diff --git a/doc/design/images/multigpu_allreduce.graffle b/doc/design/images/multigpu_allreduce.graffle
new file mode 100644
index 0000000000000000000000000000000000000000..cb5bc420ceafe8ba4c87694d44ee4e5e4ad06779
Binary files /dev/null and b/doc/design/images/multigpu_allreduce.graffle differ
diff --git a/doc/design/images/multigpu_allreduce.png b/doc/design/images/multigpu_allreduce.png
new file mode 100644
index 0000000000000000000000000000000000000000..87a1b3e8f6dd4a713ec9df9f0037d1da04e9178a
Binary files /dev/null and b/doc/design/images/multigpu_allreduce.png differ
diff --git a/doc/design/images/multigpu_before_convert.graffle b/doc/design/images/multigpu_before_convert.graffle
new file mode 100644
index 0000000000000000000000000000000000000000..6c35ab1b21fb76ceae82d3693ed0d085b5bc0855
Binary files /dev/null and b/doc/design/images/multigpu_before_convert.graffle differ
diff --git a/doc/design/images/multigpu_before_convert.png b/doc/design/images/multigpu_before_convert.png
new file mode 100644
index 0000000000000000000000000000000000000000..9c8f7711165d80a2fa3911280fdee91855a401b1
Binary files /dev/null and b/doc/design/images/multigpu_before_convert.png differ
diff --git a/doc/design/paddle_nccl.md b/doc/design/paddle_nccl.md
new file mode 100644
index 0000000000000000000000000000000000000000..c7dac70998a6cfec3a6d2fc72b698ff9722e6805
--- /dev/null
+++ b/doc/design/paddle_nccl.md
@@ -0,0 +1,65 @@
+# Design Doc: NCCL support in Paddle Fluid
+
+## Abstract
+
+This Design Doc refers to the NCCL feature in paddle. We propose an approach to support NCCL library both on a single machine and multiple machines. We wrapper the NCCL primitives `Broadcast`, `Allreduce`, `Reduce` as operators to utilize Multi-GPU powers in one script.
+
+
+## Motivation
+
+[NCCL](https://developer.nvidia.com/nccl) is a NVIDIA library support Multi-GPU communicating and optimized for NVIDIA GPUs, it provides routines such as all-gather, all-reduce, broadcast, reduce, reduce-scatter, that can achieve high bandwidth over PCIe and NVLink high-speed interconnect. With NCCL library, we can easily accelerate the training in parallel.
+
+- Pros
+1. easily plug-in with [NCCL2](https://developer.nvidia.com/nccl) library.
+1. high performance in NVIDIA GPUs.
+1. MPI like primitives, which have low learning cost for users.
+
+- Cons
+1. Only design for NVIDIA GPUs, not a general multi-device solution.
+1. Although NCCL1 is opensourced under BSD license, but NCCL2 is not opensourced anymore.
+
+At the beginning of training, the framework needs to distribute the same parameters to every GPU, and merge the gradients at any time user interests.
+
+As a result, during training, we need the operations of peer to peer copy between different GPUs, aggregating gradients/parameters from GPUs, and broadcasting parameters to GPUs. Every GPU only need to run the operator with correct place information.
+
+Besides, it needs interfaces to synchronize model update with each different GPU Cards.
+
+## Implementation
+
+As mentioned above, we wrap the NCCL routines as several kinds of operators. Need to note that NCCL need to create Communicator between gpu at the beginning, so there is a NCCLInit operator created.
+
+### Transpiler
+
+To be compatible with [parameter server design doc](https://github.com/PaddlePaddle/Paddle/blob/develop/doc/design/ops/dist_train.md), the transpiler compiles the user defined operation graph into sub-graphs to be executed on different devices.
+
+1. The user-defined model will be a single device program
+
+2. Broadcast/Reduce operators between GPUs will be inserted into the program, even for the multi-node, may insert the `Send`, `Recv` operator.
+
+ *Broadcast, AllReduce in a single machine. And Broadcast, AllReduce, [Send, Recv](https://github.com/PaddlePaddle/Paddle/blob/develop/doc/design/ops/dist_train.md#graph-converter) in multiple machines*
+
+ 
+
+After compiling, the graph as shows
+
+
+
+Operators are added to the sub-graphs. Every GPU assigned a role of `rank0`, `rank1` etc.
+
+- **Broadcast**. Broadcast operator distribute initialized parameter to all the GPUs from the GPU who owns it. e.g. from`rank0` GPU.
+- **AllReduce**. AllReduce operator synchronizes parameters/gradients between GPUs. AllReduce implemented in the Ring-Based communicating method, avoid of the bottle neck in a single GPU.
+
+Need to notice that AllReduce operator force GPUs synchronized at that point. The whole training process in asynchronous or synchronous mode depends on the AllReduce point in the graph.
+
+As it shown in the picture, when each GPU compute the gradient of `W`, followed with a `AllReduce` operator, accumulate the `dW` to full batch of data, then run the optimize process individually and apply the gradient to its `W`.
+
+- **AllReduce**
+ Need to note that our AllReduce operator is a ring-base AllReduce implementation. If we use the NCCL2 AllReduce primitive, every GPU optimized full batch of data, wasted (n-1) GPU compute resources. In addition, NCCL2 built-in AllReduce will only utilize the communicating resource during synchronization, then update the gradient will be a subsequent phase. In fact, we can amortize the update gradient time cost into the communicating phase. The process is
+1. Every parameter has its root card. That card will responsible for aggregating the gradients from GPUs.
+2. The whole model's parameter will be hashed to different root card, ensure the load balance between GPUs.
+3. Logically neighberhood card will start send parameter to the next one. After one round, the parameter main card will aggregate the full gradients.
+4. Then the root card will optimize the parameter.
+5. This parameter card will send its optimized result to its neighberhood, then the neighberhood will send parameter to its next one.
+6. Finish the sychronization round.
+
+The total time cost will be 2 * (n-1) * per-parameter-send-time, we reach the goal of amortize the upgrade time into communicating phase.