Plug-In DSP usage Scale

PaulBrewer

Is there an agreed way for comparing how much DSP a plugin uses, specifically in PTools HD ?

DamagedTrax

too many factors. processor, coding, format (tdm, tas, vst with wrapper)

the only agreed test on pt hd that i know of is to test actual processor power. thats the deverb test that they run on DUC forums.

for plugs most people just go on % of cpu (or cards in the case of hd)

PaulBrewer

But surely there could be an agreed set of parameters that a plug was compared too ?

Say A TDM system.
HD2/44.1kHz/24Bit/48 tracks = 36 instances or whatever .....

Has this ever been done ?

I agree on a native system it would be hard to do.

woodsdenis

PaulBrewer wrote: »

But surely there could be an agreed set of parameters that a plug was compared too ?

Say A TDM system.
HD2/44.1kHz/24Bit/48 tracks = 36 instances or whatever .....

Has this ever been done ?

I agree on a native system it would be hard to do.

If your on TDM use your Window/System Usage. TDM plugs use the same resources because they are non native. A HD2 Accell on a G5 will perform approximately the same as a HD2 on a Mac Pro. RTAS is a different kettle of fish entirely, there will be a huge difference between different systems/ram/ etc. There were benchmarks done with things like Dverb, which is a standard plugin in all PT systems, to evaluate how good a particular version of a Computer/System/PT version is.

Basically you load up as many RTAS Dverbs as you can until your system
gives you a "not enough CPU" message. I am sure its on the DUC somewhere.

PaulBrewer

woodsdenis wrote: »

If your on TDM use your Window/System Usage. TDM plugs use the same resources because they are non native. A HD2 Accell on a G5 will perform approximately the same as a HD2 on a Mac Pro. RTAS is a different kettle of fish entirely, there will be a huge difference between different systems/ram/ etc. There were benchmarks done with things like Dverb, which is a standard plugin in all PT systems, to evaluate how good a particular version of a Computer/System/PT version is.

Basically you load up as many RTAS Dverbs as you can until your system
gives you a "not enough CPU" message. I am sure its on the DUC somewhere.

Thanks , It seems to me it could be useful info if it was available.

woodsdenis

PaulBrewer wrote: »

Thanks , It seems to me it could be useful info if it was available.

http://duc.digidesign.com/showthread.php?p=1317682#post1317682

There you go

PaulBrewer

Thanks for that Denis.

Anima

What exactly does that prove?

DamagedTrax

Anima wrote: »

What exactly does that prove?

absolutly nothing except how powerful your processor is when running PT.

woodsdenis

DamagedTrax wrote: »

absolutly nothing except how powerful your processor is when running PT.

Precisely, you could do the same for Logic running Platinum Verb for example.

PaulBrewer

My thinking was if a Plug could be 'rated' for CPU or DSP usage it would be useful info for planning on how to 'retrieve' some of that power when one hits the inevitable 'Phuk off, ya can't use any more plug ins Ya EEjit' message from ProTools.

I know that USR for example make a 'Lite' version of a plug that comes with a main plug that's less of a load.

DamagedTrax

since every system is differant the only way to know is to add the plug and see what % cpu it uses.. even that is flawed as some daws scale their usage. with reaper the jump from 10% to 12% is a lot less plugs than the jump from 40% to 42%. then theres load balancing to account for. its a minefield :eek:

Anima

DamagedTrax wrote: »

since every system is differant the only way to know is to add the plug and see what % cpu it uses.. even that is flawed as some daws scale their usage. with reaper the jump from 10% to 12% is a lot less plugs than the jump from 40% to 42%. then theres load balancing to account for. its a minefield :eek:

Scale their usage? How do you mean?

PaulBrewer

DamagedTrax wrote: »

since every system is differant the only way to know is to add the plug and see what % cpu it uses.. even that is flawed as some daws scale their usage. with reaper the jump from 10% to 12% is a lot less plugs than the jump from 40% to 42%. then theres load balancing to account for. its a minefield :eek:

But that's my point - if you check and record a Plug's DSP usage against an agreed 'standard' system them whilst you may not have the absolute measurement you'll know that Plug A uses twice as much steam as Plug B as an example ... and plan accordingly.

DamagedTrax

Anima wrote: »

Scale their usage? How do you mean?

im not sure how it works but its documented that reaper gives you better plugin performance per cpu cycle as you use more cpu.

PaulBrewer

I see URS do exactly this -
say how many instances of a plug will run per DSP chip.

http://www.ursplugins.com/ursBLT/ursBLT.html

trackmixstudio

Very interesting.
As an experiment based on what Paul said about URS I tested to see how many Channel Strip Pro instances I could run and then how many Strip (lite version)
Running a quad 2.6 mac pro with 6gb ram.
It came out as 94 CSP







and a whopping 225 strips







I would be very interested to see how many you can get on PTHD if you would run a test Paul.

woodsdenis

To do this natively you would need the exact same spec of machine. In RTAS/Native my guess I would you would get nowhere near that number. Any one with LE could do this. I would be interested.

In HD it is not really comparable, it would depend on the amount of cards that you had in your system. a HD1 no chance. A HD6 maybe , I dont know.

I suppose you could find out how many you could get on 1 chip and multiply?


Strip Pro Instances - 5 per HD Accel chip and 2 Instances per HD chip


Edit, Here is the answer from their website, which would make 120 instances on a HD3 system versus 94 on your system, in theory. In practice if you are running ADC you would get a lot less than that.
The advantage with HD is you dont touch the computers processing for this kind of test.