diff --git a/sound/pci/emu10k1/emu10k1_callback.c b/sound/pci/emu10k1/emu10k1_callback.c
index 5f6c47cbb809..ef26e4d3e2a3 100644
--- a/sound/pci/emu10k1/emu10k1_callback.c
+++ b/sound/pci/emu10k1/emu10k1_callback.c
@@ -255,7 +255,7 @@ lookup_voices(struct snd_emux *emu, struct snd_emu10k1 *hw,
 		/* check if sample is finished playing (non-looping only) */
 		if (bp != best + V_OFF && bp != best + V_FREE &&
 		    (vp->reg.sample_mode & SNDRV_SFNT_SAMPLE_SINGLESHOT)) {
-			val = snd_emu10k1_ptr_read(hw, CCCA_CURRADDR, vp->ch);
+			val = snd_emu10k1_ptr_read(hw, CCCA_CURRADDR, vp->ch) - 64 + 3;
 			if (val >= vp->reg.loopstart)
 				bp = best + V_OFF;
 		}
@@ -364,7 +364,7 @@ start_voice(struct snd_emux_voice *vp)
 
 	map = (hw->silent_page.addr << hw->address_mode) | (hw->address_mode ? MAP_PTI_MASK1 : MAP_PTI_MASK0);
 
-	addr = vp->reg.start;
+	addr = vp->reg.start + 64 - 3;
 	temp = vp->reg.parm.filterQ;
 	ccca = (temp << 28) | addr;
 	if (vp->apitch < 0xe400)
@@ -432,6 +432,9 @@ start_voice(struct snd_emux_voice *vp)
 		/* Q & current address (Q 4bit value, MSB) */
 		CCCA, ccca,
 
+		/* cache */
+		CCR, REG_VAL_PUT(CCR_CACHEINVALIDSIZE, 64),
+
 		/* reset volume */
 		VTFT, vtarget | vp->ftarget,
 		CVCF, vtarget | CVCF_CURRENTFILTER_MASK,
diff --git a/sound/pci/emu10k1/emu10k1_patch.c b/sound/pci/emu10k1/emu10k1_patch.c
index 481fe03fef4d..2a13fb32c1d2 100644
--- a/sound/pci/emu10k1/emu10k1_patch.c
+++ b/sound/pci/emu10k1/emu10k1_patch.c
@@ -31,6 +31,7 @@ snd_emu10k1_sample_new(struct snd_emux *rec, struct snd_sf_sample *sp,
 	int shift;
 	int offset;
 	int truesize, size, blocksize;
+	int loop_start, loop_end, loop_size, data_end, unroll;
 	struct snd_emu10k1 *emu;
 
 	emu = rec->hw;
@@ -64,12 +65,35 @@ snd_emu10k1_sample_new(struct snd_emux *rec, struct snd_sf_sample *sp,
 		}
 	}
 
+	loop_start = sp->v.loopstart;
+	loop_end = sp->v.loopend;
+	loop_size = loop_end - loop_start;
+	if (!loop_size)
+		return -EINVAL;
+	data_end = sp->v.end;
+
 	/* recalculate offset */
 	sp->v.start += BLANK_HEAD_SIZE;
 	sp->v.end += BLANK_HEAD_SIZE;
 	sp->v.loopstart += BLANK_HEAD_SIZE;
 	sp->v.loopend += BLANK_HEAD_SIZE;
 
+	// Automatic pre-filling of the cache does not work in the presence
+	// of loops (*), and we don't want to fill it manually, as that is
+	// fiddly and slow. So we unroll the loop until the loop end is
+	// beyond the cache size.
+	// (*) Strictly speaking, a single iteration is supported (that's
+	// how it works when the playback engine runs), but handling this
+	// special case is not worth it.
+	unroll = 0;
+	while (sp->v.loopend < 64) {
+		truesize += loop_size;
+		sp->v.loopstart += loop_size;
+		sp->v.loopend += loop_size;
+		sp->v.end += loop_size;
+		unroll++;
+	}
+
 	/* try to allocate a memory block */
 	blocksize = truesize << shift;
 	sp->block = snd_emu10k1_synth_alloc(emu, blocksize);
@@ -89,12 +113,26 @@ snd_emu10k1_sample_new(struct snd_emux *rec, struct snd_sf_sample *sp,
 	offset += size;
 
 	/* copy provided samples */
-	size = sp->v.size << shift;
-	if (snd_emu10k1_synth_copy_from_user(emu, sp->block, offset, data, size, xor)) {
-		snd_emu10k1_synth_free(emu, sp->block);
-		sp->block = NULL;
-		return -EFAULT;
+	if (unroll && loop_end <= data_end) {
+		size = loop_end << shift;
+		if (snd_emu10k1_synth_copy_from_user(emu, sp->block, offset, data, size, xor))
+			goto faulty;
+		offset += size;
+
+		data += loop_start << shift;
+		while (--unroll > 0) {
+			size = loop_size << shift;
+			if (snd_emu10k1_synth_copy_from_user(emu, sp->block, offset, data, size, xor))
+				goto faulty;
+			offset += size;
+		}
+
+		size = (data_end - loop_start) << shift;
+	} else {
+		size = data_end << shift;
 	}
+	if (snd_emu10k1_synth_copy_from_user(emu, sp->block, offset, data, size, xor))
+		goto faulty;
 	offset += size;
 
 	/* clear rest of samples (if any) */
@@ -102,6 +140,11 @@ snd_emu10k1_sample_new(struct snd_emux *rec, struct snd_sf_sample *sp,
 		snd_emu10k1_synth_memset(emu, sp->block, offset, blocksize - offset, fill);
 
 	return 0;
+
+faulty:
+	snd_emu10k1_synth_free(emu, sp->block);
+	sp->block = NULL;
+	return -EFAULT;
 }
 
 /*