diff --git a/src/main.cpp b/src/main.cpp
index 0ab2c1f..0ba8acf 100644
--- a/src/main.cpp
+++ b/src/main.cpp
@@ -16,7 +16,7 @@ enum class TransferType {
 };
 
 struct BenchmarkConfig {
-    uint32_t buffer_size = 256ULL << 20;
+    uint32_t buffer_size = 128ULL << 20;
     uint32_t iterations = 32;
     TransferType type = TransferType::Memory;
 };
@@ -26,6 +26,19 @@ struct BenchmarkResult {
     double device_to_host_time = 0.0f;
 };
 
+struct Gpu {
+    VkPhysicalDevice physicalDevice = nullptr;
+    VkPhysicalDeviceProperties properties = {};
+    VkDevice device = nullptr;
+    VkQueue queue = nullptr;
+    VkCommandPool pool = nullptr;
+    // for benchmarks
+    VkDeviceMemory stagingMem = nullptr;
+    VkDeviceMemory deviceMem = nullptr;
+    VkBuffer stagingBuf = nullptr;
+    VkBuffer deviceBuf = nullptr;
+};
+
 // ---------- helpers ----------
 VKAPI_ATTR VkBool32 VKAPI_CALL debugCallback(
     VkDebugUtilsMessageSeverityFlagBitsEXT, VkDebugUtilsMessageTypeFlagsEXT,
@@ -124,19 +137,6 @@ double benchCopy(VkDevice dev, VkCommandPool pool, VkQueue queue, VkBuffer src,
     return std::chrono::duration<double, std::milli>(t1 - t0).count();
 }
 
-struct Gpu {
-    VkPhysicalDevice physicalDevice = nullptr;
-    VkPhysicalDeviceProperties properties = {};
-    VkDevice device = nullptr;
-    VkQueue queue = nullptr;
-    VkCommandPool pool = nullptr;
-    // for benchmarks
-    VkBuffer stagingBuf = nullptr;
-    VkBuffer deviceBuf = nullptr;
-    VkDeviceMemory stagingMem = nullptr;
-    VkDeviceMemory deviceMem = nullptr;
-};
-
 Gpu initGpu(VkPhysicalDevice phy, uint32_t buffer_size) {
     Gpu gpu = {};
     gpu.physicalDevice = phy;
@@ -202,10 +202,14 @@ void reportBenchmark(const BenchmarkResult &result, const Gpu &gpu,
     std::cout << "Device: " << gpu.properties.deviceName << std::endl;
     std::cout << "Buffer size : " << config.buffer_size / (1 << 20) << " MiB\n";
     std::cout << "Iterations  : " << config.iterations << "\n";
-    std::cout << "H→D average : " << (gib / (result.host_to_device_time * 1e-3))
-              << " GiB/s\n";
-    std::cout << "D→H average : " << (gib / (result.device_to_host_time * 1e-3))
-              << " GiB/s\n";
+    std::cout << "H->D average : "
+              << (gib / (result.host_to_device_time * 1e-3)) << " GiB/s\n";
+    std::cout << "H->D avg transfer time: " << result.host_to_device_time
+              << "ms" << std::endl;
+    std::cout << "D->H average : "
+              << (gib / (result.device_to_host_time * 1e-3)) << " GiB/s\n";
+    std::cout << "D->H avg transfer time: " << result.device_to_host_time
+              << "ms" << std::endl;
     std::cout << std::endl;
 }
 
@@ -217,7 +221,7 @@ BenchmarkResult runBenchmark(Gpu &gpu, BenchmarkConfig config) {
     std::memset(mapped, 0xAB, config.buffer_size);
     vkUnmapMemory(gpu.device, gpu.stagingMem);
 
-    // ---- warm-up ----
+    // warm-up, probably not significant
     benchCopy(gpu.device, gpu.pool, gpu.queue, gpu.stagingBuf, gpu.deviceBuf,
               config.buffer_size);