diff --git a/src/frontend/imgui/Window.hpp b/src/frontend/imgui/Window.hpp
index 3b7b79a8..0a6ac612 100644
--- a/src/frontend/imgui/Window.hpp
+++ b/src/frontend/imgui/Window.hpp
@@ -15,7 +15,7 @@ struct Window {
   [[nodiscard]] bool gotClosed(SDL_Event event);
   ImFont *uiFont, *codeFont;
   u32 windowID;
-  float volumeL = 0.001, volumeR = 0.001;
+  float volumeL = 0.05, volumeR = 0.05;
   void LoadROM(n64::Core& core, const std::string& path);
 private:
   bool lockVolume = true;
diff --git a/src/n64/core/RSP.cpp b/src/n64/core/RSP.cpp
index 6613b6eb..3f36884c 100644
--- a/src/n64/core/RSP.cpp
+++ b/src/n64/core/RSP.cpp
@@ -20,11 +20,11 @@ void RSP::Reset() {
   memset(dmem, 0, DMEM_SIZE);
   memset(imem, 0, IMEM_SIZE);
   memset(vpr, 0, 32 * sizeof(VPR));
-  memset(gpr, 0, 32);
+  memset(gpr, 0, 32 * sizeof(u32));
   memset(&vce, 0, sizeof(VPR));
-  acc = {.h={}, .m={}, .l={}};
-  vcc = {.l = {}, .h = {}};
-  vco = {.l = {}, .h = {}};
+  memset(&acc, 0, 3 * sizeof(VPR));
+  memset(&vcc, 0, 2 * sizeof(VPR));
+  memset(&vco, 0, 2 * sizeof(VPR));
   semaphore = false;
 }
 
diff --git a/src/n64/core/RSP.hpp b/src/n64/core/RSP.hpp
index 470f0f94..bd7733d2 100644
--- a/src/n64/core/RSP.hpp
+++ b/src/n64/core/RSP.hpp
@@ -409,6 +409,7 @@ struct RSP {
   void vmadm(u32 instr);
   void vmadn(u32 instr);
   void vmov(u32 instr);
+  void vmulf(u32 instr);
   void vmudl(u32 instr);
   void vmudh(u32 instr);
   void vmudm(u32 instr);
diff --git a/src/n64/core/rsp/decode.cpp b/src/n64/core/rsp/decode.cpp
index b8573d4d..c94b9200 100644
--- a/src/n64/core/rsp/decode.cpp
+++ b/src/n64/core/rsp/decode.cpp
@@ -94,17 +94,16 @@ inline void cop2(RSP& rsp, u32 instr) {
   //util::print("Cop2 {:02X}\n", mask);
   switch(mask) {
     case 0x00:
-      //util::print("Cop2 sub {:02X}\n", mask_sub);
-      switch(mask_sub) {
-        case 0x00: rsp.mfc2(instr); break;
-        case 0x02: rsp.cfc2(instr); break;
-        case 0x04: rsp.mtc2(instr); break;
-        case 0x06: rsp.ctc2(instr); break;
-        case 0x10: case 0x14:
-        case 0x1C: case 0x1E: case 0x1F:
-          util::print("Invalid COP2 SUB {:05b} at PC: {:016X}\n", mask_sub, rsp.oldPC);
-          break;
-        default: util::panic("Unhandled RSP COP2 sub ({:05b})\n", mask_sub);
+      if((instr >> 25) & 1) {
+        rsp.vmulf(instr);
+      } else {
+        switch (mask_sub) {
+          case 0x00: rsp.mfc2(instr); break;
+          case 0x02: rsp.cfc2(instr); break;
+          case 0x04: rsp.mtc2(instr); break;
+          case 0x06: rsp.ctc2(instr); break;
+          default: util::panic("Unhandled RSP COP2 sub ({:05b})\n", mask_sub);
+        }
       }
       break;
     case 0x04: rsp.vmudl(instr); break;
diff --git a/src/n64/core/rsp/instructions.cpp b/src/n64/core/rsp/instructions.cpp
index 975e8a44..ed64632a 100644
--- a/src/n64/core/rsp/instructions.cpp
+++ b/src/n64/core/rsp/instructions.cpp
@@ -402,14 +402,14 @@ void RSP::spv(u32 instr) {
   int e = E1(instr);
   u32 addr = gpr[BASE(instr)] + SignExt7bit(OFFSET(instr), 3);
 
-  int start = E1(instr);
+  int start = e;
   int end = start + 8;
 
   for(int offset = start; offset < end; offset++) {
     if((offset & 15) < 8) {
       WriteByte(addr++, vpr[VT(instr)].byte[BYTE_INDEX((offset & 7) << 1)]);
     } else {
-      WriteByte(addr++, vpr[VT(instr)].byte[ELEMENT_INDEX(offset & 7)] >> 7);
+      WriteByte(addr++, vpr[VT(instr)].element[ELEMENT_INDEX(offset & 7)] >> 7);
     }
   }
 }
@@ -581,7 +581,10 @@ void RSP::vadd(u32 instr) {
   VPR vte = GetVTE(vpr[VT(instr)], E2(instr));
 
   for(int i = 0; i < 8; i++) {
-    s32 result = vs.selement[i] + vte.selement[i] + (vco.l.element[i] != 0);
+    s16 vsE = vs.selement[i];
+    s16 vteE = vte.selement[i];
+
+    s32 result = vsE + vteE + (vco.l.element[i] != 0);
     acc.l.element[i] = result;
     vd.element[i] = clamp_signed(result);
     vco.l.element[i] = 0;
@@ -595,7 +598,7 @@ void RSP::vaddc(u32 instr) {
   VPR vte = GetVTE(vpr[VT(instr)], E2(instr));
 
   for(int i = 0; i < 8; i++) {
-    u32 result = vs.element[i] + vte.element[i];
+    s32 result = vs.element[i] + vte.element[i];
     acc.l.element[i] = result;
     vd.element[i] = result;
     vco.l.element[i] = (result >> 16) & 1 ? 0xffff : 0;
@@ -744,6 +747,23 @@ inline bool IsSignExtension(s16 hi, s16 lo) {
   return false;
 }
 
+void RSP::vmulf(u32 instr) {
+  int e = E2(instr);
+  VPR& vs = vpr[VS(instr)];
+  VPR& vd = vpr[VD(instr)];
+  VPR vte = GetVTE(vpr[VT(instr)], e);
+
+  for(int i = 0; i < 8; i++) {
+    s32 prod = vs.selement[i] * vte.selement[i];
+    s64 accum = (prod * 2) + 0x8000;
+
+    SetACC(i, accum);
+
+    s16 result = clamp_signed(accum >> 16);
+    vd.element[i] = result;
+  }
+}
+
 void RSP::vmudl(u32 instr) {
   int e = E2(instr);
   VPR& vs = vpr[VS(instr)];