This article will review some factors that affect CPU performance with Cherry Audio instruments. If you are experiencing CPU performance issues, some of the details in this article may help you determine if there is anything you can do to improve your system's performance.
Cherry Audio works hard to make compatible products for many generations of computers. However, it's important to remember that all computers have their limits. We have access to incredibly powerful computers and expect our plugins to deliver the best sound possible. Every synth voice and effect in an instance of a plugin or standalone instrument requires some processing. It's entirely possible with Cherry Audio's Dreamsynth to have a preset that uses 16 voice polyphony, comprised of three dual-waveform oscillators per voice, and a bunch of modulation and effects on top of that. Depending on the chords being played, that can add up! A preset like that makes for a HUGE sounding synth but comes at a cost in CPU.
Things to consider that dictate the performance of a CPU:
- Thermals: CPUs get hotter the harder they work. Thermal performance can vary depending on your computer's form factor, age, and heat mitigation.
- Bus speed: Bus speed is how quickly the motherboard can transmit and receive data between the components in a system. The bus speed still constrains a high-speed CPU and lots of RAM. As an aside, the quantity of RAM makes no difference in CPU issues. The only aspect of RAM relevant in a processing context is the bus speed. An insufficient bus speed can cause bottlenecks, wasting CPU cycles. More about bus speeds can be found in this article.
- CPU generation: The older the generation of your CPU, the less likely it will handle big jobs. This doesn't mean your computer isn't fabulous; it just means it's not great at everything it used to be when it was the newest, best thing. Aside from GUI operations and audio mixing, all DSP in a Cherry Audio synth happens in one thread. Why? Because the overhead of adding multi-threaded support to DSP operations like those found in Cherry Audio synths would negate any advantage. Any decent DAW will run multiple instances of plugins across multiple cores, which will leverage today's multi-core systems. So, if you have an 8-core system, the throughput of a single core is the marker of the performance of a single instance of a synth. Generally speaking, most modern multi-core systems have no problem running our synths. Still, if you are trying to use oversampling and a high sample rate, you may be disappointed if it's an older generation CPU.
- The number of cores: Despite having just said the actual test is the ability of one core, if you have a dual-core system, you're asking a lot of your CPU. A dual-core system is likely also an older generation as well.
- Graphics processing: Does your computer share the CPU to handle graphical calculations?
- Background tasks and processes: Aside from your operating system and any other applications, like a DAW, network requests, video streaming/recording software, messaging apps, and web browsers can cause additional intermittent demands on the CPU. Windows Task Manager and macOS Activity Monitor are utilities that allow you to see some of what your computer is doing if you are curious.
- Your audio interface's sample rate and buffer sizes: Higher sampling rates offer extended high-frequency harmonic content of the audio and offer lower latencies. Higher sample rates also require more of your CPU. Buffer size is the number of samples your computer takes to process audio. In the case of a virtual synth, the time it takes from when you press a key on your controller to create the sound you hear. The lower the buffer size, the higher the demand on the CPU.
- Audio interface drivers: Although this may not directly affect the CPU performance in the same way as the above items, stable and robust device drivers with updated support affect performance.
- Oversampling: Setting oversampling in a Cherry Audio synth to something above 1x will come with the benefits of the feature but also tax your CPU. It's best practice to leave oversampling at 1x and then freeze or print at a higher oversampling rate when your part is locked in.
A preliminary checklist for CPU related issues:
-
Are the expectations of the system realistic?
- An older system running a DAW and some high-demand presets might not be able to keep up.
-
Is there something else using up system resources?
- Streaming software, screen capture, video game platforms, messaging apps, and web browsers can eat up a lot of resources and cause issues with your audio device. Optimizing a system for music production is essential, especially with an older system.
-
Is the system as optimized for music production as it can be?
- Windows and macOS both have settings for network, power management, and numerous other settings that allow you to customize and optimize the system's performance. The following articles review steps you can take to optimize a system.
-
Is the audio device set to a high sample rate?
- 44.1K is the lowest, but still great.
- 48K is an excellent middle ground.
- 96K and above will require some heavy lifting.
-
Is the audio device buffer set to a very low value, and does it need to be?
- If you need low latency to perform the synth parts in real-time, a buffer size of 128 or 256 samples is often fine; the latency introduced depends on the sample rate. Buffer sizes below 128 samples are a recipe for issues unless you have a powerful computer and a stable audio interface drivers running quality clocks.
- If you don't need low latency, set your buffer size as high as you can tolerate. For example, if you are at the mixing stage, a buffer size of 2048 samples will lighten the load on your CPU while still having a fast enough response on parameter changes to be useable.
-
Does the instrument have oversampling set to something higher than 1x?
- Oversampling is a whole other topic. The relevant takeaway in this article is that if you have CPU issues, oversampling should be kept at 1x.
Danny L
Comments