diff --git a/extra/nv_gpu_driver/vbios.h b/extra/nv_gpu_driver/vbios.h
index 6721caf7e4..18dfec63d8 100644
--- a/extra/nv_gpu_driver/vbios.h
+++ b/extra/nv_gpu_driver/vbios.h
@@ -27,6 +27,20 @@
 #include "gpu/vbios/bios_types.h"
 #define FALCON_APPLICATION_INTERFACE_ENTRY_ID_DMEMMAPPER     (0x4)
 
+typedef struct
+{
+    NvU8 version;
+    NvU8 headerSize;
+    NvU8 entrySize;
+    NvU8 entryCount;
+} __attribute__((packed)) FALCON_APPLICATION_INTERFACE_HEADER_V1;
+
+typedef struct
+{
+    NvU32 id;
+    NvU32 dmemOffset;
+} __attribute__((packed)) FALCON_APPLICATION_INTERFACE_ENTRY_V1;
+
 typedef struct
 {
     NvU32 signature;
diff --git a/tinygrad/runtime/autogen/nv/nv.py b/tinygrad/runtime/autogen/nv/nv.py
index d7695c7582..992c72556b 100644
--- a/tinygrad/runtime/autogen/nv/nv.py
+++ b/tinygrad/runtime/autogen/nv/nv.py
@@ -7157,6 +7157,28 @@ BCRT30_RSA3K_SIG_SIZE = 384 # macro
 FWSECLIC_READ_VBIOS_STRUCT_FLAGS = (2) # macro
 FWSECLIC_FRTS_REGION_MEDIA_FB = (2) # macro
 FWSECLIC_FRTS_REGION_SIZE_1MB_IN_4K = (0x100) # macro
+class struct_c__SA_FALCON_APPLICATION_INTERFACE_HEADER_V1(Structure):
+    pass
+
+struct_c__SA_FALCON_APPLICATION_INTERFACE_HEADER_V1._pack_ = 1 # source:True
+struct_c__SA_FALCON_APPLICATION_INTERFACE_HEADER_V1._fields_ = [
+    ('version', ctypes.c_ubyte),
+    ('headerSize', ctypes.c_ubyte),
+    ('entrySize', ctypes.c_ubyte),
+    ('entryCount', ctypes.c_ubyte),
+]
+
+FALCON_APPLICATION_INTERFACE_HEADER_V1 = struct_c__SA_FALCON_APPLICATION_INTERFACE_HEADER_V1
+class struct_c__SA_FALCON_APPLICATION_INTERFACE_ENTRY_V1(Structure):
+    pass
+
+struct_c__SA_FALCON_APPLICATION_INTERFACE_ENTRY_V1._pack_ = 1 # source:True
+struct_c__SA_FALCON_APPLICATION_INTERFACE_ENTRY_V1._fields_ = [
+    ('id', ctypes.c_uint32),
+    ('dmemOffset', ctypes.c_uint32),
+]
+
+FALCON_APPLICATION_INTERFACE_ENTRY_V1 = struct_c__SA_FALCON_APPLICATION_INTERFACE_ENTRY_V1
 class struct_c__SA_FALCON_APPLICATION_INTERFACE_DMEM_MAPPER_V3(Structure):
     pass
 
@@ -7345,6 +7367,8 @@ __all__ = \
     'FALCON_APPLICATION_INTERFACE_DMEM_MAPPER_V3_CMD_FRTS',
     'FALCON_APPLICATION_INTERFACE_DMEM_MAPPER_V3_CMD_SB',
     'FALCON_APPLICATION_INTERFACE_ENTRY_ID_DMEMMAPPER',
+    'FALCON_APPLICATION_INTERFACE_ENTRY_V1',
+    'FALCON_APPLICATION_INTERFACE_HEADER_V1',
     'FALCON_UCODE_DESC_HEADER', 'FALCON_UCODE_DESC_HEADER_FORMAT',
     'FALCON_UCODE_DESC_V3', 'FALCON_UCODE_DESC_V3_44_FMT',
     'FALCON_UCODE_DESC_V3_SIZE_44',
@@ -8326,6 +8350,8 @@ __all__ = \
     'struct_c__SA_BIT_DATA_FALCON_DATA_V2', 'struct_c__SA_BUSINFO',
     'struct_c__SA_EcidManufacturingInfo',
     'struct_c__SA_FALCON_APPLICATION_INTERFACE_DMEM_MAPPER_V3',
+    'struct_c__SA_FALCON_APPLICATION_INTERFACE_ENTRY_V1',
+    'struct_c__SA_FALCON_APPLICATION_INTERFACE_HEADER_V1',
     'struct_c__SA_FALCON_UCODE_DESC_HEADER',
     'struct_c__SA_FALCON_UCODE_DESC_V3',
     'struct_c__SA_FALCON_UCODE_TABLE_ENTRY_V1',
diff --git a/tinygrad/runtime/support/nv/ip.py b/tinygrad/runtime/support/nv/ip.py
index 751160d2b3..b53d5a18c1 100644
--- a/tinygrad/runtime/support/nv/ip.py
+++ b/tinygrad/runtime/support/nv/ip.py
@@ -121,8 +121,13 @@ class NV_FLCN(NV_IP):
     def __patch(cmd_id, cmd):
       patched_image = bytearray(image)
 
+      hdr = nv.FALCON_APPLICATION_INTERFACE_HEADER_V1.from_buffer_copy(image[(app_hdr_off:=self.desc_v3.IMEMLoadSize+self.desc_v3.InterfaceOffset):])
+      ents = (nv.FALCON_APPLICATION_INTERFACE_ENTRY_V1 * hdr.entryCount).from_buffer_copy(image[app_hdr_off + ctypes.sizeof(hdr):])
+      for i in range(hdr.entryCount):
+        if ents[i].id == nv.FALCON_APPLICATION_INTERFACE_ENTRY_ID_DMEMMAPPER: dmem_offset = ents[i].dmemOffset
+
       # Patch image
-      dmem = nv.FALCON_APPLICATION_INTERFACE_DMEM_MAPPER_V3.from_buffer_copy(image[(dmem_mapper_offset:=self.desc_v3.IMEMLoadSize+0xae0):])
+      dmem = nv.FALCON_APPLICATION_INTERFACE_DMEM_MAPPER_V3.from_buffer_copy(image[(dmem_mapper_offset:=self.desc_v3.IMEMLoadSize+dmem_offset):])
       dmem.init_cmd = cmd_id
       patched_image[dmem_mapper_offset : dmem_mapper_offset+len(bytes(dmem))] = bytes(dmem)
       patched_image[(cmd_off:=self.desc_v3.IMEMLoadSize+dmem.cmd_in_buffer_offset) : cmd_off+len(cmd)] = cmd
@@ -164,7 +169,7 @@ class NV_FLCN(NV_IP):
     mbx = self.execute_hs(self.sec2, self.booter_image_sysmem[0], code_off=self.booter_code_off, data_off=self.booter_data_off,
       imemPa=0x0, imemVa=self.booter_code_off, imemSz=self.booter_code_sz, dmemPa=0x0, dmemVa=0x0, dmemSz=self.booter_data_sz,
       pkc_off=0x10, engid=1, ucodeid=3, mailbox=self.nvdev.gsp.wpr_meta_sysmem)
-    assert mbx[0] == 0x0, f"Booster failed to execute, mailbox is {mbx[0]:08x}, {mbx[1]:08x}"
+    assert mbx[0] == 0x0, f"Booter failed to execute, mailbox is {mbx[0]:08x}, {mbx[1]:08x}"
 
     self.nvdev.NV_PFALCON_FALCON_OS.with_base(self.falcon).write(0x0)
     assert self.nvdev.NV_PRISCV_RISCV_CPUCTL.with_base(self.falcon).read_bitfields()['active_stat'] == 1, "GSP Core is not active"