diff --git a/os/src/drivers/sd_card/omap4/bench/main.cc b/os/src/drivers/sd_card/omap4/bench/main.cc
index 3db3993d3..fafbbdec7 100644
--- a/os/src/drivers/sd_card/omap4/bench/main.cc
+++ b/os/src/drivers/sd_card/omap4/bench/main.cc
@@ -8,85 +8,55 @@
 #include <base/sleep.h>
 #include <base/printf.h>
 #include <timer_session/connection.h>
+#include <os/attached_ram_dataspace.h>
 
 /* local includes */
 #include <driver.h>
 
 
-class Jiffies_thread : Genode::Thread<4096>
-{
-	private:
-
-		Timer::Connection    _timer;
-		Genode::Lock mutable _lock;
-		Genode::size_t       _milliseconds;
-
-		enum { MS_PER_STEP = 10 };
-
-		void entry()
-		{
-			for (;;) {
-				_timer.msleep(MS_PER_STEP);
-				{
-					Genode::Lock::Guard guard(_lock);
-					_milliseconds += MS_PER_STEP;
-				}
-			}
-		}
-
-	public:
-
-		Jiffies_thread() : _milliseconds(0) { start(); }
-
-		Genode::size_t milliseconds() const
-		{
-			Genode::Lock::Guard guard(_lock);
-			return _milliseconds;
-		}
-
-		void reset()
-		{
-			Genode::Lock::Guard guard(_lock);
-			_milliseconds = 0;
-		}
-};
-
-
 /*
  * \param total_size    total number of bytes to read
  * \param request_size  number of bytes per request
  */
-static void test_read(Block::Driver &driver,
-                      Jiffies_thread &jiffies,
-                      char          *out_buffer,
-                      Genode::size_t total_size,
-                      Genode::size_t request_size)
+static void test_read(Block::Driver  &driver,
+                      Timer::Session &timer,
+                      char           *buffer_virt,
+                      Genode::addr_t  buffer_phys,
+                      Genode::size_t  buffer_size,
+                      Genode::size_t  request_size)
 {
 	using namespace Genode;
 
 	PLOG("read: request_size=%zd bytes", request_size);
 
-	jiffies.reset();
+	size_t const time_before_ms = timer.elapsed_ms();
 
-	size_t num_requests = total_size / request_size;
+	size_t num_requests = buffer_size / request_size;
 
 	for (size_t i = 0; i < num_requests; i++)
 	{
 		size_t const block_count  = request_size / driver.block_size();
 		addr_t const block_number = i*block_count;
 
-		driver.read(block_number, block_count, out_buffer + i*request_size);
+		if (driver.dma_enabled()) {
+			driver.read_dma(block_number, block_count,
+			                buffer_phys + i*request_size);
+		} else {
+			driver.read(block_number, block_count,
+			            buffer_virt + i*request_size);
+		}
 	}
 
-	size_t const duration_ms = jiffies.milliseconds();
+	size_t const time_after_ms = timer.elapsed_ms();
+	size_t const duration_ms   = time_after_ms - time_before_ms;
 
 	/*
 	 * Convert bytes per milliseconds to kilobytes per seconds
 	 *
 	 * (total_size / 1024) / (duration_ms / 1000)
 	 */
-	size_t const total_size_kb = total_size / 1024;
-	size_t const throughput_kb_per_sec = (1000*total_size_kb) / duration_ms;
+	size_t const buffer_size_kb = buffer_size / 1024;
+	size_t const throughput_kb_per_sec = (1000*buffer_size_kb) / duration_ms;
 
 	PLOG("      -> duration:   %zd ms", duration_ms);
 	PLOG("         throughput: %zd KiB/sec", throughput_kb_per_sec);
@@ -99,18 +69,27 @@ int main(int argc, char **argv)
 
 	printf("--- OMAP4 SD card benchmark ---\n");
 
-	static Block::Omap4_driver driver;
-	static Jiffies_thread jiffies;
+	bool const use_dma = true;
 
-	long const block_sizes[] = {
+	static Block::Omap4_driver driver(use_dma);
+
+	static Timer::Connection timer;
+
+	long const request_sizes[] = {
 		512, 1024, 2048, 4096, 8192, 16384, 32768, 0 };
 
-	enum { TOTAL_SIZE = 10*1024*1024 };
+	/* total size of communication buffer */
+	size_t const buffer_size = 10*1024*1024;
 
-	static char buffer[TOTAL_SIZE];
+	/* allocate read buffer */
+	static Attached_ram_dataspace buffer(env()->ram_session(), buffer_size, false);
 
-	for (unsigned i = 0; block_sizes[i]; i++)
-		test_read(driver, jiffies, buffer, TOTAL_SIZE, block_sizes[i]);
+	char * const buffer_virt = buffer.local_addr<char>();
+	addr_t const buffer_phys = Dataspace_client(buffer.cap()).phys_addr();
+
+	for (unsigned i = 0; request_sizes[i]; i++)
+		test_read(driver, timer, buffer_virt, buffer_phys, buffer_size,
+		          request_sizes[i]);
 
 	printf("--- OMAP4 SD card benchmark finished ---\n");
 	sleep_forever();