Anyone using FX Teleport with Usine?

[bsork]

I'm wondering if anyone has tried out FX Teleport with Usine.

From what I've read about it, it looks like a smart and relatively cheap way of utilizing older machines with network cards. I know I could probably do something of the sort using OSC and running Usine on both machines, but FXT seems to be a lot easier solution.

[senso]

I use to try;
It works fine but I don't really need it because, it increases the latency (a lot...) and the lan transportation cost finally more than putting the plug directly on the main machine.

[bsork]

Well, I was thinking of using it mainly for send effects where a little bit extra latency doesn't matter much, and where the number of audio channels back and forth will be low- like for instance reverb.

Anyway, I think I'll try to get an IP connection working between the old and new computers, and I'll I take from there. I guess the load put on both CPUs depends a lot on the network cards installed, so I do that first and take it from there. I think I soon have to send the new one away for service (bad FW plug), so if I can update the old one without going through an external HD or a memory stick, I guess I haven't wasted too much time with this, even if FXT doesn't do much good.

[bsork]

Here's a little report after not a very extensive testing in case anyone's interested:

My new HP has a Gigabit NIC, but the old Dell has only 100 Mbit, so the networking is limited to 100 Mbit. A little thing to note, is that most newer cards can detect a direct connection to another card (no switch or hub), and adjust themselves so that a "crossover" cable isn't needed. Only one of the cards need to have that.

I tested with Glaceverb with everything at full, and a Helios patch playing a big chord repeatedly, ASIO settings first at 44.100/buffer size 128, and then at 48.000/buffer size 64.

Here's the approximate numbers as reported by the patch cpu indicator:
At 44k/128:
Glaceverb local 5.5~6.0, FXT: 2.2~2.7
Helios II local 13.0~14.5, FXT: 2.4~3.2

At 48k/64:
Glaceverb local 6.8~7.4, FXT: 5.1~5.5
Helios II local 16.0~18.1, FXT: 6.8~7.2

At 44/128 FXT didn't report any dropouts, and the latency reported was twice the buffer size - which I believe is the lowest possible (but I might have misunderstood completely...). At 48/64, some dropout were reported, and for Glaceverb the peak latency reported was 192 samples, ie triple the ASIO buffer size. Why Glaceverb would cause a higher latency than Helios I don't know, but maybe it has to do with audio travelling both ways?

A thing that strikes me about the results, is that the change in CPU usage with FXT is roughly following the ASIO settings, while the local numbers didn't change nearly as much.

I haven't tried to optimize things so I guess the figures and the dropouts for 48/64 can be somewhat improved.

I think I conclude that if I should be in need of some more CPU horsepower with Usine I might go to the trouble of using two laptops, but then the slave should only do less timing critical stuff (like reverb). On the other hand, I'm also using Cubase, and I think that this product can come in quite handy when one has a (not too) old computer lying around gathering dust, especially when mixing with increased buffer sizes. The fact that FXT uses standard networking cards which practically every computer is sold with anyway, is a real bonus compared to some other technologies that communicates through audio interfaces. Especially since my audio interface only has analog ins and outs...

[senso]

Interesting conclusions.
A good way to recycle old computers.