diff --git a/api.cpp b/api.cpp
index b27c07a..a3d72ca 100644
--- a/api.cpp
+++ b/api.cpp
@@ -112,13 +112,11 @@ extern uint32_t rejected_count;
 extern int device_map[8];
 extern char *device_name[8];
 
-#define gpu_threads opt_n_threads
-
 /***************************************************************/
 
 static void gpustatus(int thr_id)
 {
-	if (thr_id >= 0 && thr_id < gpu_threads) {
+	if (thr_id >= 0 && thr_id < opt_n_threads) {
 		struct cgpu_info *cgpu = &thr_info[thr_id].gpu;
 		int gpuid = cgpu->gpu_id;
 		char buf[512]; *buf = '\0';
@@ -156,9 +154,9 @@ static void gpustatus(int thr_id)
 
 		card = device_name[gpuid];
 
-		snprintf(buf, sizeof(buf), "GPU=%d;CARD=%s;TEMP=%.1f;FAN=%d;"
+		snprintf(buf, sizeof(buf), "THR=%d;GPU=%d;CARD=%s;TEMP=%.1f;FAN=%d;"
 			"FREQ=%d;PST=%s;KHS=%.2f;HWF=%d;I=%d|",
-			thr_id, card, cgpu->gpu_temp, cgpu->gpu_fan,
+			thr_id, gpuid, card, cgpu->gpu_temp, cgpu->gpu_fan,
 			cgpu->gpu_clock, pstate, cgpu->khashes,
 			cgpu->hw_errors, cgpu->intensity);
 
@@ -186,7 +184,7 @@ static char *getsummary(char *params)
 		"ALGO=%s;GPUS=%d;KHS=%.2f;ACC=%d;REJ=%d;"
 		"ACCMN=%.3f;DIFF=%.6f;UPTIME=%.0f;TS=%u|",
 		PACKAGE_NAME, PACKAGE_VERSION, APIVERSION,
-		algo, gpu_threads, (double)global_hashrate / 1000.0,
+		algo, opt_n_threads, (double)global_hashrate / 1000.0,
 		accepted_count, rejected_count,
 		accps, global_diff, uptime, (uint32_t) ts);
 	return buffer;
@@ -198,7 +196,7 @@ static char *getsummary(char *params)
 static char *getthreads(char *params)
 {
 	*buffer = '\0';
-	for (int i = 0; i < gpu_threads; i++)
+	for (int i = 0; i < opt_n_threads; i++)
 		gpustatus(i);
 	return buffer;
 }
@@ -211,10 +209,9 @@ static char *gethistory(char *params)
 {
 	struct stats_data data[20];
 	int thrid = params ? atoi(params) : -1;
-	int gpuid = params ? device_map[(thrid & 0x7)] : -1;
 	char *p = buffer;
 	*buffer = '\0';
-	int records = stats_get_history(gpuid, data, ARRAY_SIZE(data));
+	int records = stats_get_history(thrid, data, ARRAY_SIZE(data));
 	for (int i = 0; i < records; i++) {
 		time_t ts = data[i].tm_stat;
 		p += sprintf(p, "THR=%d|GPU=%d;KHS=%.2f;DIFF=%.6f;"
diff --git a/api/index.php b/api/index.php
index d25c247..da1d76c 100644
--- a/api/index.php
+++ b/api/index.php
@@ -26,7 +26,10 @@ function getdataFromPears()
 
 function ignoreField($key)
 {
-	$ignored = array('API','VER','GPU','CARD','GPUS','CPU','TS');
+	$ignored = array(
+		'API','VER','THR','GPU',
+		'CARD','GPUS','CPU','TS',
+	);
 	return in_array($key, $ignored);
 }
 
diff --git a/nvml.cpp b/nvml.cpp
index cb0af29..e182bee 100644
--- a/nvml.cpp
+++ b/nvml.cpp
@@ -346,7 +346,7 @@ int wrap_nvml_destroy(wrap_nvml_handle *nvmlh)
 #ifdef WIN32
 #include "nvapi/nvapi_ccminer.h"
 
-static int nvapi_dev_map[NVAPI_MAX_PHYSICAL_GPUS] = { 0 };
+static int nvapi_dev_map[8] = { 0 };
 static NvDisplayHandle hDisplay_a[NVAPI_MAX_PHYSICAL_GPUS * 2] = { 0 };
 static NvPhysicalGpuHandle phys[NVAPI_MAX_PHYSICAL_GPUS] = { 0 };
 static NvU32 nvapi_dev_cnt = 0;
@@ -469,24 +469,24 @@ int wrap_nvapi_init()
 		cudaDeviceProp props;
 		if (cudaGetDeviceProperties(&props, g) == cudaSuccess)
 			device_bus_ids[g] = props.pciBusID;
+		nvapi_dev_map[g] = g;// default mapping
 	}
 
 	for (NvU8 i = 0; i < nvapi_dev_cnt; i++) {
 		NvAPI_ShortString name;
-		nvapi_dev_map[i] = i; // default mapping
 		ret = NvAPI_GPU_GetFullName(phys[i], name);
 		if (ret == NVAPI_OK) {
 			for (int g = 0; g < num_processors; g++) {
 				NvU32 busId;
 				ret = NvAPI_GPU_GetBusId(phys[i], &busId);
 				if (ret == NVAPI_OK && busId == device_bus_ids[g]) {
-					nvapi_dev_map[i] = g;
+					nvapi_dev_map[g] = i;
+					if (opt_debug)
+						applog(LOG_DEBUG, "CUDA GPU[%d] matches NVAPI GPU[%d]",
+							g, i);
 					break;
 				}
 			}
-			if (opt_debug)
-				applog(LOG_DEBUG, "CUDA GPU[%d] matches NVAPI GPU[%d]",
-					nvapi_dev_map[i], i);
 		} else {
 			NvAPI_ShortString string;
 			NvAPI_GetErrorMessage(ret, string);
@@ -518,7 +518,7 @@ int gpu_fanpercent(struct cgpu_info *gpu)
 #ifdef WIN32
 	else {
 		unsigned int rpm = 0;
-		nvapi_fanspeed(nvapi_dev_map[gpu->thr_id], &rpm);
+		nvapi_fanspeed(nvapi_dev_map[gpu->gpu_id], &rpm);
 		pct = (rpm * 100) / fan_speed_max;
 		if (pct > 100) {
 			pct = 100;
@@ -539,7 +539,7 @@ float gpu_temp(struct cgpu_info *gpu)
 	}
 #ifdef WIN32
 	else {
-		nvapi_temperature(nvapi_dev_map[gpu->thr_id], &tmp);
+		nvapi_temperature(nvapi_dev_map[gpu->gpu_id], &tmp);
 		tc = (float)tmp;
 	}
 #endif
@@ -555,7 +555,7 @@ int gpu_clock(struct cgpu_info *gpu)
 	}
 #ifdef WIN32
 	if (support == -1) {
-		nvapi_getclock(nvapi_dev_map[gpu->thr_id], &freq);
+		nvapi_getclock(nvapi_dev_map[gpu->gpu_id], &freq);
 	}
 #endif
 	return (int) freq;
@@ -571,7 +571,7 @@ int gpu_pstate(struct cgpu_info *gpu)
 #ifdef WIN32
 	if (support == -1) {
 		unsigned int pst = 0;
-		nvapi_getpstate(nvapi_dev_map[gpu->thr_id], &pst);
+		nvapi_getpstate(nvapi_dev_map[gpu->gpu_id], &pst);
 		pstate = (int) pst;
 	}
 #endif