Skip to main content

3 posts tagged with "Ascend"

Huawei Ascend series AI chips

View all tags

里程碑!华为昇腾910C完成1.6万亿参数大模型全参数训练

· 7 min read
Industry Research Team

2026年6月5日,深圳发布官宣重磅消息:深圳河套学院联合哈工大(深圳)、华为等团队,用1000颗华为昇腾910C芯片,成功完成1.6万亿参数DeepSeek-V4-Pro大模型全参数后训练

这不是一次试探性的尝试,而是一次里程碑式的技术突破。它用无可辩驳的工程结果证明:国产AI芯片足以支撑世界级超大参数模型训练

为什么这很重要?

AI芯片的两道坎:"推理"与"训练"

  • 推理(Inference):用现成模型聊天、写文案。此前国产芯片已经能做
  • 训练(Training):调整模型参数让它学习新能力。全参数训练要同时调整1.6万亿个参数,难度拉满

此前,万亿级参数模型的全参数训练一直被英伟达H100/H200垄断。国产芯片只能做推理,无法做大规模训练。

这次突破的意义:国产算力从"能用"跨越到"好用",从"推理"跨越到"训练"。

技术细节

训练配置

项目参数
芯片华为昇腾910C × 1,000颗
模型DeepSeek-V4-Pro
参数量1.6万亿(1600B)
训练类型全参数后训练(Full Parameter Post-Training)
训练框架昇思(MindSpore)+ torch_npu
完成时间2026年6月5日官宣

性能指标

指标数值评价
算力利用率>30%工业级水平(海外顶级芯片~40%)
关键训练算子效率提升14%相比上一代910B
通信带宽利用率>60%(推测)MoE模型的All-to-All通信
稳定性1000颗卡连续训练无故障集群稳定性达标

💡 关于30%算力利用率:很多人觉得30%不高,但在大模型训练领域,这已经是非常不错的工业级水平。就算用最顶级的海外芯片,很多团队的实际利用率也就在40%左右。

昇腾910C详细规格

昇腾910C是华为在2024年4月24日(华为分析师大会)公布的AI训练/推理芯片,理论算力峰值达到800 TFLOPS(BF16精度),与英伟达H100处于同等量级。

参数昇腾910C昇腾910BNVIDIA H100
架构Ascend 910CAscend 910BHopper
制程TSMC 7nm(推测)TSMC 7nmTSMC 4NP
BF16算力800 TFLOPS256 TFLOPS989 TFLOPS(稀疏)
显存64GB HBM(推测)64GB HBM2e(B1/B2)80GB HBM3
显存带宽~2TB/s(推测)600 GB/s(B1/B2)3.35 TB/s
TDP~400W(推测)300-400W700W
量产时间2026年4月(正式量产)2022年11月2022年3月

关键升级

  • 算力提升3×:从910B的256 TFLOPS提升到800 TFLOPS
  • 软件生态完善:torch_npu适配PyTorch,昇思框架成熟
  • 集群稳定性:1000颗卡连续训练无故障(这是最大的突破)

技术挑战与解决方案

挑战1:万亿级模型的显存需求

1.6万亿参数模型,仅模型参数就需要:

  • FP16精度:1.6T × 2 bytes = 3.2 TB
  • 加上梯度、优化器状态:至少10 TB显存

华为的解决方案

  • 模型并行(Model Parallel):将模型分布到1000颗910C上
  • ZeRO优化器:优化显存占用
  • 梯度累积:分阶段更新参数

挑战2:万卡集群的通信效率

1000颗芯片训练时,卡间通信成为瓶颈。MoE模型需要All-to-All通信(每个专家可能需要与其他所有专家通信)。

华为的解决方案

  • HCCS(Huawei Collective Communication Scheduler):自研高速互联协议
  • 分层通信:节点内NVLink + 节点间HCCS
  • 通信-计算重叠:在计算的同时进行数据传输

挑战3:训练稳定性

万亿级模型训练需要数周甚至数月,任何一颗卡故障都可能导致整个训练中断。

华为的解决方案

  • 故障检测与自动恢复:实时监测卡的状态,故障时自动重启并恢复训练状态
  • 检查点(Checkpoint)优化:高频保存训练状态(每N步保存一次)
  • 昇腾集群管理软件:专门为企业级训练设计

与竞品对比

厂商芯片1.6万亿参数训练生态成熟度可用性
华为昇腾910C已完成⭐⭐⭐(进步中)中国本土
NVIDIAH100/H200✅ 工业标准⭐⭐⭐⭐⭐全球(受出口管制)
AMDMI300X✅ 可行⭐⭐⭐⭐全球
GoogleTPU v5p/8t✅ JAX原生⭐⭐⭐⭐Google Cloud

结论:昇腾910C在硬件性能上已经追上H100,软件生态仍有差距,但这次训练成功证明了工程可行性

行业影响

1. 国产算力的"遵义会议"

这次突破被业内称为国产算力的"遵义会议"——从此从被动防守转向战略反攻。

具体影响

  • 打破"国产芯片只能推理"的偏见
  • 证明国产芯片可以做frontier模型训练
  • 为国产大模型(如DeepSeek-V4、文心5.0)提供算力底座

2. 对英伟达的冲击

华为昇腾910C完成万亿级训练,意味着中国AI产业对英伟达的依赖度降低

场景此前现在
推理国产芯片可用国产芯片好用
训练必须用H100/H200可以用910C
大规模训练必须用H100集群可以用910C集群

3. 对国产芯片产业的提振

这次突破将带动整个国产AI芯片产业链:

  • 芯片设计:寒武纪、沐曦、摩尔线程等加速迭代
  • 晶圆制造:中芯国际、华虹等获得更多订单
  • 封装测试:长电科技、通富微电等受益

华为昇腾芯片路线图(2025-2028)

时间芯片定位
2025年Q1昇腾910C旗舰训练/推理(已量产)
2026年Q1昇腾950PR推理优化(~500 TFLOPS BF16)
2026年Q4昇腾950DT数据中心训练
2027年Q4昇腾960下一代旗舰
2028年Q4昇腾970再下一代

训练实战经验分享

深圳河套学院团队在训练中积累了宝贵经验:

✅ 成功经验

  1. 渐进式训练:从小模型(7B)开始,逐步扩大到1.6T
  2. 混合精度训练:BF16主训练 + FP32梯度累积
  3. 通信优化:All-to-All通信与计算重叠
  4. 故障恢复:每1000步保存一次检查点

⚠️ 遇到的挑战

  1. 显存碎片:长训练过程中显存碎片化严重,需要定期整理
  2. 通信瓶颈:MoE模型的All-to-All通信占训练时间的30%+
  3. 软件Bug:torch_npu偶有内存泄漏,需要重启训练进程

相关芯片

参考资料


本文基于公开报道整理。向深圳河套学院、哈工大(深圳)、华为等团队表示敬意——你们用工程实践证明了中国AI算力的可行性。

Huawei Ascend 950 Mass Production and the Full Picture of China's AI Chip Ecosystem

· 4 min read
Industry Research Team

June 2026 — Huawei's Ascend 950 series (950PR / 950DT) has entered formal mass production and delivery, a landmark event for China's AI chip industry in 2026. Meanwhile, Cambricon's MLU690 has begun shipping and Moore Threads has announced MTT S5000 specifications, formally establishing China's tri-polar AI chip landscape.

Ascend 950 Series: A Historic Breakthrough with Self-Developed HBM

Huawei HiSilicon's Ascend 950 series is the fourth-generation Ascend AI chip, first revealed at Huawei Connect 2025 in September and entering mass production in Q1 2026.

950PR (Prefill Inference Specialized)

ItemSpecification
ArchitectureDa Vinci v5 (SIMD + SIMT dual-model)
ProcessN+2 (SMIC domestic)
HBMHiBL 1.0 (Huawei self-developed) , 128 GB
FP8 Compute1 PFLOPS (HiF8 format)
TDP~400 W
TargetInference Prefill (video recommendation, real-time interaction)

950DT (Decode + Training Specialized)

ItemSpecification
ArchitectureDa Vinci v5 (SIMD + SIMT dual-model)
ProcessN+2 (SMIC domestic)
HBMHiZQ 2.0 (Huawei self-developed) , 144 GB, 4 TB/s
FP8 Compute1 PFLOPS (HiF8 format)
TDP~500 W
TargetInference Decode + Model Training

Historical Significance

Self-developed HBM (HiBL 1.0 / HiZQ 2.0) represents the most important technical breakthrough of Huawei Ascend 950 — this is the first time a Chinese enterprise has achieved self-developed mass production of HBM memory, completely eliminating dependence on SK Hynix / Samsung HBM supply. Combined with the domestic N+2 process, Ascend 950 has achieved full-chain domestic production from HBM → Compute Die → Packaging → System.

Cambricon MLU690: China's Only Native FP8 Support

Cambricon's seventh-generation AI chip MLU 690 (Siyuan 690) began volume production and shipping in H1 2026. This is the first domestic AI chip with native FP8 precision support.

ItemMLU 690
Process5nm (TSMC / SMIC)
FP8 dense2 PFLOPS
HBM192GB HBM3E, 5 TB/s
TDP~500 W
Unit Price (OAM)~$8,000-12,000

MLU 690's FP8 compute power (2 PFLOPS dense) is on paper comparable to NVIDIA Blackwell (B200 FP8 4.5 PFLOPS sparse). Leveraging its financing advantage as a STAR Market listed company, Cambricon targets 2026 revenue of ¥15-20B (2025: ¥7.2B).

Moore Threads MTT S5000: From Graphics to Training-Inference Unified

Moore Threads publicly disclosed detailed specifications of the MTT S5000 in February 2026, featuring the fourth-generation MUSA "Pinghu" architecture, single-card AI compute of 1,000 TFLOPS, 80GB GDDR6X memory, 1.6 TB/s bandwidth.

Moore Threads pursues a full-function GPU path (graphics rendering + AI compute + general-purpose compute), closest to NVIDIA's strategy. The founding team comes from former NVIDIA China, and the MUSIFY toolchain helps auto-migrate CUDA code to the MUSA platform, lowering ecosystem migration costs.

China's Tri-Polar AI Chip Landscape

DimensionHuawei AscendCambriconMoore Threads
Core ArchitectureDa Vinci v5MLUv07MUSA 4th Gen
ProcessN+2 domestic5nm6nm
FP8 Compute~1 PFLOPS2 PFLOPS0.5 PFLOPS (estimated)
HBM Self-Sufficiency✅ Self-developed HiBL/HiZQ❌ Purchased❌ Purchased
EcosystemCANN + MindSporeNeuWare + MindSporeMUSA + MUSIFY
AdvantageFull-chain domesticHighest FP8 computeFull-function + CUDA migration
2025 Revenue(Huawei internal)¥7.2B¥2.2B

Global Market Comparison (Q2 2026 Update)

TierVendorFlagship ChipFP8/PFLOPSHBMMass Production
Tier 1NVIDIARubin R20025 PF (sparse)288GB HBM42026 H2
Tier 2AMDMI40020 PF (dense)432GB HBM42026
HuaweiAscend 950DT1 PF (dense)144GB self-developed HBM2026 Q1
CambriconMLU6902 PF (dense)192GB HBM3E2026 H1
AWSTrainium 35.7 PF (dense)144GB HBM2025 Q4 GA
Tier 3IntelGaudi 31.8 PF128GB HBM2eIn production
GoogleTPU v74.6 PF(TFLOPS)192GB HBM2025
Moore ThreadsMTT S50001 PF80GB GDDR6X2025 Q1

Note: NVIDIA uses sparse compute as standard, while AMD / Huawei / Cambricon use dense — not directly comparable.

Outlook for H2 2026

  • NVIDIA Rubin R200: Official shipment in H2 2026, 288GB HBM4, 6-chip CoWoS-L packaging
  • Huawei Ascend 960: Roadmap H2 2027, expected FP8 compute doubled to 2 PFLOPS
  • Cambricon MLU790: Expected 2027, 3nm, 384GB HBM4, 2.5 PFLOPS
  • Moore Threads: Next-gen GPU expected with HBM3, 2× MTT S5000 compute

By 2026, China's AI chip industry has formed a complete product matrix from Training (Cambricon MLU690 / Ascend 950DT) → Inference (Ascend 950PR / Moore Threads S5000) → Systems (CloudMatrix / Distributed Clusters).


This article is based on public information from Huawei Connect 2025 (2025-09-18), industry analysis reports from April 2026, and the latest market data as of June 2026.

China AI Chip Landscape 2025: Ascend, Cambricon, Hygon — Who Will Dominate?

· 5 min read
Industry Research Team

Escalating U.S. export controls are forcing China's AI chip industry to accelerate self-reliance. By 2025, the discussion around domestic Chinese AI chips has shifted from "are they usable?" to "which one should I choose?"

This article systematically reviews the major players, core products, and actual deployment status of domestic AI chips, helping developers and procurement decision-makers understand the competitive landscape.