KVR Audio

Aiynzahev · Post by **Aiynzahev** » Sat Jan 12, 2013 8:59 pm

Anyone know what the comparison would be?

As in equals about what a Pentium/Core2duo/Sandybridge etc

I do remember seeing a thread I think on GS where someone was taking about various synths to find that the Nord Lead 3 had something like over double the DSP power of the first TI.

Just curious is all.

Numanoid · Post by **Numanoid** » Sat Jan 12, 2013 9:03 pm

Aiynzahev wrote:Anyone know what the comparison would be?

I think a good match would be the one in Titan-Chrono

tehlord · Post by **tehlord** » Sat Jan 12, 2013 9:04 pm

I recall reading it's about the same as an iPhone 3

I have no idea if it's accurate, bullshit or anywhere in between.

EvilDragon · Post by **EvilDragon** » Sat Jan 12, 2013 9:21 pm

Virus TI uses two of these: http://www.freescale.com/webapp/sps/sit ... e=DSP56321

2x275 MHz.

An i7 kills this duo thousands of times in floating point operations for sure.

UltimateOutsider · Post by **UltimateOutsider** » Sat Jan 12, 2013 9:52 pm

The processor specs alone are sort of meaningless, IMO. But if you have a Virus TI2, it's pretty easy to see how poorly the Virus fares against a PC.

My PC has an old i7 950 CPU, and I can get up to 580 notes of Kontakt polyphony on it with my interface set to 256 samples, without any breakdown in audio.

My poor TI2 Desktop, however... I was recently working on a 12-part multitimbral project on my Virus, and started running into polyphony problems at around 10-12 notes.

Granted, some of those Virus parts were multi-voice, but none had more than 3 voices, and several had only 1. But I'll estimate a worst-case situation of 12 parts, with 3 voices each: 36 voices.

So that comes out to the Virus TI2/Desktop/Polar is about 2-6% as powerful as a three-year-old PC. (And a Snow is exactly half as powerful as that.) If you think purely in those terms, VSTis become a little more attractive.

EvilDragon · Post by **EvilDragon** » Sat Jan 12, 2013 10:04 pm

It's kinda apples and oranges comparison. Sample playback takes a lot less CPU resources than generating the sound in realtime.

UltimateOutsider · Post by **UltimateOutsider** » Sat Jan 12, 2013 10:32 pm

EvilDragon wrote:It's kinda apples and oranges comparison. Sample playback takes a lot less CPU resources than generating the sound in realtime.

Yeah, I was just thinking the same thing. Might do a more 1-1 comparison; design a similar patch on a VST and the Virus and see how far I can push them.

Still pretty sure the PC will win though.

Aiynzahev · Post by **Aiynzahev** » Sat Jan 12, 2013 11:09 pm

Well then it's clear its processing power is insignificant in light of pcs. I just wondered if Access felt they still needed dsp. Anyway, good sound is not relative to power, Sylenth1 proves that. So does Sonic Projects OP-X.

Many poorer synths can be very hungry and still sound bad.

BDeep · Post by **BDeep** » Sat Jan 12, 2013 11:44 pm

I'm by no means an expert, but aren't the chips used in HW synths specifically geared towards realtime dsp processing? In other words, because they're not designed to do a gazillion other things, they're much more efficient than an intel processor in performing audio type tasks. In that light, a direct comparison would be difficult, as they're basically apples and oranges.

Maybe a dsp guru can chime in...

Aiynzahev · Post by **Aiynzahev** » Sat Jan 12, 2013 11:49 pm

BDeep wrote:I'm by no means an expert, but aren't the chips used in HW synths specifically geared towards realtime dsp processing? In other words, because they're not designed to do a gazillion other things, they're much more efficient than an intel processor in performing audio type tasks. In that light, a direct comparison would be difficult, as they're basically apples and oranges.

Maybe a dsp guru can chime in...

I dunno, if you say 50% of your CPU is busy doing stuff, and 50% is dedicated to processing.

Well, if the processor is 4 times as powerful, then half of it is still twice as powerful.

But like I said, it's not that it ultimately matters to the sound, but it was interesting to me because the NL3 has much less polyphony, no effects and yet it had something like 6 times the power. And it's long been said that if you listen to the raw oscillators of the Virus they are not much in comparison to the raw sounds of the Nords.

The Nords do sound good raw, Virus sounds very good with it's complexity.

BDeep · Post by **BDeep** » Sat Jan 12, 2013 11:56 pm

Don't know if its as simple as doing other stuff. A dsp is designed for realtime audio, an intel isn't and therefor (I think) has to work harder to achieve the same results.
Please understand, I do think that today's processors are far more powerful as the HW chips, but I don't think there's a 1:1 relationship.

urosh · Post by **urosh** » Sat Jan 12, 2013 11:57 pm

BDeep wrote:I'm by no means an expert, but aren't the chips used in HW synths specifically geared towards realtime dsp processing? In other words, because they're not designed to do a gazillion other things, they're much more efficient than an intel processor in performing audio type tasks. In that light, a direct comparison would be difficult, as they're basically apples and oranges.

They are designed for signal processing but that means stuff like single cycle multiplication and addition, or even some combined instructions (like single cycle MAC), fast onboard memory access, stuff like that. OTOH, new x86 have some tricks up their sleeves like SSE instruction (in essence, parallel processing, thus multiple simultaneous operations). So, apple to apple comparasion would be to comapare execution time of some particualar algo on both systems (there are various DSP benchmark test out there based on this concept, but something like 64QAM demodulation is different thing compared to VA synths, so you need audio specific stress tests). Never the less, modern x86s are considerably (to put it mildly) faster than 56K used in Virus.

BDeep · Post by **BDeep** » Sat Jan 12, 2013 11:57 pm

And for the Nord vs Virus, I really have no clue, might just be better models, might just be inefficient programming

Aiynzahev · Post by **Aiynzahev** » Sat Jan 12, 2013 11:59 pm

BDeep wrote:Don't know if its as simple as doing other stuff. A dsp is designed for realtime audio, an intel isn't and therefor (I think) has to work harder to achieve the same results.
Please understand, I do think that today's processors are far more powerful as the HW chips, but I don't think there's a 1:1 relationship.

That's interesting, I didn't think of that.

BDeep · Post by **BDeep** » Sun Jan 13, 2013 12:06 am

Another thing that might be in play here, is the restrictions of the available power to the developers, which lead them to more thought on how they designed the code, and more creativity. Think about the limited memory on older Korg and Roland romplers, yet (some of) their sounds rival (some of) today's multi GB sample libraries.
Of course that is pure speculation.

Virus TI2 DSP power in relation to todays Intel chips