ACE a cpu-killer?

Official support for: u-he.com
RELATED
PRODUCTS

Post

Urs wrote:The problem is, the cheapest synth that delivers what ACE does costs $3000+ second hand and is still monophonic. ACE is a cheapskate even if it means you have to buy a second computer to run it. My 3 year old model runs 24 ACEs without a problem, so hmmm... dunno...

Actually, if ACE was a cpu friendly fellow like Zebra or Helix, I would have priced it 790$, and I'm sure I would have also sold 10 times as my licenses by now. It is just how it is, computers are freakishly slow when it comes to sound quality.
Image
Thank you Urs :D!
Image
stay juicy!

Post

Also, Sonic Core's Xite-1 DSP Audio Platform will do everything ACE does and is way more powerful than CPUs now - but you pay £2500........

Urs..you should have a go at Scope development....

Post

strav101 wrote:Also, Sonic Core's Xite-1 DSP Audio Platform will do everything ACE does and is way more powerful than CPUs now - but you pay £2500........

Urs..you should have a go at Scope development....
Nah, the development cycles seem way too long, and too many companies have gone bust over Sharc development.

In a year or two ACE will seem as cpu friendly as Zebra does now.

Post

Yes - I would agree reading this from Wikipedia : :)


Haswell is the code name for a processor that is being developed by Intel and is the planned successor to Sandy Bridge. Haswell will be designed for the 22 nm manufacturing process, introduced with Ivy Bridge, and is planned for a commercial release in 2012. Haswell should be the first Intel x86-64 CPU to implement FMA (Fused Multiply-Add also known as Multiply-accumulate) instructions.[1] Haswell's design is in the early stages of development. Like Nehalem, Haswell is being developed by Intel's Oregon team rather than by Intel's Israel Development Center like Core and Sandy Bridge.[2] One tech journalism site claims that Larrabee graphics capabilities will be integrated in the Haswell microarchitecture.[3]
[edit]Architecture

Haswell is expected to have the following features:[1] [1]
Possibly completely redesigned microarchitecture, like NetBurst was after P6
22 nm process
8 cores by default
Entirely new cache design
Revolutionary power saving systems
Possible on-package vector coprocessors
The addition of fused multiply-add instructions
Possibly a one-year transition from Sandy Bridge[citation needed]

Post

strav101 wrote:
Haswell is expected to have the following features:[1] [1]
Possibly completely redesigned microarchitecture, like NetBurst was after P6
22 nm process
8 cores by default
Entirely new cache design
Revolutionary power saving systems
Possible on-package vector coprocessors
The addition of fused multiply-add instructions
Possibly a one-year transition from Sandy Bridge[citation needed]
You forgot one feature:

Will easily run the ACE software synth

Post

hakey wrote:NI's Massive, AFAIK, doesn't do any audio rate modulation.
AFAYK is not F enough in this case...
If you like 80s retro sounds, check out my latest tune…

Post

With Intel i7 920, Ace cpu usage seems tame to me.
Anyone else using an i7 cpu?

Post

gurulogic wrote:With Intel i7 920, Ace cpu usage seems tame to me.
Anyone else using an i7 cpu?
Yes, the OP.
If you like 80s retro sounds, check out my latest tune…

Post

Urs wrote:
goldenanalog wrote:I don't think that there is a $3000+ 'second' hand monophonic synth that sounds like ACE
I didn't mean the sound. I'm fully aware that nothing sounds like ACE and vice versa.

What's bought by ACE's cpu load is the fun that is patching anything into anything else. No more no less. The only thing you can currently buy that delivers the same experience (at an even lower pain level, to be fair) IMHO is the Cwejman semi modular synth, or a fully loaded module rack plus a giant bank of multiples and VCAs. And this doesn't do supersaw either, at a much higher cost of ACE + PC.

Answering a previous post, the quality settings are indeed mostly about oversampling, but also about evaluating frequency modulation (LFO/VCO/VCF frequency, including cross modulation) only every single or sixteenth sample, as compared to e.g. 4 times a sample. Everything else is still evaluated on a 1:1 basis.

;) Urs
Urs: Thank-you very much for the explaination! OK, I get it, now. As usual: it's all about the math, isn't it?

And, of course, as PC technology advances, ACE will become easier and easier on our machines.

I'm still left with the same opinion(s) of ACE, but I withdraw my 'push' for precision selection of polyphony; I simply don't want to see you try to rewrite ACE just to please (perhaps) a few individuals such as myself that want to build a CPU-fence around ACE @ runtime, so that it can be fully utillized (and trusted) in a live situation.

On the merits of sound alone, ACE is a masterpiece....what I was thinking about in terms of (indirect, but parallel) comparison was the Hartmann Neuron:

http://www.audiomidi.com/aboutus/review ... ron_vs.cfm

which like ACE cuts a wide swath through the sea of: ME TOO! vsti's on it's way to unchartered sonic territories....truly revolutionary beasts.

No, I don't think that a CPU-efficient version would sell for $790.00US 10X as well in our present market; timing is everything, Urs.

So perhaps that's what we are really talking about, is ACE being ahead of it's time in terms of it's demands on PC resources; perhaps in 2 years, this thread will be obsolete.

I love both Poly-Ana and WOW 2600; they sound great to me, and are controllable. ACE is marginally useable (to me, anyway), but I have NO REGRETS about giving you money for it, Urs. It is a bargain by any set of metrics; I just can't do what I want with it w/o buying a computer another order of magnitude more powerful then I currently own to safely run it on.

-goldenanalog

Post

goldenanalog wrote:ACE is marginally useable (to me, anyway), but I have NO REGRETS about giving you money for it, Urs. It is a bargain by any set of metrics; I just can't do what I want with it w/o buying a computer another order of magnitude more powerful then I currently own to safely run it on.

-goldenanalog
If nothing else, use it as a monophonic/duophonic synth.

Post

PaulSC wrote:
gurulogic wrote:With Intel i7 920, Ace cpu usage seems tame to me.
Anyone else using an i7 cpu?
Yes, the OP.
Oops, I should read more thoroughly!
now I'm wondering what I have to do to get 50% cpu on with one instance of ACE on my i7 like the OP? I can just push 13% @ 128 samples if I try really really hard. Maybe I'm not digging deep enough?
Anyone want to post a test .fxp and midi file so we can compare results?

Post

ACE certainly needs quite some CPU power, but I also think that an I7 processor should be able to handle that.

I made this little ACE-track http://www.trippler.net/files/ace/coldasace.mp3 on a dual core 2.4 GHz PC without any freezing of tracks. CPU meter reached around 75% at particular parts of the track.

Post

I wonder if maybe there is some kind of denormal problems...

is there some particular patch exposing the problem?

Have you tried reading this?

http://www.digitalfishphones.com/main.p ... &subItem=6

- Mario

Post

gurulogic wrote: now I'm wondering what I have to do to get 50% cpu on with one instance of ACE on my i7 like the OP? I can just push 13% @ 128 samples if I try really really hard. Maybe I'm not digging deep enough?
Anyone want to post a test .fxp and midi file so we can compare results?
Just take one of the presets like "BT Broad'n Smooth (mw)" and make it run an arpeggiated pattern - then change the quality setting (top left) from draft, through standard to good and then finally accurate.

You can bring an i7 to its knees, in the same way you can with various settings on any of the recent 'analog' VSTs. The main 'request/complaint' I think is that the 'voices' option only gives a few vague choices rather than an exact number of how many voices can play, which Urs replied as to why it's like that.

I'm sure I'll be corrected if that summary is incorrect :wink:

Post

mabian wrote:I wonder if maybe there is some kind of denormal problems...
It's just (again, I think) to do with too many voices of high quality (calculation) at the same time. When the options are set (in)correctly, voices won't be dropped and calculations won't be shortcut in order to save CPU, thus dropouts/collaps occurs.

This isn't a problem - it just means you have to know the limitations.

It's the same as upping up the photon count, render bounces and displacement size in 3D applications and wondering why a simple scene with a few items is suddenly taking 6 hours instead of 6 seconds. There's a trade-off between perfect quality and speed that is always made - just the majority of instruments limit you so you don't complain they take too much CPU.

Post Reply

Return to “u-he”