DSP comparison request

[siriusbliss]

Can any of you geniuses give a layman's (or technical breakdown is OK too) comparison of the DSP horsepower of Xite vs. say, one of today's quad-cores?

I'm just curious since people are already drooling over next gen. quad/octal cores, 64-bit, 16G RAM hoohahs.

I obviously know that having all those DSP's in Xite are light years ahead of anything out there.
Just wanted a good description to be able to explain to the unawares out there (and a friend is considering getting a Xite).

Thanks,

Greg

[netguyjoel]

holy poop!

what a killer question! i would like to know too..

thanx Greg!

as all ways....

Joel

[Neutron]

Its a completely different thing even if you measure it in mflops or whatever.

CPU is good for general purpose computing. leave it free to do what it does so well. (and windows/osx are not real time operating systems)
DSP is good for processing real time signals on time every time. it is much closer to a digital hardware synth than a VST effect or instrument.

[netguyjoel]

http://www.hitequest.com/Kiss/risc_cisc.htm

great comparison


Joel

[Sounddesigner]

If you meausure Realtime performance XITE will be waaay more powerfull then any computer, regardless to amount of cores.

[siriusbliss]

I totally understand the pure horsepower of Xite over the 'distracted' multi-tasked, multi-threaded, bus-laden CPU's, but I'm just looking for real apples to apples numbers (disregard OS bullcrap - who cares?).

And I would also prefer this not also spin into the obvious native VST vs. Xite hardware redundant comparison argument.

Greg

p.s. thanks for the links - will check...

[siriusbliss]

http://www.hitequest.com/Kiss/risc_cisc.htm

great comparison


Joel

yeah, this is a good overview (from 2002), thanks.

Greg

[Shroomz~>]

You can't compare something like XITE-1 to a general computer (no matter how fast). It just doesn't make any sense to me. Which is better - guitar or piano?

[siriusbliss]

You can't compare something like XITE-1 to a general computer (no matter how fast). It just doesn't make any sense to me. Which is better - guitar or piano?

I understand.

Again, I'm only requesting direct apple-to-apples comparisons in the realm of DSP horsepower.

All the other subjective comparisons are not required for this inquiry.

I also realize it's not a direct comparison, but ignoramuses out there are arguing over DSP horsepower on latest-n-greatest $5K+ computers vs. dedicated DSP. Or they're <ahem> comparing VST's to Scope synths. I could compare guitar-to-guitar til my face turns blue, but there are still more direct comparisons there (as an example).

This is only an inquiry - not an argument over the obvious reality we're dealing with with our Scope systems.

One main reason I'm asking this is, that I'm getting ready to do some long-awaited articles on Xite, and I am seeing that this device may just broadside the industry a little out of left field (while everybody's waiting on the latest VST this or 64-bit that). I think we need REAL simple distinctions to get the point across as to how powerful Scope is.

Greg

[Mr Arkadin]

i just don't know how you are going to directly compare them. i mean you for example see how many instances of Minimax you can run on XITE before it keels over and how many instances of Minimmog V on a Quad computer, but even that isn't a direct comparison: who here feels that Minimmog V is anywhere near as good as Minimax? So the extra quality of Minimax may mean less instances. Also that test just shows instances and not a working system which would require a host to run on the computer for both systems - which of course will affect the results of the computer.

i just don't see how you're going to get any mathematic figures, i don't think it can be done and make any real sense.

Have a go by all means though .

[Tau]

I agree with Mr. Arkadin - it's going to be hard to compare both systems in terms of raw power and come out with apples and apples... IMO, the greatest advantage of DSPs in general, is their ability to process audio in real time, and that is essential in many studio applications such as recording, playing and interfacing with outboard or DAW, and also in live applications, providing monitoring, sub-mixing or FOH. It's really the same as hardware...

But it all comes down to the needs of each person. Great sounds and mixes can be had either way. I believe that when your friends see you working with your XITE, they'll not only get a better idea of the differences and possibilities, but they'll be wanting one for themselves Well, maybe, maybe not...


T

[siriusbliss]

Thanks for the further input guys.

Yes, I wasn't sure I'd get any real geeky numbers or not, but thought I would at least ask.

Besides, I'm not even sure once I HAD numbers (like MIPS, or GHz throughput, or FLOPS overhead, or whatever), that anyone would really see how that would interpolate into real-world scenarios (such as the reverb test).

And, my friends/associates - one doing inde films, and the other MIDI-centric DJ stuff - have already seen my current 23 DSP Scope system, and are blown away by the horsepower and getting lots of tracks and processing with the CPU still only peaking at about 50%, and the Scope humming along at about 30%.

Still would be nice to have even some generic techy distinctions (such as the RISC/CISC comparison) for some extra 'oomph'.

Greg

[spacef]

Can any of you geniuses give a layman's (or technical breakdown is OK too) comparison of the DSP horsepower of Xite vs. say, one of today's quad-cores?

I'm just curious since people are already drooling over next gen. quad/octal cores, 64-bit, 16G RAM hoohahs.

I obviously know that having all those DSP's in Xite are light years ahead of anything out there.
Just wanted a good description to be able to explain to the unawares out there (and a friend is considering getting a Xite).

Thanks,

Greg

A fun comparison, which is true but not applicable in audio, is to compare Scope (not even xite) with photoshop.
Think about it: if you know photoshop, you know that even on a Core2Duo or a Quad, photoshop is unable to deal with as many "live" layers as scope does everyday.
For example, a synth or mixers, with all this potis images, the background, the text, the various layers, is always real time, when photoshop struggles in basic task with not even half of those layers.
and scope does it without complaining, and it can updates vu meters and potis or faders in real time.

Now for audio generation (synths), it is different, because polyphony is the weakness of the dsps, because Envelopes take a lot of ressources. But i think scope quality is always better, and it still does things impossible to do with softs like synthmaker and synthedit. Also, there is a problem for samples because on scope they need to pass-through the pci-buss, whereas native samplers don't need to know what is a pci-bus (nowadays they would just use hard drives, ie sata, or other formats).

It is the quality of anything made for scope that is superior to native formats, but for this i think you need to hear it.... It is like the difference between a cheap pream and a quality preamp: for example, not everyone is able to see the point in a 2000 euros preamp compared to a 500 euros preamp. You would need to hear it, but also understand that is better, which may be a question of experience...
Now, if there are scripting abilities in xite for "major and renowed third parties companies/developpers" then i would expect that all that makes other boards succesful will be also on xite (and the "names" too)... But until then, Scope is closer to hardware than vst instruments which, sorry to anyone, are still far away in terms of quality of many things (filters or analog "emulation").

Just coffee talks, not sure comparing it to photoshop is a big selling argument for the prospective buyers

[siriusbliss]

The Photoshop comparison is a unique angle. May actually help give a distinction to the overall paradigm leap that Scope/Xite is.

I'm trying to avoid sound comparisons because frankly, no offense, but I think the younger generation has been listening to crappy mp3's for so long, that they wouldn't know good quality audio (or be able to feel the grit of a good filter) if it hit them in the head.

The less subjective comparison is yes, the real-time, low latency aspect, graphics response, and being able to do large projects.

All I really need is something like 'Xite is approximately the same horsepower as having 4 quad-cores and 32Gig of RAM with a 1TB FSB' (or whatever)
Also, should probably distinguish the PCI-e vs. firewire/USB throughput speeds, but that's somewhat of a side-note.

Thanks again,

Greg

[sonolive]

hi greg,

have a look here ...

http://www.dspguide.com/ch28.htm

no answwer like "it is twice more power than ..." but a good introduction in dsp vs other cpu

cheerz
olive

[spacef]

The less subjective comparison is yes, the real-time, low latency aspect, graphics response, and being able to do large projects.

from the start, Scope doesn't need any sound driver (asio, wave...) and therefore suffers "no" latency.
I recently measured the "latency" on scope and according to my measures (which are made with a plugin, not with stock atoms), it is around 1/3 of a milisecond, and was always below 20/22 samples (this doesn't measure a general latency on scope for synths for example, this would need another measurment).

In my measures there was various other factors such as various calculation before the triggger (due to the plugin i was testing) and the general enveloppe-minimal attack time , but i think this measure is very close to reality and that you can trigger an envelope with less than 1/3rd of a milisecond "latency", may be even less (i was making my measure on a "non-manually triggered" plugins, so many othr factors were to be taken into account). But, with the sdk, you can get faster envelopes than with the stock-envelopes of the SDK.

I mean, if you want to play scope at the same time as vstis, you will have to delay scope by an amount equal to at least half your total asio latency, because in any case it is faster than asio....

Leads to another question; what is the minimal ulli/asio latency on xite ?
i didn't check that at musikmess...

[next to nothing]

"I'm trying to avoid sound comparisons because frankly, no offense, but I think the younger generation has been listening to crappy mp3's for so long, that they wouldn't know good quality audio (or be able to feel the grit of a good filter) if it hit them in the head."

You're more right than you might think.

http://technology.timesonline.co.uk/tol ... 847674.ece

[astroman]

I can verify the 'getting used to...' effect after listening to several old YouTube clips, where the 'rare' aspect compensated for quality
But a iPod is an almost audiophile player, according to my own Sennheisers I plugged in...
The encoding is sh*t, not the device...

cheers, Tom

[astroman]

I totally understand the pure horsepower of Xite over the 'distracted' multi-tasked, multi-threaded, bus-laden CPU's, but I'm just looking for real apples to apples numbers (disregard OS bullcrap - who cares?)...

I've posted a PDF whitepaper about a (serious) comparison between TigerSharc and PPC G4 a couple of times, but the company (Bittware) took it offline.
They coded a FFT application which was optimized for the respective architecture.

The interesting point was they also predicted results according to estimated 'processing power'
While the TS DSP was slightly off, the G4 delivered only 30% (!) of the predicted figure in a real world application

your layman's bottomline as requested:
CPU power is speced by artificial tests with data aready loaded into the CPU core.
As soon as your app has to move it around - and a realtime audio processor got to move lots of stuff... you can just forget about it. Seriously.
Mind you: the example above dealt with hand-optimized machine code by industry experts on a rather simple (and well known) subject. Expect typical audio application to be anything between 5 to 20 times less efficient.

Yes indeed, you have a ton of CPU power today... half of which is wasted on threads you really don't want, and the other half only delivers 25% (let's be generous) due to architecture, inefficient tools and whatever...

That said about native CPUs and marketing blurb - DSP don't suffer that hard from the latter as they adress a less emotional market.

cheers, Tom

ps: this is not intended a native CPU diss, but against marketing's bigger-is-better bs

[nightscope]

http://sonalksisdev.blogspot.com/search ... ve&x=0&y=0

ns