diff --git a/src/common/fp/op/FPConvert.cpp b/src/common/fp/op/FPConvert.cpp
index 559d0242..49524906 100644
--- a/src/common/fp/op/FPConvert.cpp
+++ b/src/common/fp/op/FPConvert.cpp
@@ -59,7 +59,7 @@ FPT_TO FPConvertNaN(FPT_FROM op) {
 
 template <typename FPT_TO, typename FPT_FROM>
 FPT_TO FPConvert(FPT_FROM op, FPCR fpcr, RoundingMode rounding_mode, FPSR& fpsr) {
-    const auto [type, sign, value] = FPUnpack<FPT_FROM>(op, fpcr, fpsr);
+    const auto [type, sign, value] = FPUnpackCV<FPT_FROM>(op, fpcr, fpsr);
     const bool is_althp = Common::BitSize<FPT_TO>() == 16 && fpcr.AHP();
 
     if (type == FPType::SNaN || type == FPType::QNaN) {
diff --git a/src/common/fp/unpacked.h b/src/common/fp/unpacked.h
index 132fedd0..943ee461 100644
--- a/src/common/fp/unpacked.h
+++ b/src/common/fp/unpacked.h
@@ -54,6 +54,12 @@ constexpr FPUnpacked ToNormalized(bool sign, int exponent, u64 value) {
 template<typename FPT>
 std::tuple<FPType, bool, FPUnpacked> FPUnpack(FPT op, FPCR fpcr, FPSR& fpsr);
 
+template<typename FPT>
+std::tuple<FPType, bool, FPUnpacked> FPUnpackCV(FPT op, FPCR fpcr, FPSR& fpsr) {
+    fpcr.FZ16(false);
+    return FPUnpack(op, fpcr, fpsr);
+}
+
 template<typename FPT>
 FPT FPRoundBase(FPUnpacked op, FPCR fpcr, RoundingMode rounding, FPSR& fpsr);