Dune 2 - I'm speechless, what have they done

[MBTC]

Originally Posted by grs

I wasn't claiming 300%. There were other factors like the amount of plugins on other tracks etc.

I got the 300%-500% number from when you said you could barely run 2-3 instances on the i7-950 to being able to run 7-10 instances on the i7-4770K. That's roughly a 300% increase minimum.

Again, know that I'm not challenging you on this -- I believe you, I'm just trying to understand how it might be possible. I'm all about more CPU power=better, but there are subtle differences between various processors and I'm trying to understand how the results you got with your upgrade and Spire could correspond to that.

[grs]

Originally Posted by MBTC

I got the 300%-500% number from when you said you could barely run 2-3 instances on the i7-950 to being able to run 7-10 instances on the i7-4770K. That's roughly a 300% increase minimum.

Again, know that I'm not challenging you on this -- I believe you, I'm just trying to understand how it might be possible. I'm all about more CPU power=better, but there are subtle differences between various processors and I'm trying to understand how the results you got with your upgrade and Spire could correspond to that.

Ah, I see how you could take that as 2-3 instances to 7-10 as 300-500% - 2 to 10 equals 500%. This is not what I meant. What I meant was more literally 2 to 3 "ish"(2.5?) instances to 7 to 10 "ish"(8.5?), so 2.5 to 8.5, that's what I am experiencing.
So now I will correct myself to mean roughly 3 instances would kill my 950 (stutter audio playback with high 1024 buffer) and 6 instances don't kill my 4770 (same audio card and buffer) and as yet although I have 13 instances of Spire loaded in my current production only 3 to 7 play at any one moment, also the CPU meter in Windows 8.1 never pops up higher than 45% - so I'm cruising in CPU nirvana.

Also I said later that it was in context of separate songs with different other tracks.

I wish there were better audio vst tests for CPUs but they seem randomly focused on Cubase, Reaper, Fruityloops or Massive etc. so not everyone can chime in and test their system against the others.

[MBTC]

Originally Posted by grs

I wish there were better audio vst tests for CPUs but they seem randomly focused on Cubase, Reaper, Fruityloops or Massive etc. so not everyone can chime in and test their system against the others.

Which host do you use? Some time back, just to be sure the results I was seeing with Massive were not DAW-specific, I loaded up a trial of Ableton just for comparison purposes, even though I'm not very familiar with it.

[grs]

Yeah, Ableton. Maybe you can PM me the project file or post it for the public to try out. although you may not be able to save in trial version

[MBTC]

Yeah I don't think I ever saved the project file, but it's not really needed, you just insert an instance of Massive, load that patch, play a single sustained note, rinse and repeat until you hit an upper limit (but of course you need Massive -- I assume there is a trial out there but I'm not sure if it has additional overhead).

As a side note I do recommend getting some version of NI Komplete which would have Massive included in it for free. I was recently watching a movie called "Sound City", about the famous L.A. record studio (when it closed, Dave Grohl bought the extremely rare Neve mixing board that was responsible for the sound of so many classic records, and made a documentary of it). To some extent, modern music production gets slapped around in it a bit (they bash Pro Tools mostly), but one of the musicians that comes and jams with Dave is Trent Reznor from Nine Inch Nails. In the background you can see that he is using Native Instruments stuff (my point is that it says a lot about Komplete).

[TweakHead]

There are some new instructions added. Plus, it's not just the processor but the kind of chip they're using on mother boards now. A lot has changed, for the better.

Also, one has to realize that, for instance, some of these new instructions can make a huge difference - more so then you'd get from simple tests - if developers decide to code them in to their products, making them a huge advantage.

Modern CPUs have things like hyper threading and turbo boost, are mounted on much more efficient chips. First generation i7 compared to recent ones? You're in for a huge surprise... It's really come a long way m8!

[MBTC]

Originally Posted by TweakHead

There are some new instructions added. Plus, it's not just the processor but the kind of chip they're using on mother boards now. A lot has changed, for the better.

Also, one has to realize that, for instance, some of these new instructions can make a huge difference - more so then you'd get from simple tests - if developers decide to code them in to their products, making them a huge advantage.

Modern CPUs have things like hyper threading and turbo boost, are mounted on much more efficient chips. First generation i7 compared to recent ones? You're in for a huge surprise... It's really come a long way m8!

HT and turbo have been around a while though. I bought a 4770k based system about a year ago and actually returned it because real-world performance was not that much better than the i7-965 in my primary system (the 965 came out five years ago). The 870 I mentioned that I'm using for my audio host is overclocked and is slightly slower than the 965.

Even if the new ones do have new instructions (I'm not aware of specifics around that), applications would have to be compiled to take advantage of them, making them backward incompatible with old processors (and of course they aren't). The biggest hurdle for audio is that most audio-related algorithms, except for perhaps LAME encoding and so forth, do not lend themselves well to multi-threading. So while the multi-threaded performance of some of these newer six-core processors is indeed impressive in synthetic benchmarks or in situations that strongly take advantage of multi-threading, it's hard to justify an upgrade at that point.

These numbers compare a 4770K to a 975 which performs about the same as my 965.

http://www.tomshardware.com/charts/c...d%5B5750%5D=on

You'll notice the truly impressive performance gains are in things like the encryption score (a usage that can utilize multiple threads nicely), you can ignore the ridiculous numbers like the PCMark 7 number (drill down into the chart and you'll see what I mean, that's a website error), etc. Something like the Photoshop score shows a significant benefit, but that's because image processing is a very parallelizable type of operation. If you start looking at everything else you start to wonder "really? a 10-20% difference represents five years of CPU progress"?

There are good reasons CPU technology has brick walled (or at least slowed greatly in terms of year over year real world performance benefits) that aren't anyone's fault, but also Intel has been focusing on adding video performance and reducing reliance on Nvidia and AMD for that which I don't think is a good use of their time and R&D resources, personally.

All of this said, I don't want anyone to think I'm doubting that the newer ones are better. But since overall synthetic benchmark results tend to show about a 40-50% increase over the last 5 years, and real world results tend to show much less than that, I'm inclined to believe that a plug-in that's getting 2x-3x gains from newer processors has something else going on in terms of multi-threading technique that I'd be curious to know more about.

[grs]

OK, couldn't sleep last night so I ran my own Ableton with spire bench mark.
Quick results are..
Usable instances
4770k 27
i7-950 8
4770k better-ness 3.4
---------------
cpu use (task manager process) divided by instance
Instances 1 2 3 4 5 6 7 8
4770K 4.0 2.5 2.7 2.8 2.6 2.8 3.0 3.0
i7-950 8.0 6.5 7.0 6.8 7.4 7.7 7.4 8.0
better-ness 2.0 2.6 2.6 2.5 2.8 2.7 2.5 2.7

Test detalis: Spire patch BA Octava, (two oscilators 9 voices unison). Ableton Live 9.1

[MBTC]

Do you have a midi file you're using to play them? On my i7-870 using the Spire demo, BA Octava on Cubase I'm only able to insert 25 instances before the output is no longer sent to the mixer (I'm guessing that's a Cubase thing or maybe demo limit), and of course so many of them sounds horrible but I guess subjectively I could say it's usable (peaking out on the Cubase CPU meter sometimes but no crackling or anything that would prevent use).

But that's just dinking around playing notes by hand so I thought a MIDI file might be a better point of reference if you have it to share.

Also with 25 instances generating the demo white noise swoosh it makes it hard to claim it sounds usable.

[grs]

Firstly, it is a stress test like the one you mentioned with Massive. It is the same patch playing a held note,, you turn down the volume of the track to like almost nothing and then duplicate it.
"Usable instances" refers to the one patch over and over. "Usable" relates to how many instances will be duplicated until live peeks over 100% usually you hear crackling around 90% to 95%.
just make the same note on the same patch (BA Octava, to be a comparrison to my results) playing at the same time in cubase and see if you can get to 25.
I used an F below middle C held for two bars with a tempo of 77. loop that 2 bars and copy that midi into every Spire track.
To measure I used windows task manager process view and put update speed on low. After turning on each instance I wrote down the cpu usage in windows task manager for the process of the DAW and the cpu usage in Ableton Live top right corner.