Low copy performance in OpenCL (enqueueWriteBuffer & enqueueReadBuffer) - memory

Question

I was getting very low performance while copying memory between the GPU and the CPU (both ways) with enqueueWriteBuffer and enqueueReadBuffer. So I have written a test to be sure the problem was in those two functions and I still get a very low performance.
My tests are performing several copies, including 1GB copy and still the best result is around 3GB/s. In contrast, the CUDA test "bandwidthTest.exe" achieves around 12GB/s with a copy size of 30MB. I am running all the test in a laptop with a NVIDIA 1050 GTX and CUDA 10.0.
Any ideas why the performance might be so low?
This is the code I am using for testing, I am building it using Qt. So, there are some dependencies (QTime, QDebug):
#include <QCoreApplication>
#include "CL/cl.hpp"
#include <vector>
#include <QDebug>
#include <QTime>
int main(int argc, char *argv[])
{
QCoreApplication a(argc, argv);
std::vector<cl::Platform> platforms;
cl::Platform::get(&platforms);
std::vector<std::string> platformsNames;
for (auto & platform : platforms) {
platformsNames.push_back(platform.getInfo<CL_PLATFORM_NAME>());
}
std::vector<cl::Device> all_devices;
platforms[0].getDevices(CL_DEVICE_TYPE_ALL, &all_devices);
auto device = all_devices[0];
qDebug() << "Platform used: " << platformsNames[0].c_str() << ". Device used: " << device.getInfo<CL_DEVICE_NAME>().c_str();
qDebug() << "Max work group size: " << device.getInfo<CL_DEVICE_MAX_WORK_GROUP_SIZE>();
qDebug() << "Max items size: " << device.getInfo<CL_DEVICE_MAX_WORK_ITEM_SIZES>();
cl::Context context(device);
cl::CommandQueue queue(context);
// 1KB, 1MB, 10MB, 100MB, 1GB
size_t sizes[] = {1024, 1024 * 1024, 10 * 1024 * 1024, 100 * 1024 * 1024, 1024 * 1024 * 1024};
qDebug() << "Write Buffers";
for (auto size : sizes) {
cl::Buffer buffer(context, CL_MEM_READ_WRITE, size);
std::vector<unsigned char> t(size);
QTime timerFFT;
timerFFT.start();
auto iterations = 100.0f;
for (auto i = 0; i < iterations; i++) {
int err = queue.enqueueWriteBuffer(buffer, true, 0, size, t.data());
}
auto elapsed = timerFFT.elapsed() / 1000.0f;
qDebug() << "GB/s: " << size / (1024.0f * 1024.0f * 1024.0f) / elapsed * iterations;
}
qDebug() << "Read Buffers";
for (auto size : sizes) {
cl::Buffer buffer(context, CL_MEM_READ_WRITE, size);
std::vector<unsigned char> t(size);
QTime timerFFT;
timerFFT.start();
auto iterations = 100.0f;
for (auto i = 0; i < iterations; i++) {
int err = queue.enqueueReadBuffer(buffer, true, 0, size, t.data());
}
auto elapsed = timerFFT.elapsed() / 1000.0f;
qDebug() << "GB/s: " << size / (1024.0f * 1024.0f * 1024.0f) / elapsed * iterations;
}
return a.exec();
}