Ascend 950
PulseAugur coverage of Ascend 950 — every cluster mentioning Ascend 950 across labs, papers, and developer communities, ranked by signal.
- used by DeepInfer: Open-Source Deep Learning Deployment Toolkit for Image-Guided Therapy 90%
- used by DeepSeek-V4.1-Flash 90%
- used by SM100 90%
- instance of DeepInfer: Open-Source Deep Learning Deployment Toolkit for Image-Guided Therapy 70%
- instance of DeepSeek-V4.1-Flash 70%
- used by Cann 70%
- used by TensorRT-LLM 70%
- partners with Cann 70%
- affiliated with DeepInfer: Open-Source Deep Learning Deployment Toolkit for Image-Guided Therapy 50%
- affiliated with TensorRT-LLM 50%
5 day(s) with sentiment data
-
DeepSeek V5 to leverage Huawei Ascend chips for pre-training
DeepSeek is planning to use Huawei's Ascend chips for the pre-training of its upcoming DeepSeek V5 large model. This marks a significant shift from previous uses of Ascend chips, which were primarily for inference tasks…
-
Huawei's Ascend 950 SuperPoD wins WIC award, surpasses 1,000 deployments
Huawei's Ascend 950 SuperPoD has received a WIC Leading Technology Award from the World Internet Conference in Wuzhen. The system, which features 1,024 NPUs per node and utilizes Lingqu interconnect chips, has achieved …
-
DeepSeek open-sources DeepInfer for efficient LLM inference
DeepSeek has open-sourced DeepInfer, an inference framework designed for efficient large-scale model deployment. The toolkit supports offline inference for models like DeepSeek-V4.1-Flash on hardware such as Nvidia SM10…
-
Huawei Atlas AI Accelerator Updates and User Performance Insights
A Reddit user, an advocate for Huawei's Atlas AI Accelerator, shared an update on the latest card lineup and their personal experience. They detailed efforts to optimize models like QFN, GLM5.3-flash, Deepseek, and Meta…
-
DeepInfer deployment toolkit requires specialized hardware, not production-ready
The DeepInfer open-source deep learning deployment toolkit requires specific hardware, such as SM100 or Ascend 950, and lacks features like KV cache and load balancing, making it unsuitable for standalone production use…
-
DeepSeek open-sources DeepInfer framework to cut AI deployment costs
Chinese AI startup DeepSeek has open-sourced its DeepInfer framework, designed to make AI models more efficient and cost-effective to deploy. The framework supports DeepSeek-V4.1-Flash on Nvidia's SM100 Blackwell GPUs a…
-
DeepSeek open-sources DeepInfer inference framework for its models
DeepSeek has open-sourced DeepInfer, the inference framework that powers its production services, including the DeepSeek-V4.1-Flash model. This toolkit, released under the Apache 2.0 license, offers a rare glimpse into …
-
DeepSeek open-sources production inference engine, DeepInfer
DeepSeek has open-sourced its production inference engine, DeepInfer, which powers its large-scale services. The released code includes a substantial scheduler, MoE expert-parallel load balancing, speculative decoding, …
-
DeepSeek open-sources DeepInfer inference engine for large-scale AI models
DeepSeek has open-sourced its high-performance inference engine, DeepInfer, which is designed to power large-scale production services for their models. This release includes optimized kernels, support for various paral…
-
DeepSeek, Huawei Open-Source Tools to Challenge Nvidia's AI Dominance · 10 sources tracked
DeepSeek and Huawei have jointly released a suite of open-source tools designed to enhance the software ecosystem for Huawei's Ascend AI chips. This initiative aims to provide a viable alternative to Nvidia's dominant C…
-
Huawei unveils Atlas 950 SuperPoD servers with 1,024 NPUs and 256TB memory · 4 sources tracked
Huawei has showcased its Atlas 950 SuperPoD servers, a system comprising 1,024 NPUs designed for AI workloads. The system features extensive inter-rack connectivity, with each of the 32 LRS units utilizing 4x100G connec…
-
Fan Shi partners with Huawei for Ascend 950 computing power
Fan Shi has entered into a significant strategic cooperation agreement with Huawei, focusing on computing power. This partnership will see Fan Shi become one of the first AI companies to utilize Huawei's high-end domest…
-
Huawei open-sources AscendNPU IR to challenge Nvidia's AI hardware dominance
Huawei is challenging Nvidia's dominance in the AI hardware market by open-sourcing the core component of its BiSheng compiler, the AscendNPU IR. This move, aimed at supporting the upcoming Ascend 950 chip, includes nat…
-
Huawei open-sources AscendNPU IR, enhancing BiSheng compiler flexibility
Huawei has open-sourced the AscendNPU IR, a core component of its BiSheng compiler. This MLIR-based abstraction layer is designed for Ascend hardware, offering support for front-end languages like Triton and extending c…
-
New benchmarks and optimization tools emerge for LLM-generated hardware kernels
Two new research papers introduce benchmarks and optimization frameworks for large language models (LLMs) generating code for hardware accelerators. The first paper, KernelGenBench, offers a unified benchmark to evaluat…
-
China builds chip market amid US sanctions, eyes Russia alliance
China is developing its own chip supply market in response to U.S. export controls, potentially fostering a trade alliance with Russia. Russian entities like Sberbank are seeking Chinese-made processors, such as Huawei'…
-
Huawei targets South Korea AI chip market with Atlas SuperPods, challenging Nvidia
Huawei is preparing to enter the South Korean AI chip market in late 2026 with its Atlas 950 SuperPod platform, featuring clusters of up to 8,192 Ascend 950 accelerators. The company aims to challenge Nvidia's dominance…
-
Huawei considers Ascend AI chip deployment in Latin America
Huawei is exploring the deployment of its Ascend AI chips within its cloud and AI services in Latin America. This potential expansion would introduce Chinese-designed hardware into a market traditionally served by US su…
-
OpenURMA: Open-Source Implementation of Huawei's Unified Bus Protocol Released
Researchers have developed OpenURMA, an open-source implementation of Huawei's Unified Bus (UB) protocol, designed to improve datacenter network performance. This implementation aims to address bottlenecks in modern RDM…
-
Sberbank seeks Chinese AI chips amid sanctions, faces competition
Russia's Sberbank is seeking Chinese-made processors to power its GigaChat AI model due to Western sanctions. Sberbank CEO German Gref indicated interest in Huawei's Ascend 950 chips, but faces significant competition f…