diff --git a/gfx/filter.c b/gfx/filter.c
index 1298b2cbb8..55e81f5e96 100644
--- a/gfx/filter.c
+++ b/gfx/filter.c
@@ -156,7 +156,7 @@ rarch_softfilter_t *rarch_softfilter_new(const char *filter_path,
    filt->max_height = max_height;
 
    filt->impl_data = filt->impl->create(input_fmt, input_fmt, max_width, max_height,
-         threads != RARCH_SOFTFILTER_THREADS_AUTO ? threads : 1, cpu_features);
+         threads != RARCH_SOFTFILTER_THREADS_AUTO ? threads : rarch_get_cpu_cores(), cpu_features);
    if (!filt->impl_data)
    {
       RARCH_ERR("Failed to create softfilter state.\n");
@@ -170,6 +170,8 @@ rarch_softfilter_t *rarch_softfilter_new(const char *filter_path,
       goto error;
    }
 
+   RARCH_LOG("Using %u threads for softfilter.\n", threads);
+
    filt->packets = (struct softfilter_work_packet*)calloc(threads, sizeof(*filt->packets));
    if (!filt->packets)
    {
diff --git a/performance.c b/performance.c
index b5054d5919..53b87dd574 100644
--- a/performance.c
+++ b/performance.c
@@ -71,6 +71,10 @@
 #include <emscripten.h>
 #endif
 
+#if defined(BSD) || defined(__APPLE__)
+#include <sys/sysctl.h>
+#endif
+
 #include <string.h>
 
 #define MAX_COUNTERS 64
@@ -289,6 +293,40 @@ static void arm_enable_runfast_mode(void)
 }
 #endif
 
+unsigned rarch_get_cpu_cores(void)
+{
+#if defined(_WIN32) && !defined(_XBOX) // Win32
+   SYSTEM_INFO sysinfo;
+   GetSystemInfo(&sysinfo);
+   return sysinfo.dwNumberOfProcessors;
+#elif defined(_SC_NPROCESSORS_ONLN) // Linux, most unix-likes.
+   long ret = sysconf(_SC_NPROCESSORS_ONLN);
+   if (ret <= 0)
+      return 1;
+   return ret;
+#elif defined(BSD) || defined(__APPLE__) // BSD
+   // Copypasta from stackoverflow, dunno if it works.
+   int num_cpu = 0;
+   int mib[4];
+   size_t len = sizeof(num_cpu);
+
+   mib[0] = CTL_HW;
+   mib[1] = HW_AVAILCPU;
+   sysctl(mib, 2, &num_cpu, &len, NULL, 0);
+   if (num_cpu < 1)
+   {
+      mib[1] = HW_NCPU;
+      sysctl(mib, 2, &num_cpu, &len, NULL, 0);
+      if (num_cpu < 1)
+         num_cpu = 1;
+   }
+   return num_cpu;
+#else
+   // No idea, assume single core.
+   return 1;
+#endif
+}
+
 uint64_t rarch_get_cpu_features(void)
 {
    uint64_t cpu = 0;
diff --git a/performance.h b/performance.h
index d4db324964..fb6dbe1c2c 100644
--- a/performance.h
+++ b/performance.h
@@ -55,6 +55,7 @@ static inline void rarch_perf_stop(struct retro_perf_counter *perf)
 }
 
 uint64_t rarch_get_cpu_features(void);
+unsigned rarch_get_cpu_cores(void);
 
 // Used internally by RetroArch.
 #if defined(PERF_TEST) || !defined(RARCH_INTERNAL)