Synth and Effects CPU Benchmarks

VST, AU, AAX, CLAP, etc. Plugin Virtual Instruments Discussion
Post Reply New Topic
RELATED
PRODUCTS

Post

While doing some research on the latest CPU's from intel and amd, then comparing then to my current one, I find that rather than cpu benchmarks, I am more curious on how many more synths of x or z I would be able to run. For instance, amd's threadripper has a massive cpu benchmark of around 26000, putting my 5000 benchmark i7 to shame. Does it stand to reason that I will be able to load 5x as many lush-101 instances given the difference between benchmarks?

Given that cpu's have benchmarks, how much maximum processing they can do, and they hold despite differences between different computers and system specs, is it at all pheasable to quantify in a similar way how taxing a synth or effect is relative to cpu. For instance, If I can load and max out my system with 15 lush-101 instances with 8 voices at maximum unison, I could divide my cpu benchmark of 5000 and come up with a number that could somehow indicate a benchmark for the synth and the same ratio will hold true for different processors. In this case 5000/15 would be 333.33, so in effect, could I say lush-101 has a benchmark of 333 then do another calculation to see how many I could run if my cpu benchmark were 26000, so that number divided by 333, yielding 78.

In short, is it possible to give benchmarks to synths or effects and have them be universal?
High Quality Soundsets for Lush-101 | Hive | Electra 2 | Diversion | Halion | Largo | Rapid | Dune II | Thorn | and more.

TTU Youtube

Post

In general yeah, a CPU benchmark will tell you roughly how many synths and effects you can run (comparatively). So if you're one machine gets a 5000 score (multicore) and you know you are able to run 10 of a certain synth until failure/crackles, then a machine with a 10000 score of the same multicore test should let you run about twice as many. There can be slight differences between machines and sometimes higher powered ones may have some throttling when it heats up so you never know for sure though.

Post

Good to know. I just thought it would be cool to see a benchmark cpu list for synths. In this case lush-101 would get a 333, maybe dune 2 would get a 100 etc. So anyone looking at the list can cross reference with there cpu benchmark and could see how many more of the synths they could run when they upgrade.
High Quality Soundsets for Lush-101 | Hive | Electra 2 | Diversion | Halion | Largo | Rapid | Dune II | Thorn | and more.

TTU Youtube

Post

Touch The Universe wrote:Good to know. I just thought it would be cool to see a benchmark cpu list for synths. In this case lush-101 would get a 333, maybe dune 2 would get a 100 etc. So anyone looking at the list can cross reference with there cpu benchmark and could see how many more of the synths they could run when they upgrade.
No you can't give a synth a specific cpu score of any kind because the cpu depends on not only the patch (which features are being used) but also how many voices are being played. So if you want to get an idea for the cpu use of a synth compared to others you need to have some sort of standardization in the test. Like playing a 5 note chord with all the oscs on and some filtering etc. Some synths have effects, some don't etc. A synth is software much like whatever they use for the benchmarks. It could be used to test the machine's power, but you're testing the machine, not the synth.

Post

Yes, that thought occurred to me as well. It would be quite difficult to standardize the test since patch complexity can be so diverse, thought I still think some sort of intelligent median could be used as a standard. If a test were done on a synth, it would have to be on patches equidistant from init to max polyphony, effects, modulations, etc, somewhere along the lines of where most patches would fall under. Even maxing out on the synth could still yield some kind of useful figures. At the very least, you could now minimum number of synths you could run.
High Quality Soundsets for Lush-101 | Hive | Electra 2 | Diversion | Halion | Largo | Rapid | Dune II | Thorn | and more.

TTU Youtube

Post

Touch The Universe wrote:Yes, that thought occurred to me as well. It would be quite difficult to standardize the test since patch complexity can be so diverse, thought I still think some sort of intelligent median could be used as a standard. If a test were done on a synth, it would have to be on patches equidistant from init to max polyphony, effects, modulations, etc, somewhere along the lines of where most patches would fall under. Even maxing out on the synth could still yield some kind of useful figures. At the very least, you could now minimum number of synths you could run.
Nah, it would be often too misleading. There are many synths that you can max out with a single instance that for normal usage are pretty low CPU. I can think of lots examples where there might be a certain effect that makes it much higher like Omnisphere or absynth. Or many synths are modular and you can add as many modules or effects as you want, yet very low CPU for many aspects, like falcon with its va vs ircam stretching. We are talking orders of magnitude differences between patches and averages or max usage just isn't useful.

For many synths it is like asking how CPU intensive a techno project is in a given daw. With infinite possible varieties of synths and effects.

Post

Old and new versions of DAWbench are the most reliable indicators of performance for us, but not too many reviews run the test of course. There's an excellent link in the Threadripper thread here from Scan that discusses Threadripper's performance and uses the tests.
http://www.guyrowland.co.uk
http://www.sound-on-screen.com
W10, i7 7820X, 64gb RAM, RME Babyface, 1050ti, PT 2023 Ultimate, Cubase Pro 13
Macbook Air M2 OSX 10.15

Post

Hehe, I had a similar idea and thread some time ago :)

viewtopic.php?f=1&t=424295&p=5929754&hi ... k#p5929754

Post

In the games industry, certain games are used as benchmarks when testing hardware. Maybe something similar could be done with synths - pick a cpu hungry synth or 3, such as Diva in Divine mode, use a standard set of patches, etc.

Post

Googly Smythe wrote:In the games industry, certain games are used as benchmarks when testing hardware. Maybe something similar could be done with synths - pick a cpu hungry synth or 3, such as Diva in Divine mode, use a standard set of patches, etc.
Yes but again, this is using synths to test hardware, which makes sense because you can pick a certain synth or effects and some settings to use for comparison. But it's hardware you are testing, or a DAW for that matter. Bot the synth/plugin itself. The OP was wanting some kind of benchmark for comparing synth cpu usage and this is just not possible, unless it was something extremely simple using only things that all synths have in common. For example a single note with a saw wave playing through a single filter with cutoff and resonance at half. But I'm not sure this would be too useful.

Post Reply

Return to “Instruments”