diff --git a/discover/runner.go b/discover/runner.go
index 4c0bce75b..5e4e05f95 100644
--- a/discover/runner.go
+++ b/discover/runner.go
@@ -496,7 +496,7 @@ func bootstrapDevices(ctx context.Context, ollamaLibDirs []string, extraEnvs []s
 func GetDevicesFromRunner(ctx context.Context, runner BaseRunner) ([]ml.DeviceInfo, error) {
 	var moreDevices []ml.DeviceInfo
 	port := runner.GetPort()
-	tick := time.Tick(500 * time.Millisecond)
+	tick := time.Tick(10 * time.Millisecond)
 	for {
 		select {
 		case <-ctx.Done():
@@ -530,7 +530,7 @@ func GetDevicesFromRunner(ctx context.Context, runner BaseRunner) ([]ml.DeviceIn
 			}
 			if resp.StatusCode != 200 {
 				logutil.Trace("runner failed to discover free VRAM", "status", resp.StatusCode, "response", body)
-				continue
+				return nil, fmt.Errorf("runner error: %s", string(body))
 			}
 
 			if err := json.Unmarshal(body, &moreDevices); err != nil {
diff --git a/runner/ollamarunner/runner.go b/runner/ollamarunner/runner.go
index c86d3c2b9..a97ef7c18 100644
--- a/runner/ollamarunner/runner.go
+++ b/runner/ollamarunner/runner.go
@@ -1247,6 +1247,8 @@ func (s *Server) info(w http.ResponseWriter, r *http.Request) {
 	m := s.model
 
 	if m == nil {
+		startLoad := time.Now()
+
 		// Dummy load to get the backend wired up
 		f, err := os.CreateTemp("", "*.bin")
 		if err != nil {
@@ -1268,9 +1270,12 @@ func (s *Server) info(w http.ResponseWriter, r *http.Request) {
 			http.Error(w, fmt.Sprintf("failed to initialize baackend: %v", err), http.StatusInternalServerError)
 			return
 		}
+		slog.Debug("dummy model load took", "duration", time.Since(startLoad))
 	}
 
+	startDevices := time.Now()
 	infos := m.Backend().BackendDevices()
+	slog.Debug("gathering device infos took", "duration", time.Since(startDevices))
 	if err := json.NewEncoder(w).Encode(&infos); err != nil {
 		http.Error(w, fmt.Sprintf("failed to encode response: %v", err), http.StatusInternalServerError)
 	}