Benchmarking News
The latest Benchmarking coverage curated for game developers.
Evolve Blog breaks down inline vs pipeline ray tracing and why the choice still matters for engine architecture. The benchmark found similar results on NVIDIA and Intel, but AMD’s Radeon …
Evolve Blog details a fix for rasterization scoring on tiling mobile GPUs, where per-draw timestamps were badly undercounting work. The benchmark will now time whole render passes instead, which should …
arXiv cs.GR details GaussianBench, a new way to judge physics-integrated 3D Gaussian scene systems by whether their internal state is physically sound, not just visually convincing. The suite checks conservation, …
arXiv cs.GR details SsgCaps, a public dataset for benchmarking sound scene generation. The set uses human-engineered scenes tied to structured prompts and only public-domain audio, making it easier for teams …
A new 3D Gaussian Splatting method tackles gradient vanishing with truncated updates, aiming to make scene reconstruction more stable and less sensitive to initialization. The approach improves both static and …
Transparency in CAD is still a balancing act between correctness and speed, and a new survey maps the trade-offs across desktop and mobile pipelines. It highlights where approximate methods are …
Weak scalability benchmarks for visualization algorithms can be misleading when input size is scaled without preserving workload characteristics. A new study from Marvin Petersen, Jonas Lukasczyk, and Christoph Garth compares …
For teams building image-editing or in-game content tools, the big takeaway is that precise visual edits are still hard: across 11 models, the best result in PaintBench was only 17.1% …
For teams working on 3D shape retrieval or mesh similarity, this paper argues the evaluation pipeline matters as much as the descriptor itself. On FAUST-Reg and TOSCA, a Heat Kernel …
A new paper tackles a subtle but important benchmarking problem in Gaussian splatting: methods often look better simply because they densify to more points. The authors propose a quota-governed target-point …
For teams using generative video in design animation, this paper tackles the missing piece: objective evaluation. It proposes a fully automated framework that checks layout fidelity, motion correctness, temporal quality, …
AIvaluateXR introduces a robust evaluation framework for deploying large language models on XR devices, crucial for enhancing human-AI interaction. With benchmarking results from 17 models across four platforms, developers can …
The introduction of a novel graph analytics benchmark could significantly enhance performance evaluation for developers working with graph-based platforms. By selecting eight core algorithms and generating new synthetic datasets, this …