diff --git a/src/cuda-packet-batcher.c b/src/cuda-packet-batcher.c
index 9906c9b326..aa50a76f88 100644
--- a/src/cuda-packet-batcher.c
+++ b/src/cuda-packet-batcher.c
@@ -79,7 +79,7 @@ static int run_batcher = 1;
  * on the traffic
  * \todo make this user configurable, as well allow dynamic update of this
  * variable based on the traffic seen */
-static uint32_t buffer_packet_threshhold = 1280;
+static uint32_t buffer_packet_threshhold = 2400;
 
 /* flag used by the SIG_ALRM handler to indicate that the batcher TM should queue
  * the buffer to be processed by the Cuda Mpm B2g Batcher Thread for further
@@ -546,13 +546,27 @@ TmEcode SCCudaPBBatchPackets(ThreadVars *tv, Packet *p, void *data, PacketQueue
     /* the sgh to which the incoming packet belongs */
     SigGroupHead *sgh = NULL;
 
-    /* get the signature group head to which this packet belongs.  If it belongs
-     * to no sgh, we don't need to buffer this packet.
-     * \todo Get rid of this, once we get the sgh from the flow */
-    sgh = SCCudaPBGetSgh(tctx->de_ctx, p);
+    if (p->flow != NULL) {
+        /* Get the stored sgh from the flow (if any). Make sure we're not using
+         * the sgh for icmp error packets part of the same stream. */
+        if (p->proto == p->flow->proto) { /* filter out icmp */
+            if (p->flowflags & FLOW_PKT_TOSERVER && p->flow->flags & FLOW_SGH_TOSERVER) {
+                sgh = p->flow->sgh_toserver;
+            } else if (p->flowflags & FLOW_PKT_TOCLIENT && p->flow->flags & FLOW_SGH_TOCLIENT) {
+                sgh = p->flow->sgh_toclient;
+            }
+        }
+    }
+
     if (sgh == NULL) {
-        SCLogDebug("No SigGroupHead match for this packet");
-        return TM_ECODE_OK;
+        /* get the signature group head to which this packet belongs.  If it belongs
+         * to no sgh, we don't need to buffer this packet.
+         * \todo Get rid of this, once we get the sgh from the flow */
+        sgh = SCCudaPBGetSgh(tctx->de_ctx, p);
+        if (sgh == NULL) {
+            SCLogDebug("No SigGroupHead match for this packet");
+            return TM_ECODE_OK;
+        }
     }
 
     /* if the payload is less than the maximum content length in this sgh we
@@ -747,7 +761,7 @@ void SCCudaPBSetUpQueuesAndBuffers(void)
     /* \todo This needs to be changed ASAP.  This can't exceed max_pending_packets.
      * Also we need to make this user configurable and allow dynamic updaes
      * based on live traffic */
-    buffer_packet_threshhold = 1280;
+    buffer_packet_threshhold = 2400;
 
     return;
 }
diff --git a/src/suricata.c b/src/suricata.c
index 2b8030c4e8..059ca69ee4 100644
--- a/src/suricata.c
+++ b/src/suricata.c
@@ -355,6 +355,7 @@ int main(int argc, char **argv)
 #endif
     int dump_config = 0;
     int list_unittests = 0;
+    int list_cuda_cards = 0;
     int daemon = 0;
     char *user_name = NULL;
     char *group_name = NULL;
@@ -412,6 +413,7 @@ int main(int argc, char **argv)
         {"pcap-buffer-size", required_argument, 0, 0},
         {"unittest-filter", required_argument, 0, 'U'},
         {"list-unittests", 0, &list_unittests, 1},
+        {"list-cuda-cards", 0, &list_cuda_cards, 1},
 #ifdef OS_WIN32
 		{"service-install", 0, 0, 0},
 		{"service-remove", 0, 0, 0},
@@ -482,6 +484,12 @@ int main(int argc, char **argv)
 #else
                 fprintf(stderr, "ERROR: Unit tests not enabled. Make sure to pass --enable-unittests to configure when building.\n");
                 exit(EXIT_FAILURE);
+#endif /* UNITTESTS */
+            } else if(strcmp((long_opts[option_index]).name, "list-cuda-cards") == 0) {
+#ifndef __SC_CUDA_SUPPORT__
+                fprintf(stderr, "ERROR: Cuda not enabled. Make sure to pass "
+                        "--enable-cuda to configure when building.\n");
+                exit(EXIT_FAILURE);
 #endif /* UNITTESTS */
             }
 #ifdef OS_WIN32
@@ -690,6 +698,10 @@ int main(int argc, char **argv)
 #ifdef __SC_CUDA_SUPPORT__
     /* Init the CUDA environment */
     SCCudaInitCudaEnvironment();
+    if (list_cuda_cards) {
+        SCCudaListCards();
+        exit(EXIT_SUCCESS);
+    }
 #endif
 
     if (!CheckValidDaemonModes(daemon, run_mode)) {
diff --git a/src/util-cuda-handlers.c b/src/util-cuda-handlers.c
index e6b51eba87..ae477b67ad 100644
--- a/src/util-cuda-handlers.c
+++ b/src/util-cuda-handlers.c
@@ -65,6 +65,7 @@
 
 #include "tmqh-simple.h"
 
+#include "conf.h"
 #include "util-error.h"
 #include "util-debug.h"
 #include "util-unittest.h"
@@ -226,9 +227,23 @@ int SCCudaHlGetCudaContext(CUcontext *p_context, int handle)
         return 0;
     }
 
+    /* Get default log level and format. */
+    char *cuda_device_id_str = NULL;
+    int cuda_device_id = SC_CUDA_DEFAULT_DEVICE;
+    if (ConfGet("cuda.device_id", &cuda_device_id_str) == 1) {
+        cuda_device_id = atoi(cuda_device_id_str);
+        if (!SCCudaIsCudaDeviceIdValid(cuda_device_id)) {
+            SCLogError(SC_ERR_CUDA_ERROR, "Invalid device id \"%s\" supplied "
+                       "in the conf file", cuda_device_id_str);
+            cuda_device_id = SC_CUDA_DEFAULT_DEVICE;
+        }
+    } else {
+        cuda_device_id = SC_CUDA_DEFAULT_DEVICE;
+    }
+
     /* Get the device list for this CUDA platform and create a new cuda context */
     devices = SCCudaGetDeviceList();
-    if (SCCudaCtxCreate(p_context, 0, devices->devices[0]->device) == -1)
+    if (SCCudaCtxCreate(p_context, 0, devices->devices[cuda_device_id]->device) == -1)
         goto error;
     data->cuda_context = p_context[0];
 
diff --git a/src/util-cuda.c b/src/util-cuda.c
index bd511a7a68..5210b821cd 100644
--- a/src/util-cuda.c
+++ b/src/util-cuda.c
@@ -4069,6 +4069,56 @@ int SCCudaInitCudaEnvironment(void)
     return -1;
 }
 
+/**********************************Cuda_Utility********************************/
+
+/**
+ * \brief List the cuda cards on the system.
+ *
+ */
+void SCCudaListCards(void)
+{
+    int i = 0;
+
+    if (devices == NULL) {
+        SCLogWarning(SC_ERR_CUDA_ERROR, "CUDA engine not initalized!  Please "
+                     "initialize the cuda environment using "
+                     "SCCudaInitCudaEnvironment().");
+        return;
+    }
+
+    printf("CUDA Cards recognized by the suricata CUDA module - \n");
+    printf("|-----------------------------------------------------------------------------|\n");
+    printf("| %-10s | %-20s | %-10s | %-10s | %-13s |\n",
+           "Device Id", "    Device Name", "  Multi-", "Clock Rate", "Cuda Compute");
+    printf("| %-10s | %-20s | %-10s | %-10s | %-13s |\n",
+           "", "", "Processors", "   (MHz)", "Capability");
+    printf("|-----------------------------------------------------------------------------|\n");
+    for (i = 0; i < devices->count; i++) {
+        printf("| %-10d | %-20s | %-10d | %-10d | %d.%-11d |\n",
+               i,
+               devices->devices[i]->name,
+               devices->devices[i]->attr_multiprocessor_count,
+               devices->devices[i]->attr_clock_rate/1000,
+               devices->devices[i]->major_rev,
+               devices->devices[i]->minor_rev);
+    }
+    printf("|-----------------------------------------------------------------------------|\n");
+
+    return;
+}
+
+int SCCudaIsCudaDeviceIdValid(int cuda_device_id)
+{
+    if (devices == NULL) {
+        SCLogWarning(SC_ERR_CUDA_ERROR, "CUDA engine not initalized!  Please "
+                     "initialize the cuda environment using "
+                     "SCCudaInitCudaEnvironment().");
+        return 0;
+    }
+
+    return (cuda_device_id < devices->count);
+}
+
 /**********************************Unittests***********************************/
 
 int SCCudaTest01(void)
diff --git a/src/util-cuda.h b/src/util-cuda.h
index 304ec0a25a..8410a009d9 100644
--- a/src/util-cuda.h
+++ b/src/util-cuda.h
@@ -28,6 +28,7 @@
 
 #include <cuda.h>
 
+#define SC_CUDA_DEFAULT_DEVICE 0
 #define SC_CUDA_DEVICE_NAME_MAX_LEN 128
 
 typedef struct SCCudaDevice_ {
@@ -176,6 +177,9 @@ SCCudaDevices *SCCudaGetDeviceList(void);
 
 int SCCudaInitCudaEnvironment(void);
 
+void SCCudaListCards(void);
+int SCCudaIsCudaDeviceIdValid(int);
+
 void SCCudaRegisterTests(void);
 
 #endif /* __SC_CUDA_SUPPORT__ */
diff --git a/suricata.yaml b/suricata.yaml
index 9160381169..4e81ea9eb2 100644
--- a/suricata.yaml
+++ b/suricata.yaml
@@ -120,6 +120,12 @@ threading:
   #
   detect_thread_ratio: 1.5
 
+# Select the cuda device to use.  The device_id identifies the device to be used
+# if one has multiple devices on the system.  To find out device_id associated
+#  with the card(s) on the system run "suricata --list_cuda_cards".
+cuda:
+  device_id: 0
+
 # Select the multi pattern algorithm you want to run for scan/search the
 # in the engine. The supported algorithms are b2g, b3g and wumanber.
 #