GeForce RTX 3060

A. K. A. Elsayed, A. Al-obaidi, H. E. Z. Farag

The rapid advancement of Artificial Intelligence (AI) is driving unprecedented computational demands, posing significant challenges to datacenter infrastructure and threatening the stability and resilience of modern power grids. This study presents an open-access dataset featuring a diverse set of AI training sessions recorded at sub-second resolution, designed to advance research on the energy consumption profiles of AI workloads and their interactions with power grid dynamics in datacenter environments. The dataset contains 32 training sessions on high-performance H100 and B200 8-GPU nodes and 40 sessions on consumer-grade NVIDIA GeForce RTX 3060 GPUs, encompassing over 1.8 million samples. Each session records power demand, CPU and GPU utilization, per-GPU power, memory usage, and temperature across diverse AI tasks (at the node scale, temperature refers to the GPUs temperature), including forecasting, classification, reinforcement learning, and text and image generation. Data quality was verified through detailed technical validation, including timing accuracy, hardware limit conformance, and cross-metric correlation analysis. Measurements remained within manufacturer-specified thermal and power envelopes, and observed correlations among power, utilization, temperature, and current were consistent with established processor and GPU behavior. The dataset provides a robust foundation for modeling AI datacenter energy behavior, system-level performance analysis, and power grid connection impact assessment studies.

Laila Nassef

In this work we systemically study performance of deep learning models for Natural Language Processing (NLP) and Computer Vision (CV) tasks using two popular representative architectures, ResNet-50 and BERT across two configurations: a standard ONNX Runtime configuration and an optimized TensorRT configuration. The main goal is to measure and compare inference time, throughput, CPU and GPU utilization as well as memory usage of all models on Domain-Specific Architectures (DSAs), in this case an NVIDIA GeForce RTX 3060 GPU. We experimentally demonstrate these trade-offs of latency-focused and throughput-focused optimizations and implications for at-scale deployment in realistic resource-constrained environments. The main findings show that the TensorRT-Optimized configuration yields a much higher throughput (up to 432 inferences per second with ResNet-50), and that the Standard configuration shows lower inference time, which is more appropriate for latency-sensitive applications. It should be noted that due to its dense transformer structure and large number of parameters, BERT has a much higher resource demand than ResNet-50, highlighting how model choices need to match performance with the constraints imposed by their deployment. Analysis of CPU and GPU utilization further illustrates the efficiency gains and potential bottlenecks associated with each configuration. Along with the benchmarking results, we also describe optimizations for serving the model: dynamic batching, mixed-precision training and memory management techniques to improve throughput as well as inference time. This study gives practitioners rich information to choose between model configurations and optimization strategies for an effective deployment of NLP and CV models on DSAs.

Yujie Zhang, Min Xie, Tingting Zhang, Ruixiao Zhao, Xinchen Li

Currently, most object detection systems based on embedded are single-modal, which results in relatively low detection accuracy. In contrast, dual-modal object detection systems offer higher accuracy but come with increased hardware resource requirements and higher power consumption. By employing a lightweight dual-modal object detection network for RGB and infrared images and a parallel-pipeline reconfigurable hardware structure based on block convolution, this paper designs a high-performance object detection system based on FPGA. Compared to the original network model, the lightweight processed model after undergoing techniques such as pruning, sparsity and quantization, demonstrates a significant reduction of parameters to 24 % of the original, with only a 1.1 % drop in mAP@0.5. Based on the Milianke F25 development board equipped with a Xilinx xczu15eg chip, the system achieves a detection speed of 29.3 ms, with an energy efficiency improvement of approximately 4.26 times compared to the NVIDIA GeForce RTX 3060.

Xun-Yao Lin

Aiming at the problems of low efficiency and large error of traditional three-dimensional clothing cutting, this paper proposes an intelligent recognition system for three-dimensional clothing cutting based on computer vision technology, and designs a multi-scale feature fusion and dynamic template matching algorithm. The algorithm realizes multi-scale feature extraction by improving the convolutional neural network, introduces the attention mechanism to enhance the key feature recognition, combines dynamic programming to optimize the template matching path, and realizes adaptive template generation and dynamic matching. A data set containing 2000 sets of images of different styles and materials is constructed for experiments. The experimental environment is Intel Core i7-12700H processor, NVIDIA GeForce RTX 3060 graphics card, 16GB memory, and Python 3.8 and PyTorch 1.12 framework. The results show that the accuracy of the algorithm is 96.8%, which is 23% higher than the traditional algorithm and 11% higher than the existing advanced algorithm, the recall rate is 95.5%, the F 1 value is 96.1%, and the average running time is 0.8 seconds, which is 1.2 seconds shorter than the traditional algorithm. Experiments have proved that the system and algorithm provide an efficient and accurate intelligent recognition solution for clothing three-dimensional cutting, effectively promoting the intelligent process of clothing production.

Filip Krużel, M. Nytko

This article analyses the usability of Intel ARC 770 based on Intel Xe Architecture for scientific computational tasks based on the Numerical Integration in Finite Element Method (FEM). To achieve a thorough comparison, we employed our proprietary auto-tuning algorithm, allowing us to assess architectural differences between Intel’s solution and the widely adopted Nvidia architecture. As a reference point, we selected the GeForce RTX 3060 Laptop due to its comparable performance and similar price at the time of release. Intel’s latest GPU architecture represents a significant step in the company’s ongoing efforts to establish a foothold in the GPGPU market, challenging Nvidia’s longstanding dominance. While our benchmarking results indicate that the ARC 770 delivers performance on par with reference Nvidia architecture, we identified distinctive behavioural characteristics that set it apart from previously analyzed architectures. These unique attributes could affect computational efficiency and workload distribution, making Intel’s approach an intriguing alternative for scientific computing. Through a series of experiments and simulations, we aim to provide deeper insight into how architectural variations influence the efficiency of FEM numerical integration. Given that the tested Intel GPU shares similarities with the professional Xe-HPC lineup, our research also explores its potential applicability beyond consumer-grade hardware, positioning it as a viable option for high-performance computing tasks.