Latest News: u-he updates all their plug-ins
|
|||
Urs wrote: Thanks Phaze!
Multicore looks promising, but it's still a hell of a lot of work, and we don't know yet if it really pans out. Our internal (working) multicore version just so gets twice the number of voices out of 8 cores. We're looking into many possibilities to gain speed though. There's a tiny chance for an algorithm that sounds almost divine, at a fraction of the cost. This needs to be investigated. And of course there's GPUs, AVX and what not, which as of yet is rather experimental or hypothetical as a viable solution. This is not the best way to spend development time. How many modern plugs can run on a Pentium III without problems? DIVA already offers several quality modes that can be bumped up to Divine for offline rendering. Buy a I7 PC and let the skilled developer focus his time on refining ITB what was once only concievable in hardware. ---- Intel Core2 Quad CPU + 4 GIG RAM |
|||
| ^ | Joined: 05 May 2002 Member: #2696 | ||
|
|||
IMHO I think that Urs and his team should definitely pursue any viable options to make the CPU load of Diva lower. It definitely cannot be a drawback. |
|||
| ^ | Joined: 06 Jan 2009 Member: #197719 Location: Croatia | ||
|
|||
electro wrote: This is not the best way to spend development time. How many modern plugs can run on a Pentium III without problems?
DIVA already offers several quality modes that can be bumped up to Divine for offline rendering. Buy a I7 PC and let the skilled developer focus his time on refining ITB what was once only concievable in hardware. Well, we do need stereo versions of that type of filters to put in our other products (not same filters, but same principle), and we need to do it this year. Of course, we don't need to be as close to the originals in our other products, which makes maybe 10 - 20% of the code, just to get the tweaks right. A more simple topology will sound as good in terms of sonic properties, but won't be exactly following an analogue counterpart. That said it would be highly desirable to run many, many more instances and it would be highly desirable to be prepared for future technologies. I do believe that we're a step ahead of the native synth game at the moment, but it won't be long until other people figure certain things out for themselves. In case of which I want to be able to reach the next step already.  Urs |
|||
| ^ | Joined: 07 Aug 2002 Member: #3542 Location: Berlin | ||
|
|||
Urs wrote: I suppose that it has to be expected that one day some developer will come along claiming that 0-delay-feedback-filters were doable with "low cpu" with any whatever "different" method, but still preserve analogue-style non-linear elements, self oscillation and oversampling. But alas, there's no way around heavy maths that comes at a high cost. See UAD Moog filter for reference - biggest cpu hog on my UAD2.
I'm considering as follows: We would publish a formula of a popular open source filter in zero-delay-feedback form and unoptimised but working code. We will also post our heavily optimised implementation as closed binary in a plugin, along with a test scenario. We will then challenge any developer to implement the same formula within a set margin of accuracy and means (SSE etc.) in a significantly faster way. If they manage to make it use less than a third of cpu on the same data, then I'll start believing in methods for low cpu versions. Why bother? Diva is the test/challenge. The sound speaks for itself and lots of people recognize it. |
|||
| ^ | Joined: 02 Feb 2005 Member: #56776 Location: in the wilds | ||
|
|||
pdxindy wrote: Why bother? Diva is the test/challenge. The sound speaks for itself and lots of people recognize it.
Well, even if the competing product isn't released yet (thus can't be compared), the claim alone can shed a negative light on our work. Just the confusion of 0-delay-filters with 0-delay-feedback-filters can create wrongful associations: http://www.kvraudio.com/forum/viewtopic.php?p=4841872#484187 2 |
|||
| ^ | Joined: 07 Aug 2002 Member: #3542 Location: Berlin | ||
|
|||
Just out of interest and technically I have not got a clue about programming and would never pretend otherwise.
Can CUDA technologies by Nvidia be utilised to help aliviate the CPU burden, similar to Acusticas Nebula or am I way of track here? How about getting a very cheap 2nd PC for running soft synths over LAN, again like Nebula, Vienna Pro and other similar products? Note : CUDA is where the GPU (Graphics card cpu) is used instead of the main CPU, for those who have not heard about it. ---- Rintrah roars & shakes his fires in the burdend air 'Need talent & opportunity' |
|||
| ^ | Joined: 27 Mar 2010 Member: #228659 Location: UK | ||
|
|||
Urs wrote: pdxindy wrote: Why bother? Diva is the test/challenge. The sound speaks for itself and lots of people recognize it.
Well, even if the competing product isn't released yet (thus can't be compared), the claim alone can shed a negative light on our work. Just the confusion of 0-delay-filters with 0-delay-feedback-filters can create wrongful associations: http://www.kvraudio.com/forum/viewtopic.php?p=4841872#484187 2 That thread shows that Diva hit the nail on the head (not that there was any doubt). That other developers are talking about Diva in their own promotions speaks volumes! Many hardcore gearheads have raved about Diva. I think there is little question that Diva is the current gold standard. I see it mentioned frequently as such and there is almost no disagreement in response. How uncommon is that? I doubt there is anything you can do that will come close to that good press. Anyway, do as you see fit of course, but I think you can focus on development and not worry about proving anything. Most people don't care about 0-delay or 0-delay-feedback... they care how the synth sounds and Diva flat out sounds gorgeous... expressive, vibrant and musical. |
|||
| ^ | Joined: 02 Feb 2005 Member: #56776 Location: in the wilds | ||
|
|||
You're in the lead Urs, I wouldn't waste *too* much time looking over your shoulder. Just keep moving forward! |
|||
| ^ | Joined: 19 Oct 2010 Member: #241844 | ||
|
|||
EvilDragon wrote: IMHO I think that Urs and his team should definitely pursue any viable options to make the CPU load of Diva lower. It definitely cannot be a drawback.
well, all i can say by my side is that, on my iMac boosted with an i7 quadcore i don't feel this as such a terrible issue I may admit though, that i'm a lucky fellow  
And if we still can count on hardware development of new generation'c computers... |
|||
| ^ | Joined: 24 May 2002 Member: #2860 Location: Bobo-dioulasso\BF__Geneva/CH | ||
|
|||
Krakatau wrote: EvilDragon wrote: IMHO I think that Urs and his team should definitely pursue any viable options to make the CPU load of Diva lower. It definitely cannot be a drawback.
well, all i can say by my side is that, on my iMac boosted with an i7 quadcore i don't feel this as such a terrible issue I may admit though, that i'm a lucky fellow
And if we still can count on hardware development of new generation'c computers... I would also like to get a lower cpu use... but I would certainly prefer the high cpu use rather than compromise sound... and the effort to use multi-cores does not seem like a fruitful direction. There must be many seriously good coders out there working on exactly that idea and it remains elusive. 5 years from now cpu's will handle Diva easily, but in the meanwhile, I don't expect any significant cpu use reduction for Diva. For me, the task of render/freeze is well worth it because Diva sounds  |
|||
| ^ | Joined: 02 Feb 2005 Member: #56776 Location: in the wilds | ||
|
|||
most amazing synth to date IMO! |
|||
| ^ | Joined: 29 Jul 2007 Member: #156526 | ||
|
|||
pdxindy wrote: Anyway, do as you see fit of course, but I think you can focus on development and not worry about proving anything. Most people don't care about 0-delay or 0-delay-feedback....  I couldn't agree more! Responding appropriately to any false claims as they appear is best. There's NO way to stop such claims before they have their intended effect. |
|||
| ^ | Joined: 22 May 2004 Member: #26645 Location: Germany | ||
|
|||
Really the difference in sound quality between fast and divine isn't THAT great for a lot of sounds (not all but a lot) and the cpu usage is far far less in fast mode so cpu wise Diva isn't all that bad really.
Fast or great for realtime use and Divine for rendering. One thing I am noticing is that the lesser quality settings sometimes actually suit certain sounds better, particuarly high resonance filter responses. So I'm kinda hoping if future improvements are made to the different quality settings, that the overal tone isn't going to change too much, just the cpu usage going on under the hood. |
|||
| ^ | Joined: 28 May 2002 Member: #2892 Location: UK | ||
|
|||
I'm assigning Diva knobs to my MS20 Legacy controller in FL Studio, and I notice that Osc2 octave and detune is the same midi parameter (Tune2). Any idea what to do about that? |
|||
| ^ | Joined: 31 Jan 2003 Member: #5693 Location: Ghent, Belgium | ||
|
|||
i am down to using draft realtime since i want more than 3 instances on my Q9650.. quad-core 3.0ghz 64bit non-hyperthreaded.. but, i also have 2 303's and 2 808's going all with fx... i scaled back to 384 samples from 256 and CPU is handle-able again.. ----  |
|||
| ^ | Joined: 04 Feb 2004 Member: #12259 Location: Northern California |
Disclaimer: All communications made available as part of this forum and any opinions, advice, statements, views or other information expressed in this forum are solely provided by, and the responsibility of, the person posting such communication and not of kvraudio.com (unless kvraudio.com is specifically identified as the author of the communication).
Powered by phpBB © phpBB Group
