diff --git a/src/backend/Core.cpp b/src/backend/Core.cpp
index 38a62de..f553bd2 100644
--- a/src/backend/Core.cpp
+++ b/src/backend/Core.cpp
@@ -65,11 +65,8 @@ void Core::LoadROM(const std::string &rom_) {
 }
 
 u32 Core::StepCPU() {
-    if (cpuType == Interpreted) {
-        auto taken = interpreter.Step() + regs.PopStalledCycles();
-        StepRSP(taken);
-        return taken;
-    }
+    if (cpuType == Interpreted)
+        return interpreter.Step() + regs.PopStalledCycles();
 
     if (cpuType == CachedInterpreter)
         return interpreter.ExecuteCached() + regs.PopStalledCycles();
@@ -85,12 +82,6 @@ u32 Core::StepCPU() {
 void Core::StepRSP(const u32 cpuCycles) {
     MMIO &mmio = mem->mmio;
 
-    if (mmio.rsp.spStatus.halt) {
-        regs.steps = 0;
-        mmio.rsp.steps = 0;
-        return;
-    }
-
     static constexpr u32 cpuRatio = 3, rspRatio = 2;
 
     regs.steps += cpuCycles;
@@ -119,8 +110,6 @@ void Core::Run(const float volumeL, const float volumeR) {
             }
 
             for (int cycles = 0; cycles < mem->mmio.vi.cyclesPerHalfline;) {
-                Scheduler::GetInstance().HandleEvents();
-
                 const u32 taken = StepCPU();
                 cycles += taken;
                 frameCycles += taken;
diff --git a/src/backend/Scheduler.cpp b/src/backend/Scheduler.cpp
index bb3db01..8356a98 100644
--- a/src/backend/Scheduler.cpp
+++ b/src/backend/Scheduler.cpp
@@ -27,13 +27,9 @@ u64 Scheduler::Remove(const EventType eventType) const {
     return ret;
 }
 
-void Scheduler::SkipToNext() {
-    ticks = events.top().time;
-}
+void Scheduler::SkipToNext() { ticks = events.top().time; }
 
-void Scheduler::Tick(const u64 t) {
-    ticks += t;
-}
+void Scheduler::Tick(const u64 t) { ticks += t; }
 
 void Scheduler::HandleEvents() {
     n64::Mem &mem = n64::Core::GetMem();
@@ -43,6 +39,9 @@ void Scheduler::HandleEvents() {
 
     while (ticks >= events.top().time) {
         switch (const auto type = events.top().type) {
+        case RSP_STEP:
+            n64::Core::GetInstance().StepRSP(1);
+            break;
         case SI_DMA:
             si.DMA();
             break;
diff --git a/src/backend/Scheduler.hpp b/src/backend/Scheduler.hpp
index 51e43d5..0808a57 100644
--- a/src/backend/Scheduler.hpp
+++ b/src/backend/Scheduler.hpp
@@ -3,7 +3,7 @@
 #include <log.hpp>
 #include <queue>
 
-enum EventType { NONE, PI_BUS_WRITE_COMPLETE, PI_DMA_COMPLETE, SI_DMA, IMPOSSIBLE };
+enum EventType { NONE, RSP_STEP, PI_BUS_WRITE_COMPLETE, PI_DMA_COMPLETE, SI_DMA, IMPOSSIBLE };
 
 struct Event {
     u64 time;
diff --git a/src/backend/core/Interpreter.cpp b/src/backend/core/Interpreter.cpp
index 0e5573c..e248a1c 100644
--- a/src/backend/core/Interpreter.cpp
+++ b/src/backend/core/Interpreter.cpp
@@ -52,18 +52,6 @@ bool Interpreter::MaybeAdvance() {
         return false;
     }
 
-    regs.push_to_stack_trace(regs.pc);
-    if ((u32)regs.pc == 0x4) {
-        auto &regs = Core::GetRegs();
-        std::sort(regs.stack_trace.begin(), regs.stack_trace.end());
-
-        std::println("Stack trace:");
-        for (int i = 0; i < regs.stack_trace.size(); i++) {
-            std::println("    [{:016X}]", regs.stack_trace[i]);
-        }
-        exit(1);
-    }
-
     regs.oldPC = regs.pc;
     regs.pc = regs.nextPC;
     regs.nextPC += 4;
@@ -105,6 +93,14 @@ u32 Interpreter::Step() {
 
     DecodeExecute(instr);
 
+    if (!mem.mmio.rsp.spStatus.halt) {
+        rspSyncCount++;
+        if (rspSyncCount >= 3)
+            Scheduler::GetInstance().EnqueueRelative(0, RSP_STEP);
+    }
+
+    Scheduler::GetInstance().HandleEvents();
+
     return 1;
 }
 
@@ -142,7 +138,7 @@ void CachedState::EvictLine(u64 addr) {
 
 u32 Interpreter::ExecuteCached() {
     auto addr = regs.pc;
-    auto block_addr = addr;
+    auto blockAddr = addr;
     auto line = cachedState.GetLine(addr);
 
     if (line) {
@@ -153,8 +149,7 @@ u32 Interpreter::ExecuteCached() {
             Instruction instr = line->code[i];
             DecodeExecute(instr);
 
-            Core::GetInstance().StepRSP(1);
-
+            Scheduler::GetInstance().HandleEvents();
             // Branch likely with false condition, it wasn't taken so don't execute the delay slot
             if (IsBranchLikely(instr) && !regs.delaySlot)
                 break;
@@ -192,7 +187,16 @@ u32 Interpreter::ExecuteCached() {
         }
     }
 
-    cachedState.InsertLine(block_addr, std::make_shared<CachedLine>(code, i, i));
+    if (!mem.mmio.rsp.spStatus.halt) {
+        for (int j = 1; j <= (int)std::floor((double)i / 3); j++) {
+            Scheduler::GetInstance().EnqueueRelative(j * 3, RSP_STEP);
+        }
+    } else {
+        regs.steps = 0;
+        mem.mmio.rsp.steps = 0;
+    }
+
+    cachedState.InsertLine(blockAddr, std::make_shared<CachedLine>(code, i, i));
 
     return ExecuteCached();
 }
diff --git a/src/backend/core/Interpreter.hpp b/src/backend/core/Interpreter.hpp
index 716a672..1f525d5 100644
--- a/src/backend/core/Interpreter.hpp
+++ b/src/backend/core/Interpreter.hpp
@@ -64,6 +64,7 @@ struct Interpreter final {
     Registers &regs;
     Mem &mem;
     u64 cop2Latch{};
+    u32 rspSyncCount = 0;
 
     bool Fetch(Instruction &, u64);
     void CacheTypeData(u8, u64, u32, u32);
diff --git a/src/backend/core/RSP.cpp b/src/backend/core/RSP.cpp
index d7fe1c6..ac0a3cc 100644
--- a/src/backend/core/RSP.cpp
+++ b/src/backend/core/RSP.cpp
@@ -85,10 +85,7 @@ auto RSP::Read(const u32 addr) -> u32 {
     default:
         {
             auto &regs = Core::GetRegs();
-            std::println("Stack trace:");
-            for (int i = 0; i < regs.stack_trace.size(); i++) {
-                std::println("    [{:016X}]", regs.stack_trace[i]);
-            }
+
             panic("Unimplemented SP register read {:08X} (cpu pc: 0x{:016X}, rsp pc: 0x{:04X}, ra: 0x{:016X})", addr,
                   (u64)regs.oldPC, pc & 0xffc, (u64)regs.gpr[31]);
         }
diff --git a/src/backend/core/registers/Registers.cpp b/src/backend/core/registers/Registers.cpp
index 7543628..e5f6a21 100644
--- a/src/backend/core/registers/Registers.cpp
+++ b/src/backend/core/registers/Registers.cpp
@@ -6,10 +6,7 @@ namespace n64 {
 #ifdef KAIZEN_JIT_ENABLED
 Registers::Registers(JIT &jit) : jit(jit) { Reset(); }
 #else
-Registers::Registers() {
-    stack_trace.resize(0x10000000);
-    Reset();
-}
+Registers::Registers() { Reset(); }
 #endif
 
 void Registers::Reset() {
@@ -17,7 +14,6 @@ void Registers::Reset() {
     lo = 0;
     delaySlot = false;
     prevDelaySlot = false;
-    std::fill(stack_trace.begin(), stack_trace.end(), 0);
     gpr.fill(0);
     regIsConstant = 1; // first bit is true indicating $zero is constant which yes it is always
 
diff --git a/src/backend/core/registers/Registers.hpp b/src/backend/core/registers/Registers.hpp
index 1c28569..f137b34 100644
--- a/src/backend/core/registers/Registers.hpp
+++ b/src/backend/core/registers/Registers.hpp
@@ -48,14 +48,6 @@ struct Registers {
     Cop0 cop0;
     Cop1 cop1;
 
-    std::vector<u64> stack_trace;
-    int stack_count = 0;
-
-    void push_to_stack_trace(s64 val) {
-        stack_trace[stack_count++] = val;
-        stack_count %= stack_trace.size();
-    }
-
     void CpuStall(u32 cycles) { extraCycles += cycles; }
 
     u32 PopStalledCycles() {