CPU Killer!

Computers will get faster, and you can always treat Massive as a monophonic synth.

 

I have a pc running windows xp pro, stripped down for top audio performance. here are the other specs:

amd x2 4800 cpu (2.4ghz) NOT overclocked
2gb 400mhz RAM
dual 7200rpm SATA drives (one dedicated for audio files)

massive is a CPU HOG. my system is essentially brand new, with a 2GHz system buss mobo, yet I can't play more than 4 or 5 notes in massive without the CPU overloading in full quality mode.

on a positive note, ableton's LIVE 6's CPU load was cut in HALF from Live 5. I'd like to see a fully dual-core cabable massive release.. I just don't believe that it's dual-core enabled.

 

The only way you should be getting only 5-6 voices on that system is if your latency is set very, very low, like the very lowest possible (32 or 64 samples or so, perhaps).

In such a case you waste a tremendous amount of CPU with any plugin, and especially plugins that are hard on the CPU (like massive) will be difficult to get much out of.

If that is the case you might consider bumping the latency up to 128 or 256 samples (@44.1) which is still plenty usable for most purposes.

If you're already running at something more like 128 or 256 samples of latency, then something else is surely wrong. My rather old Athlon 64 gets about 12-16 voices, probably you should be getting about the same on each of two instances of Massive.

 

my latency is set to 25ms!!!! that's as high as I can set it. I'm using ASIO drivers... my interface is a firewire Presonus Firebox - yes I'm using their latest driver. seems ridiculous doesn't it?

 

With respect that is a dumb thing to say - yes of course they will get faster - what do I do with my £4,500 machine that was purchased just under a year ago? run one instance of battery 3, an instance of Massive and an instance of absynth and the whole thing starts to crack.

 

It's not that dumb... it's just a statement about the obvious, one that reflects on the concept that Massive may be ahead of its time. It's a frequent occurence on video games as well, where most of the recently released games require more power than any current computer can output but will play fine on a new computer a year down the road.

You spent 4500 brit pounds on a computer? That's like $8500 US. Why did you spend such a huge amount of money on that computer? Does it have special hardware or something? I usually get a computer that is near the top of the performance chart and it only costs around $2000.00 US. Ah, wait, I just noticed you bought a Mac... I'm running PC's, which are a whole lot cheaper for equivalent performance. But still, $8500 seems like a high price to pay even for the normally over-priced Macintosh.

 

Yes I spent that much on the Machine it included upgrades to 4gb RAM which bumped the price. The thing to note is the minimum specs for MASSIVE which are Mac OSX 10.4.x, G4 1.4 GHz or Intel® Core™ Duo 1.66 GHz, 768 MB RAM - Win XP, Pentium or Athlon 1.4 GHz, 512 MB RAM

When you check those out and then look at what is really happening it becomes a bit shocking.

So yes MASSIVE may be ahead of its time but the minimum specs would suggest otherwise in terms of CPU drain.

 

I'm not having too many problems on my dualcore 2x2ghz but I find it's better to set the max number of voices to 8-10 (rather than the default 16) but keep the sound quality high rather than try and save CPU by reducing sound quality.

 

Yiu've got other problems happening there, then. On my single core 3500+ (which is overclocked to 2.4 GHz), I see 60% CPU for 16 voices in ultra quality...

And no, it's not dual-core optimized. The only NI app that takes advantage of dual cores as far as I know is Kontakt.

ew

 

Hi,

It would be nice if NI would spend a little time and money to optimize the audio engine of Massive. I hope that they will release an update. It is impossible that it is as slow (compare the cpu usage for example to complex FM8 sounds or older complex Absynth sounds). I bet that a factor of 2 improvement could be achieved easily.

And to those of you who say something like this "for a quality like that I am quite happy with the 8-12 voices I can get...":
come on! compare it to the FM8, the sound quality is clearly as good as Massive with aprroximately 5-10 times less cpu usage (most FM8 sounds use 1-3% while Massive uses 10-50% per voice on an Athlon XP 3000+). I think we should motivate NI to improve this poor performance. Are you really satisfied? Massive has some really outstanding ,thick patches but in general it is not suitable for serious multitimbral composing currently. I think that this is not normal because today's computer are really-really fast (only the developers do not take the time and money to optimize software, they think if they can sell it so why bother optimizing).

So, are you really satisfied with this "massive" performance?

 

I am not happy with the CPU usage of Massive, too. But you cannot compare Massive with a FM8. The FM8 sound engine is based on frequency modulation, which is much easier for a CPU to compute than a complex subtractive synthesis.

 

Actually as I know FM synthesis is the cheapest synthesis method available CPU-wise. That's one of the main reasons it is used in polyphonic ringtones in mobiles.

 

Actually, the FM8 sound engine uses phase modulation (as did the Yamaha "FM" synths) and NOT frequency modulation...

ew

 

Yes, I know, but still mathematically it is not that different:

signal = sin(2 * pi * F + P)

No matter if you change frequency (F) or phrase (P), you're changing the function's argument.

;-)

 

Hi,

Of course they (Massive and FM8) cannot be compared directly but I am quite shure (I am programming since 20 years in C++) that a >10 times (!!!) factor between the two means some luxury (as NI would say) programming techniques (too many indirections, software layers, etc.). If you don't agree, look at the CPU usage of the additive and fm synthesizers contained in the Reactor5 package.

Anyway I hope that my view will be confirmed in the form of a software update with very important optimizations (if they think that massive should be suited for any multitimbral use).

By the way there are 8 modulators in Massive. Can they switched off somewhere?
Another question: the first voice uses lets say 22%, adding the 2nd, 3rd and 4th noice does not increase the CPU usage? So for 4 voices we have 22%. Then I would think it would be possible to have 22/4=6% for one voice (neglecting the effect section of course). Am I wrong? That would be much better because there are many situations (monophonic basses, solo tracks, etc.) where this would be of great advantage when creating multitimbral music on the edge of CPU power. This 4X factor is too much not to consider!.

What do you think?

 

As far as I know; yes, you're wrong. The x4 routine's brought on by the parallel processing functions of the SSE (or Altivec for PPC Macs, or whatever the new IntelMacs are using- considering it's Intel, it could very well be SSE there as well) routines used. If you threw those out, your CPU usage would be MUCH higher than it is right now.

ew

 

I think ewistrand is correct: with the vector extensions, you are better at processing groups of at least 4 vectors. According to my own tests I did when developing my own AU plugin, processing 4 numbers with a vector operation is faster than processing 1 number.

 

I see, so you say that there is already a 4x parallelization present. That's nice to hear, but 1 voice remains 10-20 times costly as FM8...

 

You're comparing apples and oranges...

ew

 

Ok, I see there is massive defense for Massive, but whether it is an apple or not, it is a bad fruit anyway if it requires 2 times the CPU power to "produce" no sound at all than the FM8 to generate an average fm7 sound (with all the oscillators, modulators, etc. running).

Try the following: select "New Sound" from Massive's File menu, switch off all modules (including all oscillators) and hit a key. On my computer the CPU usage goes from 1% to 5%. On the same computer FM8 uses typically only 2-3% (for very complex voices). And this is what Massive needs for no sound produced!

Is it an apple or an orange? In any case, it is clearly NOT optimized!

PS: I think that it would be better for all of us if we could motivate NI to optimize this synth, I see no reason for other behaviour. Finally, it would bring more profit for them (and so for you and us) too... The concept of NI must be obviously that we should wait 1 year until computers will be fast enough to compensate for the missing optimization work or we should throw out 1 year old hardware right now and invest another $500 to update to an intermediate compromise pc.