diff --git a/reframe/core/pipeline.py b/reframe/core/pipeline.py
index 5a8c617f56..06c3134d4b 100644
--- a/reframe/core/pipeline.py
+++ b/reframe/core/pipeline.py
@@ -673,6 +673,18 @@ def pipeline_hooks(cls):
     # FIXME: There is not way currently to express tuples of `float`s or
     # `None`s, so we just use the very generic `object`
 
+    #: Require that a reference is defined for each system that this test is
+    #: run on.
+    #:
+    #: If this is set and a reference is not found for the current system, the
+    #: test will fail.
+    #:
+    #: :type: boolean
+    #: :default: :const:`False`
+    #:
+    #: .. versionadded:: 4.0.0
+    require_reference = variable(typ.Bool, value=False)
+
     #:
     #: Refer to the :doc:`ReFrame Tutorials </tutorials>` for concrete usage
     #: examples.
@@ -898,7 +910,8 @@ def pipeline_hooks(cls):
     #: responsibility to check whether the build phase failed by adding an
     #: appropriate sanity check.
     #:
-    #: :type: boolean : :default: :class:`True`
+    #: :type: boolean
+    #: :default: :class:`True`
     build_locally = variable(typ.Bool, value=True, loggable=True)
 
     def __new__(cls, *args, **kwargs):
@@ -2110,6 +2123,13 @@ def check_performance(self):
                             # Pop the unit from the ref tuple (redundant)
                             ref = ref[:3]
                     except KeyError:
+                        if self.require_reference:
+                            raise PerformanceError(
+                                f'no reference value found for '
+                                f'performance variable {tag!r} on '
+                                f'system {self._current_partition.fullname!r}'
+                            ) from None
+
                         ref = (0, None, None)
 
                     self._perfvalues[key] = (value, *ref, unit)
@@ -2143,7 +2163,8 @@ def check_performance(self):
                     for var in variables:
                         name, unit = var
                         ref_tuple = (0, None, None, unit)
-                        self.reference.update({'*': {name: ref_tuple}})
+                        if not self.require_reference:
+                            self.reference.update({'*': {name: ref_tuple}})
 
                 # We first evaluate and log all performance values and then we
                 # check them against the reference. This way we always log them
@@ -2152,9 +2173,10 @@ def check_performance(self):
                     value = sn.evaluate(expr)
                     key = f'{self._current_partition.fullname}:{tag}'
                     if key not in self.reference:
-                        raise SanityError(
-                            f'tag {tag!r} not resolved in references for '
-                            f'{self._current_partition.fullname}'
+                        raise PerformanceError(
+                            f'no reference value found for '
+                            f'performance variable {tag!r} on '
+                            f'system {self._current_partition.fullname!r}'
                         )
 
                     self._perfvalues[key] = (value, *self.reference[key])
@@ -2181,7 +2203,7 @@ def check_performance(self):
                                  'expected {1} (l={2}, u={3})' % tag))
                     )
                 except SanityError as e:
-                    raise PerformanceError(e)
+                    raise PerformanceError(e) from None
 
     def _copy_job_files(self, job, dst):
         if job is None:
diff --git a/unittests/test_pipeline.py b/unittests/test_pipeline.py
index 5edbd26aef..18be91d1c1 100644
--- a/unittests/test_pipeline.py
+++ b/unittests/test_pipeline.py
@@ -1315,6 +1315,49 @@ def __init__(self):
     yield MyTest()
 
 
+@pytest.fixture
+def dummytest_modern(testsys_exec_ctx, perf_file, sanity_file):
+    '''Modern version of the dummytest above'''
+
+    class MyTest(rfm.RunOnlyRegressionTest):
+        perf_file = perf_file
+        reference = {
+            'testsys': {
+                'value1': (1.4, -0.1, 0.1, None),
+                'value2': (1.7, -0.1, 0.1, None),
+            },
+            'testsys:gpu': {
+                'value3': (3.1, -0.1, 0.1, None),
+            }
+        }
+
+        @sanity_function
+        def validate(self):
+            return sn.assert_found(r'result = success', sanity_file)
+
+        @performance_function('unit')
+        def value1(self):
+            return sn.extractsingle(r'perf1 = (\S+)', perf_file, 1, float)
+
+        @performance_function('unit')
+        def value2(self):
+            return sn.extractsingle(r'perf2 = (\S+)', perf_file, 1, float)
+
+        @performance_function('unit')
+        def value3(self):
+            return sn.extractsingle(r'perf3 = (\S+)', perf_file, 1, float)
+
+    yield MyTest()
+
+
+@pytest.fixture(params=['classic', 'modern'])
+def dummy_perftest(request, dummytest, dummytest_modern):
+    if request.param == 'modern':
+        return dummytest_modern
+    else:
+        return dummytest
+
+
 def test_sanity_success(dummytest, sanity_file, perf_file, dummy_gpu_exec_ctx):
     sanity_file.write_text('result = success\n')
     perf_file.write_text('perf1 = 1.3\n'
@@ -1390,7 +1433,7 @@ def test_reference_unknown_tag(dummytest, sanity_file,
             'foo': (3.1, -0.1, 0.1, None),
         }
     }
-    with pytest.raises(SanityError):
+    with pytest.raises(PerformanceError):
         _run_sanity(dummytest, *dummy_gpu_exec_ctx)
 
 
@@ -1456,6 +1499,27 @@ def test_reference_tag_resolution(dummytest, sanity_file,
     _run_sanity(dummytest, *dummy_gpu_exec_ctx)
 
 
+def test_required_reference(dummy_perftest, sanity_file,
+                            perf_file, dummy_gpu_exec_ctx):
+    sanity_file.write_text('result = success\n')
+    perf_file.write_text('perf1 = 1.3\n'
+                         'perf2 = 1.8\n'
+                         'perf3 = 3.3\n')
+
+    dummy_perftest.require_reference = True
+    dummy_perftest.reference = {
+        'testsys:login': {
+            'value1': (1.4, -0.1, 0.1, None),
+            'value3': (3.1, -0.1, 0.1, None),
+        },
+        'foo': {
+            'value2': (1.7, -0.1, 0.1, None)
+        }
+    }
+    with pytest.raises(PerformanceError):
+        _run_sanity(dummy_perftest, *dummy_gpu_exec_ctx)
+
+
 def test_performance_invalid_value(dummytest, sanity_file,
                                    perf_file, dummy_gpu_exec_ctx):
     sanity_file.write_text('result = success\n')