How did Fabfilter make the EQ more efficient?

How did Fabfilter make the EQ more efficient?:
 
"Highly improved CPU optimization: Pro-Q 2 uses less memory and is more than twice as efficient as its predecessor!"
 
Was it the EQ math? The eye candy graphical animation? Something else?
 
It seems to me that it would be hard to make a dsp EQ twice as inefficient as some other EQ, as long as you are comparing like to like conditions such as "natural phase" or "linear phase" mathematics.
 
This leaves me wondering; what was going on with the previous version of FF EQ that was not mentioned in the numerous reviews I have read? It leaves me wondering how/where/why Fab Filter was able to find so much opportunity for improvement while adding all sorts of features like extra poles (a.k.a. the new slope choices) and even slicker graphic features.
 
Anyone?
 
 
 

Nice explanation, Bit.  

Bit has given a good summary of what optimization entails.  It's a very complex field ... many books and advanced university courses are devoted to the subject, but it's not something that very many programmers are good at.  There's not a lot of software that ever gets optimized to a great degree, and there's no end point where you say it's done -- there's always some little-known trick that shaves another microsecond or two off an operation.  Thus, optimization efforts are never finished, only abandoned. 
 
Hmm, that reminds me of another process most of us are familiar with ... 
 
Anyway, the fact that Pro-Q's efficiency has been improved this much is a testament to the FF guys' programming skills.  They know what they're doing, for sure. 
 
Cheers,
Eddie
 

dual mono post ^
 

mike_mccue
It seems like "optimization" should be a selling point, perhaps to justify a premium price, and not-yet-optimized might be considered not quite as good.
 
It would be nice to have a way to learn what is and isn't.
 
I have all the FF stuff. I think of it as ok, and I'll be buying the Pro-Q2 *just because*.
 
I hope they do some work on the sound and the curves of their Pro-C next. If they optimize it too then I'm gonna be giddy.

 
The level of optimization can be inferred by seeing how much CPU is used by the plugin.  The fact that Pro-Q 2 uses very little CPU compared to similar products is certainly a selling point, at least for me.  There are many plugins that I'd use more often if only they were more efficient. 
 
I think it's a safe bet that Pro-C 2 will also get the full FF optimization treatment.  They've probably learned a lot from their efforts on Pro-Q.
 
Cheers,
Eddie
 
 
 
 

As the teraflops increase, this becomes less of an issue. Moore's Law is on display in our techno oriented hobby/occupation on a yearly basis. And as such, "optimization" becomes less and less of an issue. If your CPU goes from 1% to 2%, did a tree really fall in the forest? I don't think it did. Look back a few years at having to type in "win" at the C: prompt? It's just getting faster and faster. And therefore, IMO, "optimization" is a footnote. It's like driving down the autobahn in the right lane with your blinker stuck on? People are going to come flying past you in the left lane @ 200+ mph...
 
Teraflops baby? Imagine your DAW only 100, or 1000 times faster? And with corresponding increases in memory and disk capacity and access speeds? Everything solid state, no moving parts? Everything in a RAID5 array, or whatever is next. Everything in the cloud, including VSTs, and your project "drive?" and your library "drive?" And your C: "drive?" Will optimization matter then? When you have a 10gigabit link to your home? I don't think it will.

Well it comes my friend, and it comes quicker than you think. When I was a teenager, home PCs were in their infancy? I remember loading "wordstar" and "calcstar" on my mothers "5,000 dollar PC" on 5.25 floppy disks, with her dot matrix printer? And it's increasing exponentially. I work for a service provider, and we struggle to keep up with demand. It gets hotter and faster every year. Capacity grows, speed increases, or more accurately, "speeds increase" content grows. It is a good thing for what we do, because again as I said? "Teraflops..." In a few more years, we will laugh at the us of today, worrying about CPU cycles. You'll put on your virtual reality glasses and sit down at your DAW which will be on your phone? And you'll manipulate things by moving your hands and or your eyes. We are in the midst of a revolution... It will be like the lawnmower man. Be thankful that you are here and part of it, in 10 years, you won't recognize it... Band in a box? "Band on your phone..." And a song in a few minutes (original) with you singing it? Or someone on the other side of the world. It's all changing "right now" before our eyes. 
 
I believe you'll choose to use virtual instruments "virtually?" You will choose to use effects "virtually" as you need them? They will be licensed on a per-use-basis. Rather than owning the software? You will choose to use it, per project. Then you will distribute that project and people will pay a small fee to enjoy it (a one time fee) and it will all be encrypted in such a way that it discourages sharing as is done today.
 
The future is content? And what I am describing is not new, but it's my take on what is coming. From someone who does "what we do?" And is in the IT field. I may be wrong? But this is what I see based on where I've been, where I am, and where I see us going.  (there will still be pirates matie...)
 
Don't look at where we are? Look at where we're going to be. (I'll accept being the author of that quote...) copyright, trademark 9/7/2014

<<edit>>

mike_mccue
It seems like "optimization" should be a selling point, perhaps to justify a premium price...

That probably won't happen, because buyers choose products on gut-level criteria. A thousand optimizations won't ever be as effective as one Dan Worrall video.
 

Actually there is a lot of concern right now that in the next couple years Moors law might break.  We are starting to reach the limits of what is possible with silicon chips, once they start to get to the physical limits of smaller and smaller chips.
 
They are thinking right around 7nm is the limit, and we could hit that by 2020.
http://www.eetimes.com/document.asp?doc_id=1319330
 
After that we would need to switch to a new technology, maybe quantum computers or something.

The Band19
As the teraflops increase, this becomes less of an issue.  ... And therefore, IMO, "optimization" is a footnote. It's like driving down the autobahn in the right lane with your blinker stuck on? People are going to come flying past you in the left lane @ 200+ mph...

Great point.  Today's machines are fast beyond anything we could have imagined back in the 90s when PCs were just coming to become commonplace.  So a lot of software runs acceptably fast without a lot of effort spent on optimization.
 
Also, there's another factor.  Any piece of software, with rare exceptions, is going to spend an inordinate amount of time in a small percentage of the code.  5% of the code accounting for 80% or more of the cycles consumed is far from uncommon.  So optimization requires that the code base settle down before it's even worth optimizing in many cases.  In the case of FF, version one of Pro Q worked just fine speed-wise.  Given that much of the code was solid when they enhanced it to version two, they had something that was not a moving target, code that they code reliably analyze to see where the bulk of the  cycles were being consumed.

mike_mccue
I have all the FF stuff. I think of it as ok, and I'll be buying the Pro-Q2 *just because*.

 
Uh oh, cclarry's goan be all over you like white on rice for THAT decision.

Value packs, pop top lids, and a gift card for Costco in every garage.