecting that PCI-slot is too slow to provide enough data-stream between MB and CW card, so as result it will works fine but if You (I mean if I) buy 14 dsp CW card for my current 6 DSP I will not get real 14+6=20 dsp power because PCI bandwidth will cut the real stream so its useless to build 3x14 dsp card system and PCI-e is the only way to get really 20,30,40 dsp power.

Thats why I want to get some tech specification for CW cards but I didnt find it.
User avatar
garyb
Moderator
Posts: 23364
Joined: Sun Apr 15, 2001 4:00 pm
Location: ghetto by the sea

Re: Scope PCI-card effectivity

Post by garyb »

if you add cards, YOU WILL GET THE FULL BENEFIT! :lol:

many, many users have multiple cards, yet you don't see angry threads about not being able to use them do you?
Post Reply