diff --git a/CHANGELOG.md b/CHANGELOG.md
index 66e0169d7a55..3db955c68ff2 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -2,7 +2,7 @@
 ##### Unreleased
 - Added [`ArrayBuffer.prototype.{ transfer, transferToFixedLength }`](https://github.com/tc39/proposal-arraybuffer-transfer) and support transferring of `ArrayBuffer`s via [`structuredClone`](https://html.spec.whatwg.org/multipage/structured-data.html#dom-structuredclone) to engines with `MessageChannel`
 - Optimized [`Math.f16round`](https://github.com/tc39/proposal-float16array) polyfill
-- Fixed [some conversion cases](https://github.com/petamoriken/float16/issues/1046) of `Math.f16round`
+- Fixed [some conversion cases](https://github.com/petamoriken/float16/issues/1046) of [`Math.f16round` and `DataView.prototype.{ getFloat16, setFloat16 }`](https://github.com/tc39/proposal-float16array)
 - Fully forced polyfilling of [the TC39 `Observable` proposal](https://github.com/tc39/proposal-observable) because of incompatibility with [the new WHATWG `Observable` proposal](https://github.com/WICG/observable)
 - Added an extra workaround of errors with exotic environment objects in `Symbol` polyfill, [#1289](https://github.com/zloirock/core-js/issues/1289)
 - Compat data improvements:
diff --git a/packages/core-js/internals/array-buffer.js b/packages/core-js/internals/array-buffer.js
index 0671919c9acb..366420b21353 100644
--- a/packages/core-js/internals/array-buffer.js
+++ b/packages/core-js/internals/array-buffer.js
@@ -12,6 +12,7 @@ var anInstance = require('../internals/an-instance');
 var toIntegerOrInfinity = require('../internals/to-integer-or-infinity');
 var toLength = require('../internals/to-length');
 var toIndex = require('../internals/to-index');
+var fround = require('../internals/math-fround');
 var IEEE754 = require('../internals/ieee754');
 var getPrototypeOf = require('../internals/object-get-prototype-of');
 var setPrototypeOf = require('../internals/object-set-prototype-of');
@@ -62,7 +63,7 @@ var unpackInt32 = function (buffer) {
 };
 
 var packFloat32 = function (number) {
-  return packIEEE754(number, 23, 4);
+  return packIEEE754(fround(number), 23, 4);
 };
 
 var packFloat64 = function (number) {
diff --git a/packages/core-js/internals/ieee754.js b/packages/core-js/internals/ieee754.js
index 057c4bd91bf9..36775b084f4f 100644
--- a/packages/core-js/internals/ieee754.js
+++ b/packages/core-js/internals/ieee754.js
@@ -1,8 +1,5 @@
 'use strict';
 // IEEE754 conversions based on https://github.com/feross/ieee754
-var sign = require('../internals/math-sign');
-var trunc = require('../internals/math-trunc');
-
 var $Array = Array;
 var abs = Math.abs;
 var pow = Math.pow;
@@ -10,21 +7,13 @@ var floor = Math.floor;
 var log = Math.log;
 var LN2 = Math.LN2;
 
-var roundToEven = function (number) {
-  var truncated = trunc(number);
-  var delta = abs(number - truncated);
-  if (delta > 0.5 || delta === 0.5 && truncated % 2 !== 0) {
-    return truncated + sign(number);
-  } return truncated;
-};
-
 var pack = function (number, mantissaLength, bytes) {
   var buffer = $Array(bytes);
   var exponentLength = bytes * 8 - mantissaLength - 1;
   var eMax = (1 << exponentLength) - 1;
   var eBias = eMax >> 1;
   var rt = mantissaLength === 23 ? pow(2, -24) - pow(2, -77) : 0;
-  var s = number < 0 || number === 0 && 1 / number < 0 ? 1 : 0;
+  var sign = number < 0 || number === 0 && 1 / number < 0 ? 1 : 0;
   var index = 0;
   var exponent, mantissa, c;
   number = abs(number);
@@ -53,10 +42,10 @@ var pack = function (number, mantissaLength, bytes) {
       mantissa = 0;
       exponent = eMax;
     } else if (exponent + eBias >= 1) {
-      mantissa = roundToEven((number * c - 1) * pow(2, mantissaLength));
+      mantissa = (number * c - 1) * pow(2, mantissaLength);
       exponent += eBias;
     } else {
-      mantissa = roundToEven(number * pow(2, eBias - 1) * pow(2, mantissaLength));
+      mantissa = number * pow(2, eBias - 1) * pow(2, mantissaLength);
       exponent = 0;
     }
   }
@@ -72,7 +61,7 @@ var pack = function (number, mantissaLength, bytes) {
     exponent /= 256;
     exponentLength -= 8;
   }
-  buffer[--index] |= s * 128;
+  buffer[--index] |= sign * 128;
   return buffer;
 };
 
@@ -83,10 +72,10 @@ var unpack = function (buffer, mantissaLength) {
   var eBias = eMax >> 1;
   var nBits = exponentLength - 7;
   var index = bytes - 1;
-  var s = buffer[index--];
-  var exponent = s & 127;
+  var sign = buffer[index--];
+  var exponent = sign & 127;
   var mantissa;
-  s >>= 7;
+  sign >>= 7;
   while (nBits > 0) {
     exponent = exponent * 256 + buffer[index--];
     nBits -= 8;
@@ -101,11 +90,11 @@ var unpack = function (buffer, mantissaLength) {
   if (exponent === 0) {
     exponent = 1 - eBias;
   } else if (exponent === eMax) {
-    return mantissa ? NaN : s ? -Infinity : Infinity;
+    return mantissa ? NaN : sign ? -Infinity : Infinity;
   } else {
     mantissa += pow(2, mantissaLength);
     exponent -= eBias;
-  } return (s ? -1 : 1) * mantissa * pow(2, exponent - mantissaLength);
+  } return (sign ? -1 : 1) * mantissa * pow(2, exponent - mantissaLength);
 };
 
 module.exports = {
diff --git a/packages/core-js/modules/esnext.data-view.set-float16.js b/packages/core-js/modules/esnext.data-view.set-float16.js
index 97d71d55ad13..03b82d240ccc 100644
--- a/packages/core-js/modules/esnext.data-view.set-float16.js
+++ b/packages/core-js/modules/esnext.data-view.set-float16.js
@@ -4,6 +4,7 @@ var uncurryThis = require('../internals/function-uncurry-this');
 var classof = require('../internals/classof');
 var toIndex = require('../internals/to-index');
 var packIEEE754 = require('../internals/ieee754').pack;
+var f16round = require('../internals/math-f16round');
 
 var $TypeError = TypeError;
 // eslint-disable-next-line es/no-typed-arrays -- safe
@@ -15,7 +16,7 @@ $({ target: 'DataView', proto: true }, {
   setFloat16: function setFloat16(byteOffset, value /* , littleEndian */) {
     if (classof(this) !== 'DataView') throw new $TypeError('Incorrect receiver');
     var offset = toIndex(byteOffset);
-    var bytes = packIEEE754(+value, 10, 2);
+    var bytes = packIEEE754(f16round(value), 10, 2);
     return setUint16(this, offset, bytes[1] << 8 | bytes[0], arguments.length > 2 ? arguments[2] : false);
   }
 });
diff --git a/tests/unit-global/esnext.data-view.set-float16.js b/tests/unit-global/esnext.data-view.set-float16.js
index 8f8506a2c5b0..55d20a124813 100644
--- a/tests/unit-global/esnext.data-view.set-float16.js
+++ b/tests/unit-global/esnext.data-view.set-float16.js
@@ -41,4 +41,26 @@ QUnit.test('DataView.prototype.{ getFloat16, setFloat16 }', assert => {
     assert.same(view.getUint16(0, LE), bin, `DataView.prototype.setFloat16 + DataView.prototype.getUint16, LE: ${ LE }, ${ toString(f16) } -> ${ toString(bin) }`);
     assert.same(view.getFloat16(0, LE), f16, `DataView.prototype.setFloat16 + DataView.prototype.getFloat16, LE: ${ LE }, ${ toString(f16) }`);
   }
+
+  const MAX_FLOAT16 = 65504;
+  const MIN_FLOAT16 = 2 ** -24;
+
+  const conversions = [
+    [1.337, 1.3369140625],
+    [0.499994, 0.5],
+    [7.9999999, 8],
+    [MAX_FLOAT16, MAX_FLOAT16],
+    [-MAX_FLOAT16, -MAX_FLOAT16],
+    [MIN_FLOAT16, MIN_FLOAT16],
+    [-MIN_FLOAT16, -MIN_FLOAT16],
+    [MIN_FLOAT16 / 2, 0],
+    [-MIN_FLOAT16 / 2, -0],
+    [2.980232238769531911744490042422139897126953655970282852649688720703125e-8, MIN_FLOAT16],
+    [-2.980232238769531911744490042422139897126953655970282852649688720703125e-8, -MIN_FLOAT16],
+  ];
+
+  for (const [from, to] of conversions) for (const LE of [false, true]) {
+    view.setFloat16(0, from, LE);
+    assert.same(view.getFloat16(0, LE), to, `DataView.prototype.setFloat16 + DataView.prototype.getFloat16, LE: ${ LE }, ${ toString(from) } -> ${ toString(to) }`);
+  }
 });
diff --git a/tests/unit-global/esnext.math.f16round.js b/tests/unit-global/esnext.math.f16round.js
index 230e51e085c0..cdd603bb3f15 100644
--- a/tests/unit-global/esnext.math.f16round.js
+++ b/tests/unit-global/esnext.math.f16round.js
@@ -23,17 +23,17 @@ QUnit.test('Math.f16round', assert => {
   assert.same(f16round(MAX_VALUE), Infinity);
   assert.same(f16round(-MAX_VALUE), -Infinity);
 
-  const maxFloat16 = 65504;
-  const minFloat16 = 2 ** -24;
+  const MAX_FLOAT16 = 65504;
+  const MIN_FLOAT16 = 2 ** -24;
 
-  assert.same(f16round(maxFloat16), maxFloat16);
-  assert.same(f16round(-maxFloat16), -maxFloat16);
-  assert.same(f16round(minFloat16), minFloat16);
-  assert.same(f16round(-minFloat16), -minFloat16);
-  assert.same(f16round(minFloat16 / 2), 0);
-  assert.same(f16round(-minFloat16 / 2), -0);
-  assert.same(f16round(2.980232238769531911744490042422139897126953655970282852649688720703125e-8), minFloat16);
-  assert.same(f16round(-2.980232238769531911744490042422139897126953655970282852649688720703125e-8), -minFloat16);
+  assert.same(f16round(MAX_FLOAT16), MAX_FLOAT16);
+  assert.same(f16round(-MAX_FLOAT16), -MAX_FLOAT16);
+  assert.same(f16round(MIN_FLOAT16), MIN_FLOAT16);
+  assert.same(f16round(-MIN_FLOAT16), -MIN_FLOAT16);
+  assert.same(f16round(MIN_FLOAT16 / 2), 0);
+  assert.same(f16round(-MIN_FLOAT16 / 2), -0);
+  assert.same(f16round(2.980232238769531911744490042422139897126953655970282852649688720703125e-8), MIN_FLOAT16);
+  assert.same(f16round(-2.980232238769531911744490042422139897126953655970282852649688720703125e-8), -MIN_FLOAT16);
 
   assert.same(f16round(1.337), 1.3369140625);
   assert.same(f16round(0.499994), 0.5);
diff --git a/tests/unit-pure/esnext.math.f16round.js b/tests/unit-pure/esnext.math.f16round.js
index f7ba20f8a73a..8c51a0f1f183 100644
--- a/tests/unit-pure/esnext.math.f16round.js
+++ b/tests/unit-pure/esnext.math.f16round.js
@@ -22,17 +22,17 @@ QUnit.test('Math.f16round', assert => {
   assert.same(f16round(MAX_VALUE), Infinity);
   assert.same(f16round(-MAX_VALUE), -Infinity);
 
-  const maxFloat16 = 65504;
-  const minFloat16 = 2 ** -24;
+  const MAX_FLOAT16 = 65504;
+  const MIN_FLOAT16 = 2 ** -24;
 
-  assert.same(f16round(maxFloat16), maxFloat16);
-  assert.same(f16round(-maxFloat16), -maxFloat16);
-  assert.same(f16round(minFloat16), minFloat16);
-  assert.same(f16round(-minFloat16), -minFloat16);
-  assert.same(f16round(minFloat16 / 2), 0);
-  assert.same(f16round(-minFloat16 / 2), -0);
-  assert.same(f16round(2.980232238769531911744490042422139897126953655970282852649688720703125e-8), minFloat16);
-  assert.same(f16round(-2.980232238769531911744490042422139897126953655970282852649688720703125e-8), -minFloat16);
+  assert.same(f16round(MAX_FLOAT16), MAX_FLOAT16);
+  assert.same(f16round(-MAX_FLOAT16), -MAX_FLOAT16);
+  assert.same(f16round(MIN_FLOAT16), MIN_FLOAT16);
+  assert.same(f16round(-MIN_FLOAT16), -MIN_FLOAT16);
+  assert.same(f16round(MIN_FLOAT16 / 2), 0);
+  assert.same(f16round(-MIN_FLOAT16 / 2), -0);
+  assert.same(f16round(2.980232238769531911744490042422139897126953655970282852649688720703125e-8), MIN_FLOAT16);
+  assert.same(f16round(-2.980232238769531911744490042422139897126953655970282852649688720703125e-8), -MIN_FLOAT16);
 
   assert.same(f16round(1.337), 1.3369140625);
   assert.same(f16round(0.499994), 0.5);