diff --git a/av1/common/av1_rtcd_defs.pl b/av1/common/av1_rtcd_defs.pl
index 53bb4fb..0e33221 100644
--- a/av1/common/av1_rtcd_defs.pl
+++ b/av1/common/av1_rtcd_defs.pl
@@ -210,7 +210,7 @@
   # the transform coefficients are held in 32-bit
   # values, so the assembler code for  av1_block_error can no longer be used.
   add_proto qw/int64_t av1_block_error/, "const tran_low_t *coeff, const tran_low_t *dqcoeff, intptr_t block_size, int64_t *ssz";
-  specialize qw/av1_block_error avx2 neon/;
+  specialize qw/av1_block_error sse2 avx2 neon/;
 
   add_proto qw/void av1_quantize_fp/, "const tran_low_t *coeff_ptr, intptr_t n_coeffs, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan";
   specialize qw/av1_quantize_fp sse2 avx2 neon/;
diff --git a/av1/encoder/x86/error_sse2.asm b/av1/encoder/x86/error_sse2.asm
index 72e9e22..f4b4968 100644
--- a/av1/encoder/x86/error_sse2.asm
+++ b/av1/encoder/x86/error_sse2.asm
@@ -11,6 +11,21 @@
 
 ;
 
+; Increment %1 by sizeof() tran_low_t * %2.
+%macro INCREMENT_ELEMENTS_TRAN_LOW 2
+  lea %1, [%1 + %2 * 4]
+%endmacro
+
+; Load %2 + %3 into m%1.
+; %3 is the offset in elements, not bytes.
+; If tran_low_t is 16 bits (low bit depth configuration) then load the value
+; directly. If tran_low_t is 32 bits (high bit depth configuration) then pack
+; the values down to 16 bits.
+%macro LOAD_TRAN_LOW 3
+  mova     m%1, [%2 + (%3) * 4]
+  packssdw m%1, [%2 + (%3) * 4 + 16]
+%endmacro
+
 %define private_prefix av1
 
 %include "third_party/x86inc/x86inc.asm"
@@ -25,14 +40,14 @@
   pxor      m4, m4                 ; sse accumulator
   pxor      m6, m6                 ; ssz accumulator
   pxor      m5, m5                 ; dedicated zero register
-  lea     uqcq, [uqcq+sizeq*2]
-  lea     dqcq, [dqcq+sizeq*2]
-  neg    sizeq
 .loop:
-  mova      m2, [uqcq+sizeq*2]
-  mova      m0, [dqcq+sizeq*2]
-  mova      m3, [uqcq+sizeq*2+mmsize]
-  mova      m1, [dqcq+sizeq*2+mmsize]
+  LOAD_TRAN_LOW 2, uqcq, 0
+  LOAD_TRAN_LOW 0, dqcq, 0
+  LOAD_TRAN_LOW 3, uqcq, 8
+  LOAD_TRAN_LOW 1, dqcq, 8
+  INCREMENT_ELEMENTS_TRAN_LOW uqcq, 16
+  INCREMENT_ELEMENTS_TRAN_LOW dqcq, 16
+  sub    sizeq, 16
   psubw     m0, m2
   psubw     m1, m3
   ; individual errors are max. 15bit+sign, so squares are 30bit, and
@@ -41,25 +56,19 @@
   pmaddwd   m1, m1
   pmaddwd   m2, m2
   pmaddwd   m3, m3
+  ; the sum of 2 31bit integers will fit in a 32bit unsigned integer
+  paddd     m0, m1
+  paddd     m2, m3
   ; accumulate in 64bit
   punpckldq m7, m0, m5
   punpckhdq m0, m5
   paddq     m4, m7
-  punpckldq m7, m1, m5
-  paddq     m4, m0
-  punpckhdq m1, m5
-  paddq     m4, m7
   punpckldq m7, m2, m5
-  paddq     m4, m1
+  paddq     m4, m0
   punpckhdq m2, m5
   paddq     m6, m7
-  punpckldq m7, m3, m5
   paddq     m6, m2
-  punpckhdq m3, m5
-  paddq     m6, m7
-  paddq     m6, m3
-  add    sizeq, mmsize
-  jl .loop
+  jg .loop
 
   ; accumulate horizontally and store in return value
   movhlps   m5, m4
diff --git a/test/error_block_test.cc b/test/error_block_test.cc
index d98cc09..da2ceba 100644
--- a/test/error_block_test.cc
+++ b/test/error_block_test.cc
@@ -234,7 +234,7 @@
 
 using ::testing::make_tuple;
 
-#if (HAVE_SSE2 || HAVE_AVX)
+#if (HAVE_SSE2)
 INSTANTIATE_TEST_CASE_P(
     SSE2, ErrorBlockTest,
     ::testing::Values(make_tuple(&av1_highbd_block_error_sse2,
@@ -242,7 +242,10 @@
                       make_tuple(&av1_highbd_block_error_sse2,
                                  &av1_highbd_block_error_c, AOM_BITS_12),
                       make_tuple(&av1_highbd_block_error_sse2,
-                                 &av1_highbd_block_error_c, AOM_BITS_8)));
+                                 &av1_highbd_block_error_c, AOM_BITS_8),
+                      make_tuple(&BlockError8BitWrapper<av1_block_error_sse2>,
+                                 &BlockError8BitWrapper<av1_block_error_c>,
+                                 AOM_BITS_8)));
 #endif  // HAVE_SSE2
 
 #if (HAVE_AVX2)