diff --git a/configure.in b/configure.in
index f8b4c23be9..4a296e8d75 100644
--- a/configure.in
+++ b/configure.in
@@ -635,7 +635,7 @@ AC_CHECK_HEADER(pcap.h,,[AC_ERROR(pcap.h not found ...)])
     ])
 
 # enable debug validation functions & macro's output
-    AC_ARG_ENABLE(debug,
+    AC_ARG_ENABLE(debug-validation,
            AS_HELP_STRING([--enable-debug-validation], [Enable (debug) validation code output]),,[enable_debug_validation=no])
     AS_IF([test "x$enable_debug_validation" = "xyes"], [
         CFLAGS="${CFLAGS} -DDEBUG_VALIDATION"
@@ -685,7 +685,7 @@ AC_CHECK_HEADER(pcap.h,,[AC_ERROR(pcap.h not found ...)])
 
 # enable CUDA output
     AC_ARG_ENABLE(cuda,
-           AS_HELP_STRING([--enable-cuda], [Enable CUDA pattern matching]),,[enable_cuda=no])
+           AS_HELP_STRING([--enable-cuda], [Enable experimental CUDA pattern matching]),,[enable_cuda=no])
     AS_IF([test "x$enable_cuda" = "xyes"], [
         AC_ARG_WITH(cuda_includes,
                 [  --with-cuda-includes=DIR  cuda include directory],
diff --git a/src/cuda-packet-batcher.c b/src/cuda-packet-batcher.c
index 274b8bddb9..73bc8551da 100644
--- a/src/cuda-packet-batcher.c
+++ b/src/cuda-packet-batcher.c
@@ -518,6 +518,13 @@ TmEcode SCCudaPBBatchPackets(ThreadVars *tv, Packet *p, void *data, PacketQueue
      * we end up buffering the packet or not */
     p->cuda_mpm_enabled = 0;
 
+    /* packets that are too big are handled by the cpu */
+    if (p->payload_len > SC_CUDA_PB_MAX_PAYLOAD_SIZE) {
+        SCLogDebug("p->payload_len %"PRIu16" > %d, inspecting on the CPU.",
+            p->payload_len, SC_CUDA_PB_MAX_PAYLOAD_SIZE);
+        return TM_ECODE_OK;
+    }
+
     SCCudaPBThreadCtx *tctx = data;
     /* the packets buffer */
     SCCudaPBPacketsBuffer *pb = (SCCudaPBPacketsBuffer *)tctx->curr_pb;
diff --git a/src/cuda-packet-batcher.h b/src/cuda-packet-batcher.h
index eb0579f0fe..c320ffea23 100644
--- a/src/cuda-packet-batcher.h
+++ b/src/cuda-packet-batcher.h
@@ -20,6 +20,9 @@
  * be enabling cuda.  We will only end up screwing performance */
 #define SC_CUDA_PB_MIN_NO_OF_PACKETS 4000
 
+/* the maximum payload size we're sending to the card (defined in decode.h) */
+#define SC_CUDA_PB_MAX_PAYLOAD_SIZE CUDA_MAX_PAYLOAD_SIZE
+
 /**
  * \brief Implement the template SCDQGenericQData to transfer the cuda
  *        packet buffer from the cuda batcher thread to the dispatcher
@@ -87,7 +90,7 @@ typedef struct SCCudaPBPacketDataForGPU_ {
     unsigned int payload_len;
     /* holds the payload.  While we actually store the payload in the buffer,
      * we may not end up using the entire 1480 bytes if the payload is smaller */
-    uint8_t payload[1480];
+    uint8_t payload[SC_CUDA_PB_MAX_PAYLOAD_SIZE];
 } SCCudaPBPacketDataForGPU;
 
 /**
diff --git a/src/decode.h b/src/decode.h
index fe3954cfe3..a71ccab329 100644
--- a/src/decode.h
+++ b/src/decode.h
@@ -57,6 +57,10 @@
 
 #include "detect-reference.h"
 
+#ifdef __SC_CUDA_SUPPORT__
+#define CUDA_MAX_PAYLOAD_SIZE 1500
+#endif
+
 /* forward declaration */
 struct DetectionEngineThreadCtx_;
 
@@ -384,7 +388,7 @@ typedef struct Packet_
     SCMutex cuda_mutex;
     SCCondT cuda_cond;
     /* the extra 1 in the 1481, is to hold the no_of_matches from the mpm run */
-    uint16_t mpm_offsets[1481];
+    uint16_t mpm_offsets[CUDA_MAX_PAYLOAD_SIZE + 1];
 #endif
 } Packet;