Jump to content

Kent Y

Members
  • Posts

    2
  • Joined

  • Last visited

Reputation

0 Neutral
  1. TL;DR I didn't have the "high performance" power plan selected. LatencyMon mentioned checking power management and by default I had the balanced plan selected. So I went through the items in the balanced plan but didn't understand the implications of "minimum processor state" which was set at 5% and therefore I hadn't changed it. After researching it more I realized that the CPU cores were going into low clock frequency mode after 28 seconds (I saw that number mentioned in my BIOS overclocking section) and the CPU load of my project wasn't enough to keep the CPU cores out of low power mode. That is why things worked when I set the CPU affinity to use a single core. By using a single core, the load on the single CPU core was high enough to keep that core from entering low power mode. I found the HWinfo app to be useful for monitoring the CPU clock. When the CPU is running at a low frequency it takes some time to ramp up to max frequency and that creates latency in the audio processing. The high performance power plan sets the "minimum processor state" to 100% so the CPU cores are always running at max frequency. With just that single change I can now run the same project at 2.9ms/128 sample latency and never exceed 50% on engine load.
  2. My old PC is an AMD Ryzen 1700 with a MOTU M2 USB interface and I was able to get 5.8ms (256 sample) latency without getting late buffers in cakewalk. No special tuning was done. See below for full system spec. I built a new PC using an Intel I5-12500, using the same MOTU M2 interface. The best latency I can get without late buffers is 46.4ms (2048 sample). See below for full system spec. Since the 12500 CPU is faster, I was expecting latency to be similar if not better than the old PC. I've tried everything I could think of and done a lot of research on the internet but am still unable to answer why the latency is so much worse. The only thing that made any difference was to edit the registry for the Pro Audio multimedia entry (Computer\HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft\Windows NT\CurrentVersion\Multimedia\SystemProfile\Tasks\Pro Audio) and set the CPU affinity so that it would only use 2 CPU cores. That allowed me to bring the latency down to 256 samples without getting late buffers but I don't understand why utilizing fewer CPU cores improves the latency. I'm also concerned about what other limitations I may run into by using fewer CPU cores. A strange behaviour on the Intel PC is that after rebooting and loading my project in Cakewalk, I can run at 256 sample latency for the first 20 to 30 seconds without any problem. After that, the performance indicator shows the engine load increase significantly and the late buffers start occurring regularly and that causes popping sounds. I'm wondering if anyone has any insight into what could be going on? specs MOTU M2 USB interface, connected to USB 2 port (using other ports didn't make any difference) AMD PC Spec: AMD Ryzen 1700 CPU. Asus Prime B350-Plus MB, 16GB RAM/2133Mhz, GeForce GT 710 2GB GPU, Samsung 960 NVME Intel PC Spec: Intel I5-12500 CPU, MSI Pro B660-A D4 MB, 32GB RAM, 3200Mhz, GeForce GT 710 2GB GPU, Kingston KC3000 NVME. The integrated GPU of the 12500 is disabled in BIOS.
×
×
  • Create New...