Dell EMC Servers Excel in MLPerf™ Training v1.0 Benchmarks
Thu, 08 Jul 2021 15:28:25 -0000|
Read Time: 0 minutes
Dell Technologies has submitted MLPerf training v1.0 results. This blog provides an explanation of what is new with MLPerf training v1.0 and a high-level overview of our submissions. Results indicate that Dell EMC DSS8440 and PowerEdge XE8545 servers offer promising performance for Deep Learning training workloads across different areas.
MLCommons™ is a community that contains a consortium of experts in the Machine Learning/Deep Learning industry from different fields within AI technology. It consists of experts from industry, academia, startups, and individual researchers. MLPerf™ Training is the community-led test suite focusing on deep learning training. This test suite aims to measure how fast a system can train deep learning models across eight different problem types:
- Image classification
- Medical image segmentation
- Light-weight object detection
- Heavy-weight object detection
- Speech recognition
- Natural language processing
- Reinforcement learning
These benchmarks provide a consistent and reproducible way to measure accuracy and convergence on individual accelerators, systems, and cloud setups. As of June 2021, MLPerf™ Training released the latest v1.0 results in the fourth round of submissions of MLPerf Training. The following changes are new with v1.0:
- Addition of two benchmarks:
- RNN-T—RNN-T is a speech recognition model. Speech recognition accepts raw audio samples and produces a corresponding text transcription. It uses the Libri-speech dataset, which is derived from audiobooks. An example of the use of speech recognition is Google Voice Search.
- 3D-UNet—3D-Unet is a model for 3D medical image segmentation. It accepts 3D images that contain tumors; the model divides (or segments) the tumor from the other parts in the image. It uses the KiTs19 dataset. An example of the use of 3D medical image segmentation is for the identification of kidney tumors.
- Introduction of a uniform and more mature process for evaluation and submission:
- Reference Convergence Points (RCP) checker to ensure hyperparameters are assessed consistently and uniformly across different submissions.
- Other checkers such as compliance checker, system desc checker, and package checker to check the accuracy of the submission.
- Result summarizer to provide a submission summary.
- Retirement of two language translation benchmarks from v0.7:
BERT serves as a replacement for language model tasks.
The following figure demonstrates the numbers from the Deep Learning v1.0 benchmarks submitted by Dell Technologies:
Figure 1: MLPerf v1.0 results from Dell Technologies
Contributions from Dell Technologies
Our submissions focused on Dell EMC DSS 8440 and Dell EMC PowerEdge XE8545 servers. The DSS 8440 server is an Intel-based, PCIe Gen3 4U server that supports up to 10 double-wide PCIe GPUs, focused on Machine Learning/Deep Learning applications such as training. The 4U PowerEdge XE8545 server supports the latest 3rd Gen AMD EPYC processors, PCIe Gen4, and the latest NVIDIA A100 Tensor Core GPUs for cutting edge machine learning workloads. Both of these system configurations are NVIDIA-Certified, which means they have been validated for best performance and optimal scalability. The submission from Dell Technologies also included multinode training entries to showcase scale-out performance.
Multinode training is important. Training is compute intensive, therefore, more compute nodes are used while training models. Because extra compute nodes help to reduce the turnaround time, it is critical to showcase multiple nodes’ performance. Dell Technologies and NVIDIA are the only submitters that submitted multiple nodes on GPUs. The submissions from NVIDIA run on Docker with a customized Slurm environment to optimize performance; we submitted multinode submissions with Singularity on our DSS 8440 servers as well as Docker and Slurm submissions on PowerEdge XE8545 servers. Singularity is a secure containerization solution primarily used in traditional HPC GPU clusters. Setup scripts with singularity help traditional HPC customers run MLPerf™ Training on their cluster without the need to fully restructure their existing cluster setup.
The PowerEdge XE8545 server provides the best performing submission with an air-cooled solution for NVIDIA A100-SXM-80GB 500W GPUs. Typically, 500W GPUs of most vendors' systems are cooled with liquid, due to the challenges presented by the high TDP. However, Dell Technologies invested engineering and design time to solve the thermal challenge and allows customers to avoid the need for costly changes to a standard data center setup.
The DSS 8440 server submissions to MLPerf™ Training v1.0 using the latest generation NVIDIA A100 40 GB-PCIe GPUs show a 2.1 to 2.4 times increase from equivalent MLPerf™ Training v0.7 submissions using NVIDIA V100S PCIe GPUs. Dell Technologies is committed to bringing the latest performance advancements to customers as quickly as possible.
Out of 12 different organizations, Dell Technologies and NVIDIA are the only two organizations that submitted results for all eight models in the MLPerf™ training v1.0 benchmarking suite.
As a next step, we will publish more technical blogs to provide deep dives into DSS 8440 server and PowerEdge XE8545 server results.
Related Blog Posts
Multinode Performance of Dell PowerEdge Servers with MLPerfTM Training v1.1
Mon, 07 Mar 2022 19:51:12 -0000|
Read Time: 0 minutes
The Dell MLPerf v1.1 submission included multinode results. This blog showcases performance across multiple nodes on Dell PowerEdge R750xa and XE8545 servers and demonstrates that the multinode scaling performance was excellent.
The compute requirement for deep learning training is growing at a rapid pace. It is imperative to train models across multiple nodes to attain a faster time-to-solution. Therefore, it is critical to showcase the scaling performance across multiple nodes. To demonstrate to customers the performance that they can expect across multiple nodes, our v1.1 submission includes multinode results. The following figures show multinode results for PowerEdge R750xa and XE8545 systems.
Figure 1: One-, two-, four-, and eight-node results with PowerEdge R750xa Resnet50 MLPerf v1.1 scaling performance
Figure 1 shows the performance of the PowerEdge R750xa server with Resnet50 training. These numbers scale from one node to eight nodes, from four NVIDIA A100-PCIE-80GB GPUs to 32 NVIDIA A100-PCIE-80GB GPUs. We can see that the scaling is almost linear across nodes. MLPerf training requires passing Reference Convergence Points (RCP) for compliance. These RCPs were inhibitors to show linear scaling for the 8x scaling case. The near linear scaling makes a PowerEdge R750xa node an excellent choice for multinode training setup.
The workload was distributed by using singularity on PowerEdge R750xa servers. Singularity is a secure containerization solution that is primarily used in traditional HPC GPU clusters. Our submission includes setup scripts with singularity that help traditional HPC customers run workloads without the need to fully restructure their existing cluster setup. The submission also includes Slurm Docker-based scripts.
Figure 2: Multinode submission results for PowerEdge XE8545 server with BERT, MaskRCNN, Resnet50, SSD, and RNNT
Figure 2 shows the submitted performance of the PowerEdge XE8545 server with BERT, MaskRCNN, Resnet50, SSD, and RNNT training. These numbers scale from one node to two nodes, from four NVIDIA A100-SXM-80GB GPUs to eight NVIDIA A100-SXM-80GB GPUs. All GPUs operate at 500W TDP for maximum performance. They were distributed using Slurm and Docker on PowerEdge XE8545 servers. The performance is nearly linear.
Note: The RNN-T single node results submitted for the PowerEdge XE8545x4A100-SXM-80GB system used a different set of hyperparameters than for two nodes. After the submission, we ran the RNN-T benchmark again on the PowerEdge XE8545x4A100-SXM-80GB system with the same hyperparameters and found that the new time to converge is approximately 77.37 minutes. Because we only had the resources to update the results for the 2xXE8545x4A100-SXM-80GB system before the submission deadline, the MLCommons results show 105.6 minutes for a single node XE8545x4100-SXM-80GB system.
The following figure shows the adjusted representation of performance for the PowerEdge XE8545x4A100-SXM-80GB system. RNN-T provides an unverified score of 77.31 minutes:
Figure 3: Revised multinode results with PowerEdge XE8545 BERT, MaskRCNN, Resnet50, SSD, and RNNT
Figure 3 shows the linear scaling abilities of the PowerEdge XE8545 server across different workloads such as BERT, MaskRCNN, ResNet, SSD, and RNNT. This linear scaling ability makes the PowerEdge XE8545 server an excellent choice to run large-scale multinode workloads.
Note: This rnnt.zip file includes log files for 10 runs that show that the averaged performance is 77.31 minutes.
- It is critical to measure deep learning performance across multiple nodes to assess the scalability component of training as deep learning workloads are growing rapidly.
- Our MLPerf training v1.1 submission includes multinode results that are linear and perform extremely well.
- Scaling numbers for the PowerEdge XE8545 and PowerEdge R750xa server make them excellent platform choices for enabling large scale deep learning training workloads across different areas and tasks.
 MLPerf v1.1 Training RNN-T; Result not verified by the MLCommonsTM Association. The MLPerf name and logo are trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See http://www.mlcommons.org for more information.
Unveiling the Power of the PowerEdge XE9680 Server on the GPT-J Model from MLPerf™ Inference
Tue, 16 Jan 2024 18:30:32 -0000|
Read Time: 0 minutes
For the first time, the latest release of the MLPerf™ inference v3.1 benchmark includes the GPT-J model to represent large language model (LLM) performance on different systems. As a key player in the MLPerf consortium since version 0.7, Dell Technologies is back with exciting updates about the recent submission for the GPT-J model in MLPerf Inference v3.1. In this blog, we break down what these new numbers mean and present the improvements that Dell Technologies achieved with the Dell PowerEdge XE9680 server.
MLPerf inference v3.1
MLPerf inference is a standardized test for machine learning (ML) systems, allowing users to compare performance across different types of computer hardware. The test helps determine how well models, such as GPT-J, perform on various machines. Previous blogs provide a detailed MLPerf inference introduction. For in-depth details, see Introduction to MLPerf inference v1.0 Performance with Dell Servers. For step-by-step instructions for running the benchmark, see Running the MLPerf inference v1.0 Benchmark on Dell Systems. Inference version v3.1 is the seventh inference submission in which Dell Technologies has participated. The submission shows the latest system performance for different deep learning (DL) tasks and models.
Dell PowerEdge XE9680 server
The PowerEdge XE9680 server is Dell’s latest two-socket, 6U air-cooled rack server that is designed for training and inference for the most demanding ML and DL large models.
Figure 1. Dell PowerEdge XE9680 server
Key system features include:
- Two 4th Gen Intel Xeon Scalable Processors
- Up to 32 DDR5 DIMM slots
- Eight NVIDIA HGX H100 SXM 80 GB GPUs
- Up to 10 PCIe Gen5 slots to support the latest Gen5 PCIe devices and networking, enabling flexible networking design
- Up to eight U.2 SAS4/SATA SSDs (with fPERC12)/ NVMe drives (PSB direct) or up to 16 E3.S NVMe drives (PSB direct)
- A design to train and inference the most demanding ML and DL large models and run compute-intensive HPC workloads
The following figure shows a single NVIDIA H100 SXM GPU:
Figure 2. NVIDIA H100 SXM GPU
GPT-J model for inference
Language models take tokens as input and predict the probability of the next token or tokens. This method is widely used for essay generation, code development, language translation, summarization, and even understanding genetic sequences. The GPT-J model in MLPerf inference v3.1 has 6 B parameters and performs text summarization tasks on the CNN-DailyMail dataset. The model has 28 transformer layers, and a sequence length of 2048 tokens.
The official MLPerf inference v3.1 results for all Dell systems are published on https://mlcommons.org/benchmarks/inference-datacenter/. The PowerEdge XE9680 system ID is ID 3.1-0069.
After submitting the GPT-J model, we applied the latest firmware updates to the PowerEdge XE9680 server. The following figure shows that performance improved as a result:
Figure 3. Improvement of the PowerEdge XE9680 server on GPT-J Datacenter 99 and 99.9, Server and Offline scenarios 
In both 99 and 99.9 Server scenarios, the performance increased from 81.3 to an impressive 84.6. This 4.1 percent difference showcases the server's capability under randomly fed inquires in the MLPerf-defined latency restriction. In the Offline scenarios, the performance saw a notable 5.3 percent boost from 101.8 to 107.2. These results mean that the server is even more efficient and capable of handling batch-based LLM workloads.
Note: For PowerEdge XE9680 server configuration details, see https://github.com/mlcommons/inference_results_v3.1/blob/main/closed/Dell/systems/XE9680_H100_SXM_80GBx8_TRT.json
This blog focuses on the updates of the GPT-J model in the v3.1 submission, continuing the journey of Dell’s experience with MLPerf inference. We highlighted the improvements made to the PowerEdge XE9680 server, showing Dell's commitment to pushing the limits of ML benchmarks. As technology evolves, Dell Technologies remains a leader, constantly innovating and delivering standout results.
 Unverified MLPerf® v3.1 Inference Closed GPT-J. Result not verified by MLCommons Association.
The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use is strictly prohibited. See www.mlcommons.org for more information.