XTC benchmark

[maky325]

Hello

Some technical questions. As i saw in thread about XTC (and how to configure XTC doc) people like to say that XTC need more PCI bandwidth and therefore is needing more resources. Well yesterday i did benchmark with Minimax, Profit 5 and Protone. I have 21 DSP system and so far it seems that i can run EXACTLY equal number of synth/voices just like in SFP mode.

Is this normal? I mean i was expecting lower result in XTC mode becuase of information in these various documents. Now i am not complaining because of this results or something i am just looking is something wrong or not with my configuration? Maybe i can run even more in standard SFP mode or something. I am just playing with XTC these days. Difference in PCI bandwidth probably exist but is it so big or we should not think about it when we are in XTC mode?

[the19thbear]

hmmm i have never heard that xtc mode should use more bandwith! interesting!
what you DONT need in xtc mode is a mixer, and that WILL save dsp

[maky325]

hmmm i have never heard that xtc mode should use more bandwith! interesting!
what you DONT need in xtc mode is a mixer, and that WILL save dsp

Well you oviously missed sticky thread here and there is a word document named PlanetZ - Introduction to tweaking XTC.

Here is a quick copy paste:

XTC mode is the most demanding configuration for Scope board! Why? Well, because it increases:
- PCI Bandwidth usage: each plugin have ins/outs added to your basic routing (and their own internal use of PCI. Ex: Reverbs, delays and samplers which have "hidden" PCI load). The maximum PCI communications channel seems to be 64 (for standard PCI Clock), the closer one comes to this limit, and the buggier the system behaves. As there is no "hard" limit and as one can reach it easily, you won't be aware that you overloaded the system.
- You'll stress the system for Low Latency use, which increase your hardware resources usage.
- Your Vsti host might not load devices in the order you expect: Not all Scope synth are equals, some require being loaded first to have a stable operation.

[MD69]

Hi,

You mix DSP power and PCI usage which are two different things.

To make it more clear, suppose you have STS2000p as XTC device, you'll have the 7 outputs of the device AND the 16 PCI channel required by the device for adressing PC ram.
If you add your basic routing (Inputs and outputs of your host app), you will have a higher PCI usage than the same devices loaded in an SFP project.

TO give you an idea, I am very close to the PCI limit with 8 XTC Fx (P100, CD100, MASTERVERB Pro, CELMO Vintage delay, OPTIMASTER, AMPLITUBE, SL9000 Channel strip, DAS 2A) and 4 XTC Synths ((Minimax, Prophat, ProTone and B2003) and I just have 12 ins and 10 outs, but the DSP load on the other end is 60%.

cheers

[maky325]

Yes i am aware of difference between DSP and PCI performance but for example i can load same amount of P-100 or Masterverb Pro in XTC like in SFP until i get PCI overflow error and that is where i am confused a bit. I mean it is not anything major i just have some time so i am experimenting a little. By reading various docs i came to idea that XTC is more hungry but in my system it behave just like SFP (in performance area of course) so i was thinking that maybe i am doing something wrong. Though 15 Masterverbs is nice result which make me calm

[MD69]

Well il some case, it doesn't make a lot of difference, but in my exemple, you could add 12 stereo delay in SFP before you reach the PCI limit (for a stereo delay in XTC mode, the PCI usage is twice the one in SFP).