diff --git a/include/nvtop/extract_gpuinfo_common.h b/include/nvtop/extract_gpuinfo_common.h
index 9e4d1c9..e2d070f 100644
--- a/include/nvtop/extract_gpuinfo_common.h
+++ b/include/nvtop/extract_gpuinfo_common.h
@@ -107,6 +107,8 @@ enum gpuinfo_dynamic_info_valid {
   gpuinfo_power_draw_max_valid,
   gpuinfo_effective_load_rate_valid,
   gpuinfo_multi_instance_mode_valid,
+  gpuinfo_ecc_corrected_valid,
+  gpuinfo_ecc_uncorrected_valid,
   gpuinfo_dynamic_info_count,
 };
 
@@ -123,6 +125,8 @@ struct gpuinfo_dynamic_info {
   unsigned long long total_memory;  // Total memory (bytes)
   unsigned long long free_memory;   // Unallocated memory (bytes)
   unsigned long long used_memory;   // Allocated memory (bytes)
+  unsigned long long ecc_corrected; // Total correctable ECC errors
+  unsigned long long ecc_uncorrected; // Total uncorrected ECC errors
   unsigned int pcie_link_gen;       // PCIe link generation used
   unsigned int pcie_link_width;     // PCIe line width used
   unsigned int pcie_rx;             // PCIe throughput in KB/s
diff --git a/include/nvtop/interface_internal_common.h b/include/nvtop/interface_internal_common.h
index aec93d3..ad29eab 100644
--- a/include/nvtop/interface_internal_common.h
+++ b/include/nvtop/interface_internal_common.h
@@ -67,6 +67,7 @@ struct device_window {
   WINDOW *fan_speed;
   WINDOW *temperature;
   WINDOW *power_info;
+  WINDOW *ecc_info;
   WINDOW *gpu_clock_info;
   WINDOW *mem_clock_info;
   WINDOW *pcie_info;
@@ -148,6 +149,7 @@ enum device_field {
   device_fan_speed,
   device_temperature,
   device_power,
+  device_ecc,
   device_pcie,
   device_clock,
   device_mem_clock,
diff --git a/src/extract_gpuinfo_nvidia.c b/src/extract_gpuinfo_nvidia.c
index 33670a6..0417ab6 100644
--- a/src/extract_gpuinfo_nvidia.c
+++ b/src/extract_gpuinfo_nvidia.c
@@ -138,6 +138,16 @@ typedef enum {
   NVML_TEMPERATURE_GPU = 0,
 } nvmlTemperatureSensors_t;
 
+typedef enum {
+  NVML_MEMORY_ERROR_TYPE_CORRECTED = 0, 
+  NVML_MEMORY_ERROR_TYPE_UNCORRECTED = 1
+} nvmlMemoryErrorType_t;
+
+typedef enum {
+  NVML_VOLATILE_ECC = 0,
+  NVML_AGGREGATE_ECC = 1
+} nvmlEccCounterType_t;
+
 static nvmlReturn_t (*nvmlDeviceGetTemperature)(nvmlDevice_t device, nvmlTemperatureSensors_t sensorType,
                                                 unsigned int *temp);
 
@@ -145,6 +155,8 @@ static nvmlReturn_t (*nvmlDeviceGetPowerUsage)(nvmlDevice_t device, unsigned int
 
 static nvmlReturn_t (*nvmlDeviceGetEnforcedPowerLimit)(nvmlDevice_t device, unsigned int *limit);
 
+static nvmlReturn_t (*nvmlDeviceGetTotalEccErrors)(nvmlDevice_t device, nvmlMemoryErrorType_t errorType, nvmlEccCounterType_t counterType, unsigned long long *eccCounts);
+
 static nvmlReturn_t (*nvmlDeviceGetEncoderUtilization)(nvmlDevice_t device, unsigned int *utilization,
                                                        unsigned int *samplingPeriodUs);
 
@@ -465,6 +477,7 @@ static bool gpuinfo_nvidia_init(void) {
       (nvmlReturn_t(*)(nvmlDevice_t, unsigned int *, void *))nvmlDeviceGetMPSComputeRunningProcesses_v2;
   nvmlDeviceGetMPSComputeRunningProcesses[3] =
       (nvmlReturn_t(*)(nvmlDevice_t, unsigned int *, void *))nvmlDeviceGetMPSComputeRunningProcesses_v3;
+  nvmlDeviceGetTotalEccErrors = dlsym(libnvidia_ml_handle, "nvmlDeviceGetTotalEccErrors");
 
   // These ones might not be available
   nvmlDeviceGetProcessUtilization = dlsym(libnvidia_ml_handle, "nvmlDeviceGetProcessUtilization");
@@ -741,6 +754,15 @@ static void gpuinfo_nvidia_refresh_dynamic_info(struct gpu_info *_gpu_info) {
   if (last_nvml_return_status == NVML_SUCCESS)
     SET_VALID(gpuinfo_power_draw_max_valid, dynamic_info->valid);
 
+  if (nvmlDeviceGetTotalEccErrors) {
+    unsigned long long ecc_count;
+    last_nvml_return_status = nvmlDeviceGetTotalEccErrors(device, NVML_MEMORY_ERROR_TYPE_CORRECTED, NVML_VOLATILE_ECC, &ecc_count);
+    if (last_nvml_return_status == NVML_SUCCESS)
+      SET_GPUINFO_DYNAMIC(dynamic_info, ecc_corrected, ecc_count);
+    last_nvml_return_status = nvmlDeviceGetTotalEccErrors(device, NVML_MEMORY_ERROR_TYPE_UNCORRECTED, NVML_VOLATILE_ECC, &ecc_count);
+    if (last_nvml_return_status == NVML_SUCCESS)
+      SET_GPUINFO_DYNAMIC(dynamic_info, ecc_uncorrected, ecc_count);
+  }
   // MIG mode
   if (nvmlDeviceGetMigMode) {
     unsigned currentMode, pendingMode;
diff --git a/src/interface.c b/src/interface.c
index ae23199..8159a2d 100644
--- a/src/interface.c
+++ b/src/interface.c
@@ -44,8 +44,8 @@
 
 static unsigned int sizeof_device_field[device_field_count] = {
     [device_name] = 11,       [device_fan_speed] = 11,   [device_temperature] = 10, [device_power] = 15,
-    [device_clock] = 11,      [device_mem_clock] = 12,   [device_pcie] = 46,        [device_shadercores] = 7,
-    [device_l2features] = 11, [device_execengines] = 11,
+    [device_ecc] = 16,        [device_clock] = 11,       [device_mem_clock] = 12,   [device_pcie] = 46,
+    [device_shadercores] = 7, [device_l2features] = 11,  [device_execengines] = 11,
 };
 
 static unsigned int sizeof_process_field[process_field_count] = {
@@ -94,6 +94,13 @@ static void alloc_device_window(unsigned int start_row, unsigned int start_col,
                  sizeof_device_field[device_temperature] + sizeof_device_field[device_fan_speed]);
   if (dwin->power_info == NULL)
     goto alloc_error;
+  dwin->ecc_info =
+      newwin(1, sizeof_device_field[device_ecc], start_row + 1,
+             start_col + spacer * 5 + sizeof_device_field[device_clock] + sizeof_device_field[device_mem_clock] +
+                 sizeof_device_field[device_temperature] + sizeof_device_field[device_fan_speed] +
+                 sizeof_device_field[device_power]);
+  if (dwin->ecc_info == NULL)
+    goto alloc_error;
 
   // Line 3 = GPU used | MEM used | Encoder | Decoder
 
@@ -199,6 +206,7 @@ static void free_device_windows(struct device_window *dwin) {
   delwin(dwin->gpu_clock_info);
   delwin(dwin->mem_clock_info);
   delwin(dwin->power_info);
+  delwin(dwin->ecc_info);
   delwin(dwin->temperature);
   delwin(dwin->fan_speed);
   delwin(dwin->pcie_info);
@@ -350,7 +358,7 @@ static unsigned device_length(void) {
   return max(sizeof_device_field[device_name] + sizeof_device_field[device_pcie] + 1,
              sizeof_device_field[device_clock] + sizeof_device_field[device_mem_clock] +
                  sizeof_device_field[device_temperature] + sizeof_device_field[device_fan_speed] +
-                 sizeof_device_field[device_power] + 5);
+                 sizeof_device_field[device_power] + sizeof_device_field[device_ecc] + 6);
 }
 
 static pid_t nvtop_pid;
@@ -830,6 +838,22 @@ static void draw_devices(struct list_head *devices, struct nvtop_interface *inte
     mvwchgat(dev->power_info, 0, 0, 3, 0, cyan_color, NULL);
     wnoutrefresh(dev->power_info);
 
+    // ECC errors (professional/datacenter cards only; hidden otherwise)
+    werase(dev->ecc_info);
+    if (GPUINFO_DYNAMIC_FIELD_VALID(&device->dynamic_info, ecc_corrected) ||
+        GPUINFO_DYNAMIC_FIELD_VALID(&device->dynamic_info, ecc_uncorrected)) {
+      unsigned long long corrected =
+          GPUINFO_DYNAMIC_FIELD_VALID(&device->dynamic_info, ecc_corrected) ? device->dynamic_info.ecc_corrected : 0;
+      unsigned long long uncorrected =
+          GPUINFO_DYNAMIC_FIELD_VALID(&device->dynamic_info, ecc_uncorrected) ? device->dynamic_info.ecc_uncorrected : 0;
+      mvwprintw(dev->ecc_info, 0, 0, "  ECC %llu/%llu", corrected, uncorrected);
+      mvwchgat(dev->ecc_info, 0, 2, 3, 0, cyan_color, NULL);
+      // Highlight a non-zero uncorrected count in red: it signals a hardware fault
+      if (uncorrected > 0)
+        mvwchgat(dev->ecc_info, 0, 6, -1, 0, red_color, NULL);
+    }
+    wnoutrefresh(dev->ecc_info);
+
     // PICe throughput
     werase(dev->pcie_info);
     if (device->static_info.integrated_graphics) {