diff --git a/extra/remu/src/thread.rs b/extra/remu/src/thread.rs
index 1278351165..5f493a1bef 100644
--- a/extra/remu/src/thread.rs
+++ b/extra/remu/src/thread.rs
@@ -1127,13 +1127,14 @@ impl<'a> Thread<'a> {
                                 self.vec_reg.write64(vdst, ret)
                             }
                         }
-                        306 | 309 | 313 | 596 | 584 | 585 | 588 => {
+                        306 | 309 | 310 | 313 | 596 | 584 | 585 | 588 => {
                             let (s0, s1, s2) = (self.val(src.0), self.val(src.1), self.val(src.2));
                             let s0 = f16::from_bits(s0).negate(0, neg).absolute(0, abs);
                             let s1 = f16::from_bits(s1).negate(1, neg).absolute(1, abs);
                             let s2 = f16::from_bits(s2).negate(1, neg).absolute(1, abs);
                             let ret = match op {
                                 309 => s0 * s1,
+                                310 => f16::mul_add(s0, s1, f16::from_bits(self.vec_reg[vdst] as u16)),
                                 306 => s0 + s1,
                                 584 => f16::mul_add(s0, s1, s2),
                                 585 => f16::min(f16::min(s0, s1), s2),
diff --git a/extra/remu/test/hwtest.py b/extra/remu/test/hwtest.py
index 88fc6b41dd..52c019787a 100644
--- a/extra/remu/test/hwtest.py
+++ b/extra/remu/test/hwtest.py
@@ -1,6 +1,6 @@
 import numpy as np
 import unittest
-import subprocess
+import subprocess, struct
 from typing import cast
 from tinygrad.runtime.ops_amd import AMDProgram, AMDDevice
 from tinygrad import Tensor, dtypes, Device
@@ -83,7 +83,7 @@ amdhsa.version:
       + "\n" + code_start + code + f"\n.size {function_name}, .-{function_name}"
   return AMDProgram(cast(AMDDevice, Device["AMD"]), function_name, assemble(ret))
 
-def get_output(s:str, n_threads:int):
+def get_output(s:str, n_threads:int=1):
   assert n_threads <= 32
   code = "\n".join(["s_load_b64 s[0:1], s[0:1], null", "v_lshlrev_b32_e32 v0, 2, v0", s,
     "s_waitcnt 0",
@@ -94,6 +94,8 @@ def get_output(s:str, n_threads:int):
   prg(test._buf, global_size=(1, 1, 1), local_size=(n_threads, 1, 1), wait=True)
   return test.numpy()
 
+def f16_to_bits(x:float) -> int: return struct.unpack('<H', struct.pack('<e', x))[0]
+
 @unittest.skipUnless(Device.DEFAULT == "AMD", "tests RDNA3")
 class TestHW(unittest.TestCase):
   def setUp(self):
@@ -140,5 +142,15 @@ class TestHW(unittest.TestCase):
     """, n_threads=2)
     np.testing.assert_equal(out, 0b01)
 
+  def test_fmac_vop3_modifier(self):
+    init_state = f"""
+    v_mov_b32_e32 v10 {f16_to_bits(4.0)}
+    v_mov_b32_e32 v11 {f16_to_bits(3.0)}
+    v_mov_b32_e32 v1 {f16_to_bits(2.0)}
+    """
+    self.assertEqual(get_output(init_state+"\n"+"v_fmac_f16_e64 v1 v11 v10"), f16_to_bits(14.))
+    self.assertEqual(get_output(init_state+"\n"+"v_fmac_f16_e64 v1 -v11 v10"), f16_to_bits(-10.))
+    self.assertEqual(get_output(init_state+"\n"+"v_fmac_f16_e64 v1 -v11 -v10"), f16_to_bits(14.))
+
 if __name__ == "__main__":
   unittest.main()