Thanks for reply. it is kind of live use. But not on stage, but in my humble homestudio. So I do not have to be mobile, but I am not arranging, I am just playing live - that is, enjoy turning a lot of knobs without any real clue. And if it sounds crap for a given day, I at least do enjoy all the blinking lights. The card is an old RME Hammerfall (not even DSP), and I'll be using X with a standard distro. Gentoo or Arch - or maybe KX, if they really should drop ubuntu as a base, as I've read somewhere. Depending on the software I'll be using in the end. So we are talking about a simple desktop PC in a tower case with an 21" CRT Monitor. About multitrack: Being able to run 32 effects (8 Tracks x 4) would be great, 16 tracks perfect, but my budget is not infinite and I doubt I'll really use 16 Tracks at once. And most likely not really reverb on all tracks, that is just the worst case to find out a proper CPU. However, there will be differnt rooms on different tracks. There is still an external master reverb inserted into main mix. If so desired. Maybe I'll try to control some of the effects via Midi - I do own an old Doepfer Drehbank controller, but not sure, wether I'll have the patience to program it before winter. Maybe handy for the EQ. But it's also occupied for the microwave II. However, EQing with a mouse is a PITA and something I will have to solve sooner or later. Way worse then setting up reverb or delay. imho. I do not plan to use effects via aux sends, just stack one on top of the others in a serial manner - as if I would do, if I had 8 hardware effect devices plugged into my mixer. There won't be any mixing (as in mastering) - at least not on this box, I am oldfashioned and need real faders. I do have another box for recording - mostly straight to /dev/null, that is unfortunately. This box will just be plugged into the inserts of my mixer - therefore the ideal of 16tracks - which I suppose will be cheaper and better to use than buying a couple of multieffects. I do not have any more ADDA convertes anyway. And for my lousy ears, the quality of named plugins is more than adequate. So it's basically -> jack-in->effect1-effect2-effect3-effect4->jack-out with 16 permanent physical connections bewteen soundcard and mixer. Though not all of them will be used at once. I do have the soundcard (which I finally do inted to use properly after all those years) and the ADDA converters, the screen, SSD and the case. All I need is a mainboard, CPU+cooling and some RAM. And maybe an el cheapo graphics card, if going the AMD route. I doubt I'll get usable multitrack hardware for that money. Sure, I am screwd if one of those parts breaks, it's all yesterdays jam and hard to get besides waiting for ebay. The price limit is a bit difficult as prices are varying between countries. As I've said, an upper class intel i5 or AMD is about what I am thinking of, until general consesus is, that there is no way to accomplish this with standard, upper class (as opposed to high end) PCs. Then maybe I'll wait a little more and save some money. But, as you mentioned, I do remeber running softwareeffects, even not really in parallel, way back on an 2GHz AMD 3200 in decent quality even on WIndows, so a modern multicore, I would suspect, should be able to run a couple of them. Just which would be the best bet. For hardware devices: Behringer is out of discussion. As is Alesis. I do have a couple of midiverbs and found the quality of calf to be way better. If maybe not for the algorithms, then at least for the noise level. Indeed, I am planning to sell those. For peace with my drumheads and of course for better useability. So again - are plugins really FPU heavy? Or would it be more important to have feature X, like AVX or SSE27 or so? Unfortunately I am not really into this recent hardware stuff and all the latest buzzwords. And I usually do not trust benchmarks, but that's all I have and the FX series did quite well in some so called media benchmarks, like rendering, however, I am not sure whether this can be tranferred to running multiple audio effects without problems. AS maybe rendering is just two fast threads instead of 16 medium ones. Dunno. The price advantage for the AMD would probably be eaten up by the more expensive colling - I prefer it rather quiet - and the higher energy bill. But if it would be the more adequate CPU for this very workload, I would go that route. i7 are all HT, so that would leave the i5 as only other option. Am 13.09.2014 18:14, schrieb Len Ovens:

On Sat, 13 Sep 2014 19:44:19 +0200 Ede Wolf <listac(a)nebelschwaden.de> wrote: IME, HT dosen't have to be a showstopper. It's conceptually "wrong" to use it if you are relying on linux realtime processing, but in real life it's normally not a problem with audio, unless you push the system real hard doing other things. I also theorize that on an i7 it can lead to cache depletion and cause long xruns when the CPU is heavily loaded. On the other hand you can normally just disable HT in the BIOS, so with the right motherboard it shouldn't be a criteria to disqualify a CPU. -- Joakim

On Sun, 14 Sep 2014, Gene Heskett wrote: I agree that would make a lot of sense. I think there are some other tools that could do some of these things too. However, many people even in linux audio rely on OOTB setups. It is hard for a distro to imagine what setup the user might have to make the very best use out of it. Also, those who put distros together often do not have the knowledge to set something automated up either. So this will have to start will people who are willing to take the time to customize a system to use these ideas. The one problem with the older Atoms is the GPU. It does work ok, but it is always (in Linux) configured for at least two screens (a direct LCD seems to be the exception) one of which normally gets set up to not show (but still uses memory). The latest ones have a different gpu, but I do not know if it is open either. I put one in my wife's computer, but it is not keeping up with her main (mostly graphics) use. So i will be using it as a headless audio board after I get her another one (with an open GPU like my son has). -- Len Ovens www.ovenwerks.net

On Sun, 14 Sep 2014, Gene Heskett wrote: Everyone has to start somewhere. I think it is possible to have a better OOTB linux machine than windows. If the person interested in trying linux can't do better with "out of the box", they may just give up. If they already have something better, goinng farther makes sense. -- Len Ovens www.ovenwerks.net

On Sat, 2014-09-13 at 09:14 -0700, Len Ovens wrote: This isn't possible for all effect combinations that could be needed, but I agree, I would do the same providing the most common usages. I agree that aux sends are much better for reverb. Usually we want one virtual room and only need to control how much is send to the reverb. By using aux sends there's - no overlapping of reverb tails or overdosed early reflections. I don't know if this would happen when using a reverb for each channel, if it would cause an audible unwanted side effect, but theoretically it's possible when using tons of reverbs. I never tested it. - For the effect rack in the box, less DSP load, for analog reverbs, less money, since just one reverb is needed, and less background noise, by less reverb outputs.

Because it can be useful to put f.e. drums, rather parts of it, in a very small room, maybe even lofi - just to change or slightly fatten the sound but not use this very room for anything else? I do not use those settings for a global reverb and just controlling the amount with the aux sends as the settings are completely different. Same is true basically for synth - nearly each digital synth and most modern analog ones have their own effects unit used to enhance or shape a sound - and nearly always a reverb plays an important part. You hardly turn that off because of a global reverb I believe you are referring to. Again, I am not mastering or producing, I am creating sounds which I hope will play well together. Each track servers one instrument, and each instrument may or may not have an own room. Harmonizing that is a challenge, but there we go. The reverb, if not used too extreme, is foremost part of the sound, not of the mix, though of course one can easily ruin the other, if not used carefully. No doubts here, again, that's part of the challenge. But, while I would not mind to open an extra thread for this, I just would like to use this thread for CPU and/or Software recommendations - and not focus too much on the reverb. I just believe, that a reverb is most likely the most complex effect requiring the most CPU power. I may be wrong here, too, that's why I am asking. I am not even claiming I will be using a reverb on each channel, but I would like to setup the rack for each track in such a manner as I do not want to play around too much with it. Set up 16 tracks and the ones not used, well, are muted. I just want to set up the rack once with somewhat sane defaults and then use it. In other words, I want to start that application and have a rack with 16 cahnnels and 3 or 4 effects in each. Set up ready to use. Of course I will need to adapt the settings anyway, but basically it will start up the a reverb, a delay and an EQ in each channel. As those are the most commony used so far. That does not mean, that all are active all the time, but if it sounds crap, I would like it not to be the CPUs fault. I want to minimize the time spent with the computer. Making it as much an appliance as possible. But sure, it'll need some attention. Quite simple: I am looking the the best available CPU for whatever effects calf or invada are offering running as much of them in parallel as proves to be sensible - and usable. And how much is sensible for a given set shall be decided by my ears, not by the CPU. So if I am asking for a CPU that _could_ run HQ reverb on 16 channels in parallel does not mean I will be doing that. But if it could do that, it surely will be powerfull enough to run any other effect desired. That's my theory and I do not mind at all getting proved wrong. In fact, that's why I am asking. Some people here are coders, as are you, and know, what "part" of a cpu is used most for their algorithms. Or so I hope and was asking for a recommendation. Sure my fault, I wrote too much, again, I just had a simple question: Which (not too high end) CPU would you recommend for running as many lv2 effects in parallel as possible. And why, if you are kind enough to share. I do not care for ladspa, vst, dssi, softsynths or recording. Just a little midi for the delay, but that is not toooo timing critical (please don't get off on this, I am not saying it does not matter, but a _minimal_ shuffle in timing is quite acceptable for a tapped delay) Just send audio from a mixing consoles insert, process the audio as fast as possible and get it back to the mixer. On as much channels as reasonable possible. And if not all are used at once, then speedstep is happy. Am 13.09.2014 20:47, schrieb Fons Adriaensen:

On Sat, Sep 13, 2014 at 09:42:07PM +0200, Ede Wolf wrote: Most reverb algorithms (apart from convolution reverbs) do very little and simple calculations but quite a lot of memory access. I'd expect those to be dominated by cache size and speeds rather than raw CPU speed. And if you have 16 independent effect chains, the thing to go for is to ensure they will execute in parallel on an SMP system (more or less the norm today). And that depends on software architecture and nothing else. Ciao, -- FA A world of exhaustive, reliable metadata would be an utopia. It's also a pipe-dream, founded on self-delusion, nerd hubris and hysterically inflated market opportunities. (Cory Doctorow)