Abstract

The Dell Technologies HPC & AI Innovation Lab recently submitted results to the MLPerf Inference v1.1 benchmark suite. These results provide our customers with transparent information about the performance of Dell EMC servers. This blog highlights the enhancements between the MLPerf™  Inference v1.0 and MLPerf Inference v1.1 submissions from Dell Technologies. These enhancements include improved GPU performance and new software to extract performance. Also, this blog compares server and GPU configurations from the MLPerf Inference v1.0 and v1.1 submissions.

Configuration comparison

The MLPerf Inference submissions focus was on outperforming the expectations outlined by MLPerf. For an introduction to the MLPerf Inference v1.0 performance results, we recommend that you read this blog published by Dell Technologies.

The following table provides the software stack configurations from the two submissions for the closed division benchmarks:

Table 1: MLPerf Inference v1.0 and v1.1 software stacks 

The following table shows the Dell EMC servers used for the MLPerf Inference v1.0 and v1.1 submissions:

Table 2: Servers used for the MLPerf Inference v1.0 and v1.1 submissions

Besides the upgrades in the software stack that are detailed in the preceding table and the results from the latest hardware, differences between the MLPerf Inference v1.0 and v1.1 submissions include:

• The Multistream scenario has been deprecated in MLPerf v1.1.
• The total number of submitters increased from 17 to 21.
• There were 1725 total submissions to MLCommons™ in v1.1.

MLPerf Inference v1.0 compared to MLPerf Inference v1.1

We compared the MLPerf v1.0 and v1.1 submissions by looking at results from an identical server and the same GPU configurations used in both rounds of submission. For both submissions, Dell Technologies submitted results for the Dell EMC PowerEdge XE8545 server configured with four A100 SXM 80 GB GPUs. The PowerEdge XE8545 servers used a combination of the latest AMD CPUs and powerful NVIDIA A100 Tensor Core GPUs. The PowerEdge XE8545 Spec Sheet provides additional details about the server.

The following figure shows nearly level performance across the two submissions, which allows for a fair comparison between the submissions. Also, it shows that we need to be aware of the software upgrades listed in Table 1, no matter how minimal.

Figure 1: Relative performance comparison of PowerEdge XE8545 4 x A100 SXM 80 GB in MLPerf v1.0 and v1.1

Dell EMC systems improvements for MLPerf Inference v1.1

This section provides detailed comparisons of various GPUs across the MLPerf Inference v1.0 and v1.1 submissions to show an expansion of Dell EMC server and GPU configurations that are available.

A100 40 GB GPU compared with A100 80 GB GPU

Dell EMC DSS 8440 server

The Dell EMC DSS 8440 server delivers high performance at a lower cost compared to our competitors. By offering support for four, eight, or 10 GPUs, this server excels in processing capacity along with a flexible infrastructure. The DSS 8440 server delivers high performance for machine learning workloads. The DSS 8440 Spec Sheet provides more details about the server.

The following figure compares two DSS 8440 servers configured with NVIDIA A100 Tensor Core GPUs. For the v1.0 submission, the DSS 8440 server was configured with the A100 40 GB GPU (shown in blue). For the v1.1 submission, the DSS 8440 server was configured with the A100 80 GB GPU (shown in orange). Across the different models, the performance improvement was between three percent to 20 percent, favoring the system with the A100 80 GB GPU. The more than 10 percent performance improvement can be attributed to the frequency of each card; the A100 80 GB GPU is a 300W card whereas the A100 40 GB GPU is 250W card.

Figure 2:  Relative performance comparison of DSS 8440 10 x A100 PCIe 40 GB and 80 GB in MLPerf v1.0 and v1.1

Dell EMC PowerEdge R750xa server

The PowerEdge R750xa server is ideal for Artificial Intelligence (AI)/Machine Learning (ML)/Deep Learning (DL) training and inferencing, high performance computing, and virtualization. See the Dell EMC PowerEdge R750xa Spec Sheet for more information about the server.

For this comparison, the server for both submissions was consistent. For the MLPerf v1.0 submission, the PowerEdge R750xa server was configured with four A100 40 GB GPUs. For the MLPerf v1.1 submission, the PowerEdge R750xa server was configured with four A100 80 GB GPUs. The following figure shows that for the MLPerf v1.1 submission, extra performance was extracted from the system. Across the various models, the MLPerf v1.1 results are seven percent to 22 percent better than the results from the MLPerf v1.0 submission. In the Resnet50 benchmark, the MLPerf v1.1 results are an impressive 15 and 19 percent better in the Offline and Server scenarios respectively.

Figure 3: Relative performance of PowerEdge R750xa 4 x A100 40 GB GPU and 80 GB in MLPerf v1.0 and v1.1 respectively

Dell EMC PowerEdge XE8545 server

For the MLPerf v1.0 submission, the PowerEdge XE8545 server was configured with the A100 SXM4 40 GB GPU (shown in blue in figures 4 and 5) and the A100 SXM4 80 GB GPU (shown in orange in figures 4 and 5). For the MLPerf v1.1 submission, the PowerEdge XE 8545 server was configured with the A100 SXM4 80 GB GPU (shown in gray in figures 4 and 5). It was expected that for the MLPerf v1.0 submission, the A100 SXM4 80 GB GPU would outperform the A100 SXM4 40 GB GPU. Across the models in the MLPerf v1.1 submission, the A100 SXM4 80 GB GPU performed between negative one percent (a negative value indicates a performance deficit, noted for SSD ResNet34 in Figure 5) and eight percent better than the identical system in the MLPerf v1.0 submission. Interestingly, for the SSD Resnet-34 benchmark, the A100 GPU in the MLPerf v1.0 submission slightly outperformed the A100 GPU in the MLPerf v1.1 submission.

Figure 4: Performance of PowerEdge XE8545 4 x A100 40 GB and 80 GB in MLPerf v1.0 and 80 GB in MLPerf v1.1 for ResNet50 and RNNT

Figure 5: Performance of PowerEdge XE8545 4 x A100 40 GB and 80 GB in MLPerf v1.0 and 80 GB in MLPerf v1.1 for BERT and SSD ResNet34

NVIDIA A30 GPU compared with NVIDIA A40 GPU

This comparison considers the NVIDIA A40 and NVIDIA A30 Tensor Core GPU. For a fair comparison between the two GPUs, the DSS 8440 server configuration was consistent across the two submissions. For the MLPerf v1.0 submission, the DSS 8440 server was configured with ten A40 GPUs. For the MLPerf v1.1 submission, the server was configured with eight A30 GPUs. For a clear interpretation of the two GPUs, the results in Figure 6 are presented as the per card performance numbers, which means that the throughput results from the A40 GPU have been divided by ten and the results from the A30 GPU have been divided by eight.

The system configured with the A30 GPU performed 15 to 111 percent better than the A40 GPU across the various benchmarks. The A30 GPU is ideal for inference as it is configured with a High Bandwidth Memory (HBM2) and a higher GPU frequency. The A40 GPU is positioned more for Virtual Desktop Infrastructure (VDI) and other workloads. 

Figure 6: Per card relative performance comparison of the DSS 8440 server with A30 and A40 GPUs in MLPerf v1.0 and v1.1

Comparison of NVIDIA T4, A30, and A10 GPUs

This comparison considers three submissions on three different servers. The numbers are divided to display per card performance.

The Dell EMC PowerEdge XE2420 server is a specialty edge server that supports demanding applications at the edge, retail applications and analytics, manufacturing and logistics applications, and 5G cell processing. See the PowerEdge XE2420 Spec Sheet for more information. Our lab configured the system with four NVIDIA Tesla T4  GPUs that have been optimized for high utilization while also performing in an energy-efficient manner. The results from this system were published in the MLPerf Inference v1.0 Results.

The second server in this comparison is the DSS 8440 server, which was configured with eight NVIDIA A30 GPUs. The final server in this comparison is the PowerEdge XE2420 server, which was configured with two NVIDIA A10 GPUs.

The three cards in this comparison have different form factors; the A10 and A30 GPUs are larger than the T4 GPU. The following figure shows that the A30 GPU performed better than the other two GPUs. Across the various benchmarks, the A30 GPU performed between 204 and 360 percent better than the T4 GPU and between five percent and 57 percent better than the A10 GPU.

Figure 7: Comparison of T4, A30, and A10 GPUs for DLRM

Figure 8: Comparison of T4, A40, and A10 GPUs for ResNet50, RNNT, and SSD ResNet34

Comparison of NVIDIA T4 GPU, A30 Multi-Instance GPU (MIG), and A100 MIG

This comparison also considers three submissions on three different servers. The results from the Resnet50 and SSD Resnet34 benchmarks have been divided to display per card performance.

The PowerEdge XE2420 server was configured with four NVIDIA Tesla T4 GPUs. The results for this system are from the MLPerf v1.0 submission. The PowerEdge R7525 server was configured with three NVIDIA A30 GPUs. MIG was enabled on all these GPUs with a profile of 1g6gb. We did not publish the A30 MIG results on the PowerEdge R7525 server to MLCommons, but the results are compliant.

The PowerEdge R750xa server was configured with four NVIDIA A100 80 GB GPUs, which support Multi-Instance GPU (MIG) and Peripheral Component Interconnect Express (PCIe). MIG is an enhancement for NVIDIA GPUs with the Ampere architecture that allows for seven secure partitions of GPU instances. This architecture is beneficial because it allows for increased parallelism. The results from this system were submitted in the MLPerf Inference v1.1 submission. There are different sizes of MIG slices. The configuration for the A30 and A100 GPUs used the smallest slice possible. For example, the A100 GPU was divided into seven slices and the A30 GPU into four slices.

The following figures show results across the MLPerf v1.0 and v1.1 submissions from Dell Technologies  for ResNet50 and SSD ResNet34. Figure 9 shows per physical GPU results. For the ResNet50 Offline benchmark, the A30 GPU performed 232 percent better than the T4 GPU, while the A100 GPU performed 76 percent better than the A30 GPU. In the ResNet 50 Server mode, the A30 GPU outperformed the T4 GPU by 50 percent and the A100 GPU outperformed the A30 GPU by 23 percent. We observed a similar trend across the Offline and Server modes where the A100 GPU outperformed the A30 GPU, which outperformed the T4 GPU.

Figure 9: Per card performance of the T4 GPU, A30 MIG, and A100 MIG for ResNet50

In the SSD ResNet34 benchmark, we observed a similar trend where the performance of the A100 GPU was better than the performance of the A30 GPU, which performed better than the T4 GPU. In the Offline mode of the SSD ResNet34 benchmark, the A30 GPU performed 243 percent better than the T4 GPU, and the A100 GPU performed 77 percent better than the A30 GPU. In the Server mode, the A100 GPU outperformed the A30 GPU by 93 percent and the A30 GPU performed 198 percent better than the T4 GPU.

Figure 10: Per card performance of the T4 GPU, A30 MIG, and A100 MIG for SSD ResNet34

Conclusion

This blog has provided a brief introduction to MLPerf Inference benchmarking and a summary of the Dell Technologies submission from MLPerf Inference v1.0. Also, it highlighted the differences in the software stack between the MLPerf v1.0 and v1.1 submissions. This blog quantified results from various server and GPU configurations across the two rounds of MLPerf submissions and displayed noteworthy and relevant performance comparisons.

When comparing the A100 40 GB to the A100 80 GB GPUs on the Dell EMC DSS 8440 server, the latter exhibited an 11 percent increase in performance. On the Dell EMC PowerEdge R750xa server, the A100 PCIe 80 GB GPU performed 12 percent better than the A100 PCIe 40 GB GPU. The Dell EMC PowerEdge XE8545 server confirmed this result for the MLPerf v1.1 submission; the A100 SXM 80 GB GPU performed three percent better than an identical system from the MLPerf v1.0 submission.

The A30 and A40 GPU comparison showed that the former achieved a notable 42 percent performance improvement while maintaining the Dell EMC DSS 8440 server.

The comparison between the T4, A30, and A10 GPUs revealed that the A30 GPU performed significantly better than the T4 GPU and is considered a good upgrade for your ML workloads. The T4 GPU, A30 MIG, and A100 MIG were compared based on results from the ResNet50 and SSD-ResNet34 benchmarks.

Abstract                                              

The recent release of MLPerf Inference v3.1 showcased the latest performance results from Dell's new PowerEdge XE8640 and PowerEdge XE9640 servers, and another submission from the PowerEdge R760xa server. The data underscores the outstanding performance of PowerEdge servers. These benchmarks illustrate the surging demand for compute power, with PowerEdge servers consistently emerging on top across various models, claiming numerous top titles. This blog examines the expected performance for image classification, object detection, question answering, speech recognition, medical image segmentation and summarization, focusing specifically on the capabilities of the PCIe and SXM form factor NVIDIA H100 Tensor Core GPUs in the new generation PowerEdge systems.

Overview of top title results

The PowerEdge XE8640 and XE9640 servers won several #1 titles. 

For instance, the PowerEdge XE8640 server emerged as a winner in all benchmarks in the data center suite such as image classification, object detection, question answering, speech recognition, medical image segmentation, and summarization relative to other systems having four NVIDIA H100 SXM GPUs. The PowerEdge XE9640 server received #1 titles for all benchmarks previously mentioned relative to other liquid-cooled systems having four NVIDIA H100 SXM GPUs.

Comparison from the previous rounds of submission

The following figure shows the improvement customers can derive by using the new generation PowerEdge XE8640 and XE9640 servers from our previous generation PowerEdge XE8545 server.                           

Figure 1. Relative performance of PowerEdge XE8640 and PowerEdge XE9640 servers using the PowerEdge XE8545 server as a baseline reference (for the Y axis, the higher the better)

The graph shows that the relative performance improvement from the PowerEdge XE8545 server with four NVIDIA A100 SXM Tensor Core GPUs as a baseline (from MLPerf Inference v3.0) and the new generation severs such as the PowerEdge XE8640 and PowerEdge XE9640 servers using NVIDIA H100 Tensor Core GPUs. The improvement in performance is substantial, as evident from the graph. End users can derive a two- to four-times improvement in performance for different tasks in MLPerf Inference benchmarks. We see relatively higher performance with BERT benchmarks because of the NVIDIA H100 GPU’s FP8 support. 

Comparing air-cooled and liquid-cooled servers

The following figure shows the raw performance of PowerEdge XE8640 and XE9640 servers; this graph and the following graph provide relative scores. The graph includes all the benchmarks in the Inference closed data center suite that we submitted. Note that different benchmarks have different scales. All the benchmarks are presented in one graph, therefore, the y-axis is expressed logarithmically.

Figure 2. Performance of PowerEdge XE8640 and PowerEdge XE9640 servers

PowerEdge XE8640 and XE9640 servers are both great choices for inference workloads with four NVIDIA H100 SXM Tensor Core GPUs. The PowerEdge XE9640 server is a liquid-cooled server and the PowerEdge XE8640 server is an air-cooled server. The following figure shows the difference in performance between these systems; they both performed optimally. Both systems have similar effective throughput and render excellent performance as the CPU and GPU configurations are the same. 

Figure 3. Performance difference between PowerEdge XE9640 and XE8640 servers using the PowerEdge XE9640 server as a baseline

Impact of SXM over PCIe form factors

The following figure shows the performance of the PowerEdge R760xa server with NVIDIA H100 PCIe GPUs as the baseline and shows the performance improvement of PowerEdge XE9640 and PowerEdge XE8640 servers with NVIDIA H100 Tensor Core SXM GPUs. The graph demonstrates that the PowerEdge XE8640 server with NVIDIA H100 SXM GPUs performs approximately 1.25 to 1.7 times better than the PowerEdge R760xa server with NVIDIA H100 PCIe GPUs.

Figure 4. Performance difference between PowerEdge XE9640 and XE8640 servers with 4x H100 SXM and PowerEdge R760xa server with 4x H100 PCIe as a baseline

Because the NVIDIA H100 SXM GPUs have higher Thermal Design Power (TDP), if high performance is imperative, then using NVIDIA SXM GPUs is a great choice. 

Comparing efficiency of new and previous generation servers

The following figure shows the performance of the previous generation PowerEdge XE8545 server with NVIDIA A100 SXM GPUs compared to the new generation servers such as the PowerEdge R760xa server with the NVIDIA H100 PCIE form factor and the PowerEdge XE8640 and XE9640 servers with the NVIDIA H100 SXM form factor. We see that all the new generation servers rendered higher performance. Furthermore, our new generation PowerEdge R760xa server with four NVIDIA H100 PCIe GPUs is more power efficient than our previous generation PowerEdge XE8545 server with four NVIDIA A100 SXM GPUs. This result is because NVIDIA A100 SXM GPUs have higher TDP relative to the NVIDIA H100 PCIe GPU. 

Figure 5. Relative performance of PowerEdge R760xa, PowerEdge XE9640, and PowerEdge XE8640 servers using the PowerEdge XE8545 server as a baseline

Hardware overview

The following sections describe the system components. The appendix lists the system configurations in the benchmark.  

Dell PowerEdge XE8640 server

The PowerEdge XE8640 server is an air-cooled 4U server that accelerates traditional AI training and inferencing, modeling, simulation, and other high-performance computing (HPC) applications with optimized compute, turning data and automating insights into outcomes with a four-way GPU platform. Its powerful architecture and the power of two 4th Generation Intel Xeon processors with a high core count of up to 56 cores and the latest on-chip innovations to boost AI and machine learning operations.

The following figure shows the PowerEdge XE8640 server:

Figure 6.  Dell PowerEdge XE8640 server

Dell PowerEdge XE9640 server

The PowerEdge XE9640 server is a purpose-built direct liquid-cooled (DLC) 2U server for AI and HPC workloads. NVIDIA NVLink and Intel Xelink technologies in the PowerEdge XE9640 server allow seamless communication between the GPUs, pooling their memory and cores to tackle memory-coherent workloads such as large language models (LLM) efficiently.

The following figure shows the PowerEdge XE9640 server:

Figure 7. Dell PowerEdge XE8640

NVIDIA H100 Tensor core GPU

The NVIDIA H100 GPU is an integral part of the NVIDIA data center platform. Built for AI, HPC, and data analytics, the platform accelerates over 3,000 applications, and is available everywhere from the data center to the edge, delivering both dramatic performance gains and cost-saving opportunities. The NVIDIA H100 Tensor Core GPU delivers unprecedented performance, scalability,
 and security for every workload. With NVIDIA® NVLink® Switch System, up to 256
 NVIDIA H100 GPUs can be connected to accelerate exascale workloads, while the dedicated
 Transformer Engine supports trillion-parameter language models. The NVIDIA H100 GPU uses
 breakthrough innovations in the NVIDIA Hopper™ architecture to deliver industry-leading conversational AI, speeding up large language models by 30 times over the previous generation.

The following figure shows the NVIDIA H100 PCIe accelerator:

Figure 8. NVIDIA H100 PCIe accelerator

The following figure shows the NVIDIA H100 SXM accelerator:

Figure 9. NVIDIA H100 SXM accelerator 

Conclusion

The key takeaways include:

• Both the Dell PowerEdge XE8640 and Dell PowerEdge XE9640 servers are an excellent choice for inference. The performance of the air-cooled PowerEdge XE8640 server is almost identical to the liquid-cooled PowerEdge XE9640 server. While the PowerEdge XE9640 server is a 2U server, it requires additional cooling unit attachments. It is a good choice if there are space and temperature constraints, otherwise the PowerEdge XE8640 server is a great choice. 
• PowerEdge XE8640 and PowerEdge 9640 servers have received several top titles. They are clear leaders in inference compute. 
• New generation PowerEdge XE8640 and PowerEdge XE9640 servers with NVIDIA H100 GPUs have delivered 2- to 4-times improvement relative to the previous generation PowerEdge XE8545 server with NVIDIA A100 GPUs. Upgrading from the PowerEdge XE8545 sever would render higher performance.
• The PowerEdge XE9640 and PowerEdge XE8640 servers with four NVIDIA H100 SXM form-factor GPUs are significantly more effective than the PowerEdge R760xa server with four NVIDIA H100 PCIe GPUs by a factor of 1.25 to 1.7 times.

Our submission results to MLPerf Inference since its inception have continuously demonstrated significant performance improvements. We have submitted to different tasks to provide customers with a wide spectrum of possible results to review. This round marked a new and the first submission to MLPerf with PowerEdge XE8640 and XE9640 servers. Customers can rely on these high compute machines for their fast/low latency inference needs. If constrained by TDP or other factors, the PowerEdge R760xa server with the PCIe form factor is an excellent choice on which to run inference workloads.

Appendix

The following table lists the system configuration details for the servers described in this blog:

Table 1. System configurations                                                                                                                                                                                                       

MLCommons results

MLPerf system IDs:

• ID 3.0-0011
• ID 3.1-0064
• ID 3.1-0066
• ID 3.1-0067

Note: We reran the RetinaNet Offline benchmark for the PowerEdge R760xa server and the DLRMv2 benchmark for the PowerEdge XE8640 server to reflect the correct performance that the servers can render. Only these two results are not official due to MLCommons rules.

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 www.mlcommons.org for more information.