diff --git a/benchmark/string_decoder/string-decoder-create.js b/benchmark/string_decoder/string-decoder-create.js
new file mode 100644
index 0000000000..17c0f6750d
--- /dev/null
+++ b/benchmark/string_decoder/string-decoder-create.js
@@ -0,0 +1,22 @@
+'use strict';
+const common = require('../common.js');
+const StringDecoder = require('string_decoder').StringDecoder;
+
+const bench = common.createBenchmark(main, {
+  encoding: [
+    'ascii', 'utf8', 'utf-8', 'base64', 'ucs2', 'UTF-8', 'AscII', 'UTF-16LE'
+  ],
+  n: [25e6]
+});
+
+function main(conf) {
+  const encoding = conf.encoding;
+  const n = conf.n | 0;
+
+  bench.start();
+  for (var i = 0; i < n; ++i) {
+    const sd = new StringDecoder(encoding);
+    !!sd.encoding;
+  }
+  bench.end(n);
+}
diff --git a/benchmark/string_decoder/string-decoder.js b/benchmark/string_decoder/string-decoder.js
index 705ef84103..7a86fb6ef5 100644
--- a/benchmark/string_decoder/string-decoder.js
+++ b/benchmark/string_decoder/string-decoder.js
@@ -1,51 +1,79 @@
 'use strict';
-var common = require('../common.js');
-var StringDecoder = require('string_decoder').StringDecoder;
+const common = require('../common.js');
+const StringDecoder = require('string_decoder').StringDecoder;
 
-var bench = common.createBenchmark(main, {
-  encoding: ['ascii', 'utf8', 'base64-utf8', 'base64-ascii'],
-  inlen: [32, 128, 1024],
+const bench = common.createBenchmark(main, {
+  encoding: ['ascii', 'utf8', 'base64-utf8', 'base64-ascii', 'utf16le'],
+  inlen: [32, 128, 1024, 4096],
   chunk: [16, 64, 256, 1024],
-  n: [25e4]
+  n: [25e5]
 });
 
-var UTF_ALPHA = 'Blåbærsyltetøy';
-var ASC_ALPHA = 'Blueberry jam';
+const UTF8_ALPHA = 'Blåbærsyltetøy';
+const ASC_ALPHA = 'Blueberry jam';
+const UTF16_BUF = Buffer.from('Blåbærsyltetøy', 'utf16le');
 
 function main(conf) {
-  var encoding = conf.encoding;
-  var inLen = conf.inlen | 0;
-  var chunkLen = conf.chunk | 0;
-  var n = conf.n | 0;
+  const encoding = conf.encoding;
+  const inLen = conf.inlen | 0;
+  const chunkLen = conf.chunk | 0;
+  const n = conf.n | 0;
 
   var alpha;
-  var chunks = [];
+  var buf;
+  const chunks = [];
   var str = '';
-  var isBase64 = (encoding === 'base64-ascii' || encoding === 'base64-utf8');
+  const isBase64 = (encoding === 'base64-ascii' || encoding === 'base64-utf8');
   var i;
 
   if (encoding === 'ascii' || encoding === 'base64-ascii')
     alpha = ASC_ALPHA;
   else if (encoding === 'utf8' || encoding === 'base64-utf8')
-    alpha = UTF_ALPHA;
-  else
+    alpha = UTF8_ALPHA;
+  else if (encoding === 'utf16le') {
+    buf = UTF16_BUF;
+    str = Buffer.alloc(0);
+  } else
     throw new Error('Bad encoding');
 
-  var sd = new StringDecoder(isBase64 ? 'base64' : encoding);
+  const sd = new StringDecoder(isBase64 ? 'base64' : encoding);
 
   for (i = 0; i < inLen; ++i) {
     if (i > 0 && (i % chunkLen) === 0 && !isBase64) {
-      chunks.push(Buffer.from(str, encoding));
-      str = '';
+      if (alpha) {
+        chunks.push(Buffer.from(str, encoding));
+        str = '';
+      } else {
+        chunks.push(str);
+        str = Buffer.alloc(0);
+      }
+    }
+    if (alpha)
+      str += alpha[i % alpha.length];
+    else {
+      var start = i;
+      var end = i + 2;
+      if (i % 2 !== 0) {
+        ++start;
+        ++end;
+      }
+      str = Buffer.concat([
+        str,
+        buf.slice(start % buf.length, end % buf.length)
+      ]);
     }
-    str += alpha[i % alpha.length];
   }
-  if (str.length > 0 && !isBase64)
+
+  if (!alpha) {
+    if (str.length > 0)
+      chunks.push(str);
+  } else if (str.length > 0 && !isBase64)
     chunks.push(Buffer.from(str, encoding));
+
   if (isBase64) {
     str = Buffer.from(str, 'utf8').toString('base64');
     while (str.length > 0) {
-      var len = Math.min(chunkLen, str.length);
+      const len = Math.min(chunkLen, str.length);
       chunks.push(Buffer.from(str.substring(0, len), 'utf8'));
       str = str.substring(len);
     }
diff --git a/lib/string_decoder.js b/lib/string_decoder.js
index 4047489e6c..aaadfd8934 100644
--- a/lib/string_decoder.js
+++ b/lib/string_decoder.js
@@ -2,224 +2,234 @@
 
 const Buffer = require('buffer').Buffer;
 
-function assertEncoding(encoding) {
-  // Do not cache `Buffer.isEncoding`, some modules monkey-patch it to support
-  // additional encodings
-  if (encoding && !Buffer.isEncoding(encoding)) {
-    throw new Error('Unknown encoding: ' + encoding);
+// Do not cache `Buffer.isEncoding` when checking encoding names as some
+// modules monkey-patch it to support additional encodings
+function normalizeEncoding(enc) {
+  if (!enc) return 'utf8';
+  var low;
+  for (;;) {
+    switch (enc) {
+      case 'utf8':
+      case 'utf-8':
+        return 'utf8';
+      case 'ucs2':
+      case 'utf16le':
+      case 'ucs-2':
+      case 'utf-16le':
+        return 'utf16le';
+      case 'base64':
+      case 'ascii':
+      case 'binary':
+      case 'hex':
+        return enc;
+      default:
+        if (low) {
+          if (!Buffer.isEncoding(enc))
+            throw new Error('Unknown encoding: ' + enc);
+          return enc;
+        }
+        low = true;
+        enc = ('' + enc).toLowerCase();
+    }
   }
 }
 
 // StringDecoder provides an interface for efficiently splitting a series of
 // buffers into a series of JS strings without breaking apart multi-byte
-// characters. CESU-8 is handled as part of the UTF-8 encoding.
-//
-// @TODO Handling all encodings inside a single object makes it very difficult
-// to reason about this code, so it should be split up in the future.
-// @TODO There should be a utf8-strict encoding that rejects invalid UTF-8 code
-// points as used by CESU-8.
-const StringDecoder = exports.StringDecoder = function(encoding) {
-  this.encoding = (encoding || 'utf8').toLowerCase().replace(/[-_]/, '');
-  assertEncoding(encoding);
+// characters.
+exports.StringDecoder = StringDecoder;
+function StringDecoder(encoding) {
+  this.encoding = normalizeEncoding(encoding);
+  var nb;
   switch (this.encoding) {
-    case 'utf8':
-      // CESU-8 represents each of Surrogate Pair by 3-bytes
-      this.surrogateSize = 3;
-      break;
-    case 'ucs2':
     case 'utf16le':
-      // UTF-16 represents each of Surrogate Pair by 2-bytes
-      this.surrogateSize = 2;
-      this.detectIncompleteChar = utf16DetectIncompleteChar;
+      this.text = utf16Text;
+      this.end = utf16End;
+      // fall through
+    case 'utf8':
+      nb = 4;
       break;
     case 'base64':
-      // Base-64 stores 3 bytes in 4 chars, and pads the remainder.
-      this.surrogateSize = 3;
-      this.detectIncompleteChar = base64DetectIncompleteChar;
+      this.text = base64Text;
+      this.end = base64End;
+      nb = 3;
       break;
     default:
-      this.write = passThroughWrite;
+      this.write = simpleWrite;
+      this.end = simpleEnd;
       return;
   }
+  this.lastNeed = 0;
+  this.lastTotal = 0;
+  this.lastChar = Buffer.allocUnsafe(nb);
+}
 
-  // Enough space to store all bytes of a single character. UTF-8 needs 4
-  // bytes, but CESU-8 may require up to 6 (3 bytes per surrogate).
-  this.charBuffer = Buffer.allocUnsafe(6);
-  // Number of bytes received for the current incomplete multi-byte character.
-  this.charReceived = 0;
-  // Number of bytes expected for the current incomplete multi-byte character.
-  this.charLength = 0;
-};
-
-
-// write decodes the given buffer and returns it as JS string that is
-// guaranteed to not contain any partial multi-byte characters. Any partial
-// character found at the end of the buffer is buffered up, and will be
-// returned when calling write again with the remaining bytes.
-//
-// Note: Converting a Buffer containing an orphan surrogate to a String
-// currently works, but converting a String to a Buffer (via `Buffer.from()`,
-// or Buffer#write) will replace incomplete surrogates with the unicode
-// replacement character. See https://codereview.chromium.org/121173009/ .
-StringDecoder.prototype.write = function(buffer) {
-  var charStr = '';
-  var buflen = buffer.length;
-  var charBuffer = this.charBuffer;
-  var charLength = this.charLength;
-  var charReceived = this.charReceived;
-  var surrogateSize = this.surrogateSize;
-  var encoding = this.encoding;
-  var charCode;
-  // if our last write ended with an incomplete multibyte character
-  while (charLength) {
-    // determine how many remaining bytes this buffer has to offer for this char
-    var diff = charLength - charReceived;
-    var available = (buflen >= diff) ? diff : buflen;
-
-    // add the new bytes to the char buffer
-    buffer.copy(charBuffer, charReceived, 0, available);
-    charReceived += available;
-
-    if (charReceived < charLength) {
-      // still not enough chars in this buffer? wait for more ...
-
-      this.charLength = charLength;
-      this.charReceived = charReceived;
-
+StringDecoder.prototype.write = function(buf) {
+  if (buf.length === 0)
+    return '';
+  var r;
+  var i;
+  if (this.lastNeed) {
+    r = this.fillLast(buf);
+    if (r === undefined)
       return '';
-    }
-
-    // remove bytes belonging to the current character from the buffer
-    buffer = buffer.slice(available, buflen);
-    buflen = buffer.length;
-
-    // get the character that was split
-    charStr = charBuffer.toString(encoding, 0, charLength);
+    i = this.lastNeed;
+    this.lastNeed = 0;
+  } else {
+    i = 0;
+  }
+  if (i < buf.length)
+    return (r ? r + this.text(buf, i) : this.text(buf, i));
+  return r || '';
+};
 
-    // CESU-8: lead surrogate (D800-DBFF) is also the incomplete character
-    charCode = charStr.charCodeAt(charStr.length - 1);
-    if (charCode >= 0xD800 && charCode <= 0xDBFF) {
-      charLength += surrogateSize;
-      charStr = '';
-      continue;
-    }
-    charReceived = charLength = 0;
+StringDecoder.prototype.end = utf8End;
 
-    // if there are no more bytes in this buffer, just emit our char
-    if (buflen === 0) {
-      this.charLength = charLength;
-      this.charReceived = charReceived;
+// Returns only complete characters in a Buffer
+StringDecoder.prototype.text = utf8Text;
 
-      return charStr;
-    }
+// Attempts to complete a partial character using bytes from a Buffer
+StringDecoder.prototype.fillLast = function(buf) {
+  if (this.lastNeed <= buf.length) {
+    buf.copy(this.lastChar, this.lastTotal - this.lastNeed, 0, this.lastNeed);
+    return this.lastChar.toString(this.encoding, 0, this.lastTotal);
   }
+  buf.copy(this.lastChar, this.lastTotal - this.lastNeed, 0, buf.length);
+  this.lastNeed -= buf.length;
+};
 
-  // determine and set charLength / charReceived
-  if (this.detectIncompleteChar(buffer))
-    charLength = this.charLength;
-  charReceived = this.charReceived;
+// Checks the type of a UTF-8 byte, whether it's ASCII, a leading byte, or a
+// continuation byte.
+function utf8CheckByte(byte) {
+  if (byte <= 0x7F)
+    return 0;
+  else if (byte >> 5 === 0x06)
+    return 2;
+  else if (byte >> 4 === 0x0E)
+    return 3;
+  else if (byte >> 3 === 0x1E)
+    return 4;
+  return -1;
+}
 
-  var end = buflen;
-  if (charLength) {
-    // buffer the incomplete character bytes we got
-    buffer.copy(charBuffer, 0, buflen - charReceived, end);
-    end -= charReceived;
+// Checks at most the last 3 bytes of a Buffer for an incomplete UTF-8
+// character, returning the total number of bytes needed to complete the partial
+// character (if applicable).
+function utf8CheckIncomplete(self, buf, i) {
+  var j = buf.length - 1;
+  if (j < i)
+    return 0;
+  var nb = utf8CheckByte(buf[j--]);
+  if (nb >= 0) {
+    if (nb > 0)
+      self.lastNeed = nb + 1 - (buf.length - j);
+    return nb;
   }
-
-  this.charLength = charLength;
-  charStr += buffer.toString(encoding, 0, end);
-
-  end = charStr.length - 1;
-  charCode = charStr.charCodeAt(end);
-  // CESU-8: lead surrogate (D800-DBFF) is also the incomplete character
-  if (charCode >= 0xD800 && charCode <= 0xDBFF) {
-    charLength += surrogateSize;
-    charReceived += surrogateSize;
-    charBuffer.copy(charBuffer, surrogateSize, 0, surrogateSize);
-    buffer.copy(charBuffer, 0, 0, surrogateSize);
-
-    this.charLength = charLength;
-    this.charReceived = charReceived;
-
-    return charStr.substring(0, end);
+  if (j < i)
+    return 0;
+  nb = utf8CheckByte(buf[j--]);
+  if (nb >= 0) {
+    if (nb > 0)
+      self.lastNeed = nb + 1 - (buf.length - j);
+    return nb;
   }
+  if (j < i)
+    return 0;
+  nb = utf8CheckByte(buf[j--]);
+  if (nb >= 0) {
+    if (nb > 0)
+      self.lastNeed = nb + 1 - (buf.length - j);
+    return nb;
+  }
+  return 0;
+}
 
-  // or just emit the charStr
-  return charStr;
-};
-
-// detectIncompleteChar determines if there is an incomplete UTF-8 character at
-// the end of the given buffer. If so, it sets this.charLength to the byte
-// length that character, and sets this.charReceived to the number of bytes
-// that are available for this character.
-StringDecoder.prototype.detectIncompleteChar = function(buffer) {
-  var buflen = buffer.length;
-  // determine how many bytes we have to check at the end of this buffer
-  var i = (buflen >= 3) ? 3 : buflen;
-  var newlen = false;
-
-  // Figure out if one of the last i bytes of our buffer announces an
-  // incomplete char.
-  for (; i > 0; i--) {
-    var c = buffer[buflen - i];
-
-    // See http://en.wikipedia.org/wiki/UTF-8#Description
-
-    // 110XXXXX
-    if (i === 1 && c >> 5 === 0x06) {
-      this.charLength = 2;
-      newlen = true;
-      break;
-    }
+// Returns all complete UTF-8 characters in a Buffer. If the Buffer ended on a
+// partial character, the character's bytes are buffered until the required
+// number of bytes are available.
+function utf8Text(buf, i) {
+  const total = utf8CheckIncomplete(this, buf, i);
+  if (!this.lastNeed)
+    return buf.toString('utf8', i);
+  this.lastTotal = total;
+  const end = buf.length - (total - this.lastNeed);
+  buf.copy(this.lastChar, 0, end);
+  return buf.toString('utf8', i, end);
+}
 
-    // 1110XXXX
-    if (i <= 2 && c >> 4 === 0x0E) {
-      this.charLength = 3;
-      newlen = true;
-      break;
-    }
+// For UTF-8, a replacement character for each buffered byte of a (partial)
+// character needs to be added to the output.
+function utf8End(buf) {
+  const r = (buf && buf.length ? this.write(buf) : '');
+  if (this.lastNeed)
+    return r + '\ufffd'.repeat(this.lastTotal - this.lastNeed);
+  return r;
+}
 
-    // 11110XXX
-    if (i <= 3 && c >> 3 === 0x1E) {
-      this.charLength = 4;
-      newlen = true;
-      break;
+// UTF-16LE typically needs two bytes per character, but even if we have an even
+// number of bytes available, we need to check if we end on a leading/high
+// surrogate. In that case, we need to wait for the next two bytes in order to
+// decode the last character properly.
+function utf16Text(buf, i) {
+  if ((buf.length - i) % 2 === 0) {
+    const r = buf.toString('utf16le', i);
+    if (r) {
+      const c = r.charCodeAt(r.length - 1);
+      if (c >= 0xD800 && c <= 0xDBFF) {
+        this.lastNeed = 2;
+        this.lastTotal = 4;
+        this.lastChar[0] = buf[buf.length - 2];
+        this.lastChar[1] = buf[buf.length - 1];
+        return r.slice(0, -1);
+      }
     }
+    return r;
   }
+  this.lastNeed = 1;
+  this.lastTotal = 2;
+  this.lastChar[0] = buf[buf.length - 1];
+  return buf.toString('utf16le', i, buf.length - 1);
+}
 
-  this.charReceived = i;
-
-  return newlen;
-};
+// For UTF-16LE we do not explicitly append special replacement characters if we
+// end on a partial character, we simply let v8 handle that.
+function utf16End(buf) {
+  const r = (buf && buf.length ? this.write(buf) : '');
+  if (this.lastNeed) {
+    const end = this.lastTotal - this.lastNeed;
+    return r + this.lastChar.toString('utf16le', 0, end);
+  }
+  return r;
+}
 
-StringDecoder.prototype.end = function(buffer) {
-  var res = '';
-  if (buffer && buffer.length)
-    res = this.write(buffer);
-
-  var charReceived = this.charReceived;
-  if (charReceived) {
-    var cr = charReceived;
-    var buf = this.charBuffer;
-    var enc = this.encoding;
-    res += buf.toString(enc, 0, cr);
+function base64Text(buf, i) {
+  const n = (buf.length - i) % 3;
+  if (n === 0)
+    return buf.toString('base64', i);
+  this.lastNeed = 3 - n;
+  this.lastTotal = 3;
+  if (n === 1) {
+    this.lastChar[0] = buf[buf.length - 1];
+  } else {
+    this.lastChar[0] = buf[buf.length - 2];
+    this.lastChar[1] = buf[buf.length - 1];
   }
+  return buf.toString('base64', i, buf.length - n);
+}
 
-  return res;
-};
 
-function passThroughWrite(buffer) {
-  return buffer.toString(this.encoding);
+function base64End(buf) {
+  const r = (buf && buf.length ? this.write(buf) : '');
+  if (this.lastNeed)
+    return r + this.lastChar.toString('base64', 0, 3 - this.lastNeed);
+  return r;
 }
 
-function utf16DetectIncompleteChar(buffer) {
-  var charReceived = this.charReceived = buffer.length % 2;
-  this.charLength = charReceived ? 2 : 0;
-  return true;
+// Pass bytes on through for single-byte encodings (e.g. ascii, binary, hex)
+function simpleWrite(buf) {
+  return buf.toString(this.encoding);
 }
 
-function base64DetectIncompleteChar(buffer) {
-  var charReceived = this.charReceived = buffer.length % 3;
-  this.charLength = charReceived ? 3 : 0;
-  return true;
+function simpleEnd(buf) {
+  return (buf && buf.length ? this.write(buf) : '');
 }
diff --git a/test/parallel/test-string-decoder-end.js b/test/parallel/test-string-decoder-end.js
index e13ddcb167..33b47a034a 100644
--- a/test/parallel/test-string-decoder-end.js
+++ b/test/parallel/test-string-decoder-end.js
@@ -51,6 +51,8 @@ function testBuf(encoding, buf) {
   var res3 = buf.toString(encoding);
 
   console.log('expect=%j', res3);
+  console.log('res1=%j', res1);
+  console.log('res2=%j', res2);
   assert.equal(res1, res3, 'one byte at a time should match toString');
   assert.equal(res2, res3, 'all bytes at once should match toString');
 }
diff --git a/test/parallel/test-string-decoder.js b/test/parallel/test-string-decoder.js
index 4f7cbd7fad..14933c46fc 100644
--- a/test/parallel/test-string-decoder.js
+++ b/test/parallel/test-string-decoder.js
@@ -1,8 +1,13 @@
 'use strict';
 require('../common');
 var assert = require('assert');
+var inspect = require('util').inspect;
 var StringDecoder = require('string_decoder').StringDecoder;
 
+// Test default encoding
+var decoder = new StringDecoder();
+assert.strictEqual(decoder.encoding, 'utf8');
+
 process.stdout.write('scanning ');
 
 // UTF-8
@@ -27,10 +32,49 @@ test(
 test('ucs2', Buffer.from('ababc', 'ucs2'), 'ababc');
 
 // UTF-16LE
-test('ucs2', Buffer.from('3DD84DDC', 'hex'), '\ud83d\udc4d'); // thumbs up
+test('utf16le', Buffer.from('3DD84DDC', 'hex'), '\ud83d\udc4d'); // thumbs up
 
 console.log(' crayon!');
 
+// Additional UTF-8 tests
+decoder = new StringDecoder('utf8');
+assert.strictEqual(decoder.write(Buffer.from('E1', 'hex')), '');
+assert.strictEqual(decoder.end(), '\ufffd');
+
+decoder = new StringDecoder('utf8');
+assert.strictEqual(decoder.write(Buffer.from('E18B', 'hex')), '');
+assert.strictEqual(decoder.end(), '\ufffd\ufffd');
+
+decoder = new StringDecoder('utf8');
+assert.strictEqual(decoder.write(Buffer.from('\ufffd')), '\ufffd');
+assert.strictEqual(decoder.end(), '');
+
+decoder = new StringDecoder('utf8');
+assert.strictEqual(decoder.write(Buffer.from('\ufffd\ufffd\ufffd')),
+                   '\ufffd\ufffd\ufffd');
+assert.strictEqual(decoder.end(), '');
+
+decoder = new StringDecoder('utf8');
+assert.strictEqual(decoder.write(Buffer.from('efbfbde2', 'hex')), '\ufffd');
+assert.strictEqual(decoder.end(), '\ufffd');
+
+
+// Additional UTF-16LE surrogate pair tests
+decoder = new StringDecoder('utf16le');
+assert.strictEqual(decoder.write(Buffer.from('3DD8', 'hex')), '');
+assert.strictEqual(decoder.write(Buffer.from('4D', 'hex')), '');
+assert.strictEqual(decoder.write(Buffer.from('DC', 'hex')), '\ud83d\udc4d');
+assert.strictEqual(decoder.end(), '');
+
+decoder = new StringDecoder('utf16le');
+assert.strictEqual(decoder.write(Buffer.from('3DD8', 'hex')), '');
+assert.strictEqual(decoder.end(), '\ud83d');
+
+decoder = new StringDecoder('utf16le');
+assert.strictEqual(decoder.write(Buffer.from('3DD8', 'hex')), '');
+assert.strictEqual(decoder.write(Buffer.from('4D', 'hex')), '');
+assert.strictEqual(decoder.end(), '\ud83d');
+
 // test verifies that StringDecoder will correctly decode the given input
 // buffer with the given encoding to the expected output. It will attempt all
 // possible ways to write() the input buffer, see writeSequences(). The
@@ -54,9 +98,9 @@ function test(encoding, input, expected, singleSequence) {
       var message =
         'Expected "' + unicodeEscape(expected) + '", ' +
         'but got "' + unicodeEscape(output) + '"\n' +
+        'input: ' + input.toString('hex').match(/.{2}/g) + '\n' +
         'Write sequence: ' + JSON.stringify(sequence) + '\n' +
-        'Decoder charBuffer: 0x' + decoder.charBuffer.toString('hex') + '\n' +
-        'Full Decoder State: ' + JSON.stringify(decoder, null, 2);
+        'Full Decoder State: ' + inspect(decoder);
       assert.fail(output, expected, message);
     }
   });