diff --git a/CMakeLists.txt b/CMakeLists.txt
index a3c7c783b..355b060b7 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -1,4 +1,4 @@
-cmake_minimum_required (VERSION 2.6)
+cmake_minimum_required (VERSION 3.5)
 project (xgboost)
 find_package(OpenMP)
 
diff --git a/plugin/updater_gpu/README.md b/plugin/updater_gpu/README.md
index 526f64d1b..96d425531 100644
--- a/plugin/updater_gpu/README.md
+++ b/plugin/updater_gpu/README.md
@@ -1,5 +1,16 @@
 # CUDA Accelerated Tree Construction Algorithm
 
+## Benchmarks
+
+Time for 500 boosting iterations in seconds. 
+
+Dataset | Instances | Features | i7-6700K | Titan X (pascal) | Speedup
+--- | --- | --- | --- | --- | --- 
+Yahoo LTR | 473,134 | 700 | 3738 | 507 | 7.37
+Higgs | 10,500,000 | 28 | 31352 | 4173 | 7.51
+Bosch | 1,183,747 | 968 | 9460 | 1009 | 9.38
+
+
 ## Usage
 Specify the updater parameter as 'grow_gpu'. 
 
@@ -13,9 +24,11 @@ param['updater'] = 'grow_gpu'
 ## Memory usage
 Device memory usage can be calculated as approximately:
 ```
-bytes = (10 x n_rows) + (44 x n_rows x n_columns x column_density)
+bytes = (10 x n_rows) + (40 x n_rows x n_columns x column_density) + (64 x max_nodes) + (76 x max_nodes_level x n_columns)
 ```
-Data is stored in a sparse format. For example, missing values produced by one hot encoding are not stored. If a one hot encoding separates a categorical variable into 5 columns the column_density of these columns is 1/5 = 0.2.
+The maximum number of nodes needed for a given tree depth d is 2<sup>d+1</sup> - 1. The maximum number of nodes on any given level is 2<sup>d</sup>.
+
+Data is stored in a sparse format. For example, missing values produced by one hot encoding are not stored. If a one hot encoding separates a categorical variable into 5 columns the density of these columns is 1/5 = 0.2.
 
 A 4GB graphics card will process approximately 3.5 million rows of the well known Kaggle higgs dataset.
 
@@ -26,7 +39,7 @@ A CUDA capable GPU with at least compute capability >= 3.5 (the algorithm depend
 
 Building the plug-in requires CUDA Toolkit 7.5 or later.
 
-The plugin also depends on CUB 1.5.4 - http://nvlabs.github.io/cub/index.html.
+The plugin also depends on CUB 1.5.2 - https://github.com/NVlabs/cub/tree/1.5.2
 
 CUB is a header only cuda library which provides sort/reduce/scan primitives.
 
@@ -60,5 +73,3 @@ Rory Mitchell
 Report any bugs to r.a.mitchell.nz at google mail.
 
 
-
-
diff --git a/plugin/updater_gpu/src/device_helpers.cuh b/plugin/updater_gpu/src/device_helpers.cuh
index c895c16e0..3c849efdf 100644
--- a/plugin/updater_gpu/src/device_helpers.cuh
+++ b/plugin/updater_gpu/src/device_helpers.cuh
@@ -170,55 +170,6 @@ struct Timer {
   }
 };
 
-/*
- *  Utility functions
- */
-
-template <typename T>
-void print(const thrust::device_vector<T> &v, size_t max_items = 10) {
-  thrust::host_vector<T> h = v;
-  for (int i = 0; i < std::min(max_items, h.size()); i++) {
-    std::cout << " " << h[i];
-  }
-  std::cout << "\n";
-}
-
-template <typename T>
-void print(char *label, const thrust::device_vector<T> &v,
-           const char *format = "%d ", int max = 10) {
-  thrust::host_vector<T> h_v = v;
-
-  std::cout << label << ":\n";
-  for (int i = 0; i < std::min(static_cast<int>(h_v.size()), max); i++) {
-    printf(format, h_v[i]);
-  }
-  std::cout << "\n";
-}
-
-template <typename T1, typename T2> T1 div_round_up(const T1 a, const T2 b) {
-  return static_cast<T1>(ceil(static_cast<double>(a) / b));
-}
-
-template <typename T> thrust::device_ptr<T> dptr(T *d_ptr) {
-  return thrust::device_pointer_cast(d_ptr);
-}
-
-template <typename T> T *raw(thrust::device_vector<T> &v) { //  NOLINT
-  return raw_pointer_cast(v.data());
-}
-
-template <typename T> size_t size_bytes(const thrust::device_vector<T> &v) {
-  return sizeof(T) * v.size();
-}
-
-// Threadblock iterates over range, filling with value
-template <typename IterT, typename ValueT>
-__device__ void block_fill(IterT begin, size_t n, ValueT value) {
-  for (auto i : block_stride_range(static_cast<size_t>(0), n)) {
-    begin[i] = value;
-  }
-}
-
 /*
  * Range iterator
  */
@@ -282,6 +233,55 @@ template <typename T> __device__ range block_stride_range(T begin, T end) {
   return r;
 }
 
+/*
+ *  Utility functions
+ */
+
+template <typename T>
+void print(const thrust::device_vector<T> &v, size_t max_items = 10) {
+  thrust::host_vector<T> h = v;
+  for (int i = 0; i < std::min(max_items, h.size()); i++) {
+    std::cout << " " << h[i];
+  }
+  std::cout << "\n";
+}
+
+template <typename T>
+void print(char *label, const thrust::device_vector<T> &v,
+           const char *format = "%d ", int max = 10) {
+  thrust::host_vector<T> h_v = v;
+
+  std::cout << label << ":\n";
+  for (int i = 0; i < std::min(static_cast<int>(h_v.size()), max); i++) {
+    printf(format, h_v[i]);
+  }
+  std::cout << "\n";
+}
+
+template <typename T1, typename T2> T1 div_round_up(const T1 a, const T2 b) {
+  return static_cast<T1>(ceil(static_cast<double>(a) / b));
+}
+
+template <typename T> thrust::device_ptr<T> dptr(T *d_ptr) {
+  return thrust::device_pointer_cast(d_ptr);
+}
+
+template <typename T> T *raw(thrust::device_vector<T> &v) { //  NOLINT
+  return raw_pointer_cast(v.data());
+}
+
+template <typename T> size_t size_bytes(const thrust::device_vector<T> &v) {
+  return sizeof(T) * v.size();
+}
+
+// Threadblock iterates over range, filling with value
+template <typename IterT, typename ValueT>
+__device__ void block_fill(IterT begin, size_t n, ValueT value) {
+  for (auto i : block_stride_range(static_cast<size_t>(0), n)) {
+    begin[i] = value;
+  }
+}
+
 /*
  * Memory
  */
@@ -414,6 +414,7 @@ class bulk_allocator {
     }
 
     _size = get_size_bytes(args...);
+    std::cout << "trying to allocate: " << _size << "\n";
 
     safe_cuda(cudaMalloc(&d_ptr, _size));
 
diff --git a/plugin/updater_gpu/src/gpu_builder.cu b/plugin/updater_gpu/src/gpu_builder.cu
index 4279b35e8..e198c8e40 100644
--- a/plugin/updater_gpu/src/gpu_builder.cu
+++ b/plugin/updater_gpu/src/gpu_builder.cu
@@ -68,9 +68,12 @@ struct GPUData {
 
     // Calculate memory for sort
     size_t cub_mem_size = 0;
+    cub::DoubleBuffer<NodeIdT> db_key;
+    cub::DoubleBuffer<int> db_value;
+
     cub::DeviceSegmentedRadixSort::SortPairs(
-        cub_mem.data(), cub_mem_size, cub::DoubleBuffer<NodeIdT>(),
-        cub::DoubleBuffer<int>(), in_fvalues.size(), n_features,
+        cub_mem.data(), cub_mem_size, db_key,
+        db_value, in_fvalues.size(), n_features,
         foffsets.data(), foffsets.data() + 1);
 
     // Allocate memory
@@ -304,7 +307,11 @@ void GPUBuilder::Update(const std::vector<bst_gpair> &gpair, DMatrix *p_fmat,
 
 float GPUBuilder::GetSubsamplingRate(MetaInfo info) {
   float subsample = 1.0;
-  size_t required = 10 * info.num_row + 44 * info.num_nonzero;
+  uint32_t max_nodes = (1 << (param.max_depth + 1)) - 1;
+  uint32_t max_nodes_level = 1 << param.max_depth;
+  size_t required = 10 * info.num_row + 40 * info.num_nonzero
+    + 64 * max_nodes + 76 * max_nodes_level * info.num_col;
+  std::cout << "required: " << required << "\n";
   size_t available = dh::available_memory();
   while (available < required) {
     subsample -= 0.05;
diff --git a/plugin/updater_gpu/src/gpu_builder.cuh b/plugin/updater_gpu/src/gpu_builder.cuh
index fdb308285..61ccdbbcf 100644
--- a/plugin/updater_gpu/src/gpu_builder.cuh
+++ b/plugin/updater_gpu/src/gpu_builder.cuh
@@ -36,7 +36,7 @@ class GPUBuilder {
   GPUData *gpu_data;
 
   int multiscan_levels =
-      0;  // Number of levels before switching to sorting algorithm
+     5;  // Number of levels before switching to sorting algorithm
 };
 }  // namespace tree
 }  // namespace xgboost