Similar to This
Previous

NVIDIA Sweeps MLPerf Training and Endpoints as Vera Rubin Debuts With Leading Performance
blogs.nvidia.com/blog/mlperf-training-endpoints-vera-rubin/

Microsoft patent describes turning idle gaming GPUs into paid AI training hardware
videocardz.com/newz/microsoft-could-pay-gamers-for-using-spare-gpu-capacity-to-train-ai

How Nvidia’s export-banned Blackwell chips are showing up in Chinese AI models: A new research study has found that leading Chinese open-weight AI models were ...
indianexpress.com/article/technology/artificial-intelligence/nvidia-banned-gpus-training-chinese-ai-models-key-findings-10964970/

AI Agent Benchmarks Need to Measure User Intent: Proposing a new metric for whether AI does what you actually want
spectrum.ieee.org/ai-agent-benchmark

NVIDIA Blackwell Sweeps Every MLPerf 6.0 Benchmark With No Competition In Sight, While GB300 Systems Run Up to 60% Faster Than GB200
wccftech.com/nvidia-blackwell-sweeps-mlperf-6-0-benchmark-no-competition-in-sight/
GitHub - elsung/blackwell-16gb-llm-starter: Tested setup, configs, and benchmarks for local LLM inference on 16 GB NVIDIA Blackwell GPUs.
github.com/elsung/blackwell-16gb-llm-starter

NVIDIA Is Among the First to Submit MLPerf Inference v6.0 Benchmarks With Blackwell Ultra, and It’s Total Domination Over Competitors
wccftech.com/nvidia-is-among-the-first-to-submit-mlperf-inference-v6-0-benchmarks/

Squashing 'fantastic bugs' hidden in AI benchmarks: After reviewing thousands of benchmarks used in AI development, a Stanford team found that 5% could ...
techxplore.com/news/2025-12-fantastic-bugs-hidden-ai-benchmarks.html

NVIDIA Wins Every MLPerf Training v5.1 Benchmark: In MLPerf Training v5.1, NVIDIA swept all seven tests, delivering the fastest time to train across LLMs ...
blogs.nvidia.com/blog/mlperf-training-benchmark-blackwell-ultra/

Nvidia's Blackwell Ultra Dominates MLPerf Inference: Nvidia's Blackwell Ultra chip is setting new standards in AI performance. How does it achieve nearly 50% ...
spectrum.ieee.org/mlperf-inference-51

MLPerf v5.1 AI Inference Benchmark Showdown: NVIDIA Blackwell Ultra GB300 & AMD Instinct MI355X In T…
wccftech.com/mlperf-v5-1-ai-inference-benchmark-showdown-nvidia-blackwell-ultra-gb300-amd-instinct-mi355x/

Intel Project Battlematrix With Arc Pro B60 GPUs Delivers 4x Better Perf/$ Versus NVIDIA In MLPerf v5.1 …
wccftech.com/intel-project-battlematrix-arc-pro-b60-gpus-4x-perf-versus-nvidia-mlperf-v5-1-benchmarks/

Dodgy Huawei chips nearly sunk DeepSeek's next-gen R2 model: : Chinese AI model dev still plans to use homegrown silicon for inferencing
theregister.com/2025/08/14/dodgy_huawei_deepseek/
Key FactsChina’s Chip Pressure Delays DeepSeek R2 Launch
Aug 14, 2025 · Reading time: ~30 sec.
Summary created with AI assistance · Source: techspot.com
Summary created with AI assistance · Source: techspot.com
- Chinese authorities pushed DeepSeek to replace Nvidia hardware with Huawei chips for the R2 model.
- The switch caused unstable performance, slower interconnect speeds, and software limitations during training.
- Huawei engineers were deployed to assist, but training on the Ascend platform never succeeded.
- The R2 release was postponed from May to a later date amid ongoing hardware shortages.

MLPerf Client 1.0 AI benchmark released — new testing toolkit sports a GUI, covers more models and tasks, and supports more hardware acceleration paths
tomshardware.com/software/mlperf-client-1-0-ai-benchmark-released-new-testing-toolkit-sports-a-gui-covers-more-models-and-tasks-and-supports-more-hardware-acceleration-paths

Nvidia’s Blackwell Conquers Largest LLM Training Benchmark: New MLPerf training results put AMD’s MI325X accelerator on par with Nvidia’s H200
spectrum.ieee.org/mlperf-training-5

Salesforce research lays the foundations for more reliable enterprise AI agents: Business leaders face some big challenges if they want to use agents ...
zdnet.com/article/salesforce-research-lays-the-foundations-for-more-reliable-enterprise-ai-agents/

The way we measure progress in AI is terrible: Many of the most popular benchmarks for AI models are outdated or poorly designed.
technologyreview.com/2024/11/26/1107346/the-way-we-measure-progress-in-ai-is-terrible/
Next

