diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c
index add514d53d2e..ff93f8ecba28 100644
--- a/tools/perf/builtin-record.c
+++ b/tools/perf/builtin-record.c
@@ -403,8 +403,10 @@ static void create_counter(int counter, int cpu, pid_t pid)
 	if (call_graph)
 		attr->sample_type	|= PERF_SAMPLE_CALLCHAIN;
 
-	if (raw_samples)
+	if (raw_samples) {
 		attr->sample_type	|= PERF_SAMPLE_RAW;
+		attr->sample_type	|= PERF_SAMPLE_CPU;
+	}
 
 	attr->mmap		= track;
 	attr->comm		= track;
diff --git a/tools/perf/builtin-trace.c b/tools/perf/builtin-trace.c
index 8247fd04745a..bbe4c444ef8f 100644
--- a/tools/perf/builtin-trace.c
+++ b/tools/perf/builtin-trace.c
@@ -58,12 +58,19 @@ process_sample_event(event_t *event, unsigned long offset, unsigned long head)
 	struct dso *dso = NULL;
 	struct thread *thread;
 	u64 ip = event->ip.ip;
+	u32 cpu = -1;
 	u64 period = 1;
 	void *more_data = event->ip.__more_data;
 	int cpumode;
 
 	thread = threads__findnew(event->ip.pid, &threads, &last_match);
 
+	if (sample_type & PERF_SAMPLE_CPU) {
+		cpu = *(u32 *)more_data;
+		more_data += sizeof(u32);
+		more_data += sizeof(u32); /* reserved */
+	}
+
 	if (sample_type & PERF_SAMPLE_PERIOD) {
 		period = *(u64 *)more_data;
 		more_data += sizeof(u64);
@@ -120,7 +127,7 @@ process_sample_event(event_t *event, unsigned long offset, unsigned long head)
 		 * field, although it should be the same than this perf
 		 * event pid
 		 */
-		print_event(0, raw->data, raw->size, 0, thread->comm);
+		print_event(cpu, raw->data, raw->size, 0, thread->comm);
 	}
 	total += period;