diff --git a/src/video_core/control/channel_state.cpp b/src/video_core/control/channel_state.cpp
index d07c7e2a83..c3e54b2fbf 100644
--- a/src/video_core/control/channel_state.cpp
+++ b/src/video_core/control/channel_state.cpp
@@ -16,6 +16,24 @@
 #include "video_core/memory_manager.h"
 
 namespace Tegra::Control {
+namespace {
+
+// Match NVK/Nouveau's initial pushbuffer subchannel layout.
+constexpr u32 Nvk3DSubchannel = 0;
+constexpr u32 NvkComputeSubchannel = 1;
+constexpr u32 Nvk2DSubchannel = 3;
+constexpr u32 NvkCopySubchannel = 4;
+
+void BindNvkDefaultSubchannels(ChannelState& channel_state) {
+    auto& dma_pusher = *channel_state.dma_pusher;
+    dma_pusher.BindSubchannel(&*channel_state.maxwell_3d, Nvk3DSubchannel, Engines::EngineTypes::Maxwell3D);
+    dma_pusher.BindSubchannel(&*channel_state.kepler_compute, NvkComputeSubchannel, Engines::EngineTypes::KeplerCompute);
+    // Subchannel 2 is M2MF there; Eden does not expose a 0x9039 engine yet.
+    dma_pusher.BindSubchannel(&*channel_state.fermi_2d, Nvk2DSubchannel, Engines::EngineTypes::Fermi2D);
+    dma_pusher.BindSubchannel(&*channel_state.maxwell_dma, NvkCopySubchannel, Engines::EngineTypes::MaxwellDMA);
+}
+
+} // Anonymous namespace
 
 ChannelState::ChannelState(s32 bind_id_) : bind_id{bind_id_}, initialized{} {}
 
@@ -28,6 +46,7 @@ void ChannelState::Init(Core::System& system, GPU& gpu, u64 program_id_) {
     kepler_compute.emplace(system, *memory_manager);
     maxwell_dma.emplace(system, *memory_manager);
     kepler_memory.emplace(system, *memory_manager);
+    BindNvkDefaultSubchannels(*this);
     initialized = true;
 }
 
diff --git a/src/video_core/engines/puller.cpp b/src/video_core/engines/puller.cpp
index 4030f93d49..a5a6138cbe 100644
--- a/src/video_core/engines/puller.cpp
+++ b/src/video_core/engines/puller.cpp
@@ -22,6 +22,13 @@
 
 namespace Tegra::Engines {
 
+namespace {
+
+constexpr u32 Gf100BindClassMask = 0xffff;
+constexpr u32 Gf100BindValidMask =  0x1f0000 | Gf100BindClassMask;
+
+} // Anonymous namespace
+
 Puller::Puller(GPU& gpu_, MemoryManager& memory_manager_, DmaPusher& dma_pusher_,
                Control::ChannelState& channel_state_)
     : gpu{gpu_}, memory_manager{memory_manager_}, dma_pusher{dma_pusher_}, channel_state{
@@ -30,10 +37,14 @@ Puller::Puller(GPU& gpu_, MemoryManager& memory_manager_, DmaPusher& dma_pusher_
 Puller::~Puller() = default;
 
 void Puller::ProcessBindMethod(const MethodCall& method_call) {
-    // Bind the current subchannel to the desired engine id.
-    LOG_DEBUG(HW_GPU, "Binding subchannel {} to engine {}", method_call.subchannel,
+    LOG_DEBUG(HW_GPU, "Binding subchannel {} to engine {:#x}", method_call.subchannel,
               method_call.argument);
-    const auto engine_id = static_cast<EngineID>(method_call.argument);
+    u32 engine = method_call.argument;
+    if ((engine & ~Gf100BindClassMask) != 0 && (engine & ~Gf100BindValidMask) == 0) {
+        engine &= Gf100BindClassMask;
+    }
+
+    const auto engine_id = static_cast<EngineID>(engine);
     bound_engines[method_call.subchannel] = engine_id;
     switch (engine_id) {
     case EngineID::FERMI_TWOD_A: