diff --git a/reframe/core/pipeline.py b/reframe/core/pipeline.py
index 01fa2b1be5..9ff6afd034 100644
--- a/reframe/core/pipeline.py
+++ b/reframe/core/pipeline.py
@@ -1027,6 +1027,10 @@ def check_performance(self):
             return
 
         with os_ext.change_dir(self._stagedir):
+            # We first evaluate and log all performance values and then we
+            # check them against the reference. This way we always log them
+            # even if the don't meet the reference.
+            perf_values = []
             for tag, expr in self.perf_patterns.items():
                 value = evaluate(expr)
                 key = '%s:%s' % (self._current_partition.fullname, tag)
@@ -1037,9 +1041,13 @@ def check_performance(self):
                         "tag `%s' not resolved in references for `%s'" %
                         (tag, self._current_partition.fullname))
 
+                perf_values.append((value, self.reference[key]))
                 self._perf_logger.log_performance(logging.INFO, tag, value,
                                                   ref, low_thres, high_thres)
-                evaluate(assert_reference(value, ref, low_thres, high_thres))
+
+            for val, reference in perf_values:
+                refval, low_thres, high_thres = reference
+                evaluate(assert_reference(val, refval, low_thres, high_thres))
 
     def _copy_to_outputdir(self):
         """Copy checks interesting files to the output directory."""
diff --git a/unittests/test_pipeline.py b/unittests/test_pipeline.py
index 55ccfe5f2d..63aae7f073 100644
--- a/unittests/test_pipeline.py
+++ b/unittests/test_pipeline.py
@@ -605,3 +605,39 @@ def test_tag_resolution(self):
             }
         }
         self.test.check_performance()
+
+    def test_perf_var_evaluation(self):
+        # All performance values must be evaluated, despite the first one
+        # failing To test this, we need an extract function that will have a
+        # side effect when evaluated, whose result we will check after calling
+        # `check_performance()`.
+        logfile = 'perf.log'
+
+        @sn.sanity_function
+        def extract_perf(patt, tag):
+            val = sn.evaluate(
+                sn.extractsingle(patt, self.perf_file.name, tag, float))
+
+            with open('perf.log', 'a') as fp:
+                fp.write('%s=%s' % (tag, val))
+
+            return val
+
+        self.test.perf_patterns = {
+            'value1': extract_perf(r'performance1 = (?P<v1>\S+)', 'v1'),
+            'value2': extract_perf(r'performance2 = (?P<v2>\S+)', 'v2'),
+            'value3': extract_perf(r'performance3 = (?P<v3>\S+)', 'v3')
+        }
+        self.write_performance_output(performance1=1.0,
+                                      performance2=1.8,
+                                      performance3=3.3)
+        with self.assertRaises(SanityError) as cm:
+            self.test.check_performance()
+
+        logfile = os.path.join(self.test.stagedir, logfile)
+        with open(logfile) as fp:
+            log_output = fp.read()
+
+        self.assertIn('v1', log_output)
+        self.assertIn('v2', log_output)
+        self.assertIn('v3', log_output)