Skip to main content

Huawei Ascend 950 (950PR / 950DT)

Product Overview​

The Huawei Ascend 950 series (950PR / 950DT) is Ascend's fourth-generation AI chip, officially released and mass-produced in early 2026. It introduces FP8 / FP4 low-precision formats for the first time, with 1 PFLOPS of FP8 compute per card. It is also the first to adopt Huawei's in-house HBM (950PR uses HiBL 1.0, 950DT uses HiZQ 2.0), completely removing dependence on external HBM supply.

The 950PR (Prefill inference) and 950DT (Decode + training) are Huawei Ascend's first attempt at a "workload segmentation" strategy — same architecture, different memory configurations, precisely matched to the needs of different AI workloads.

Core Specifications​

ParameterAscend 950PRAscend 950DT
ArchitectureDa Vinci v5 (4th-generation Ascend)Da Vinci v5 (4th-generation Ascend)
ProcessN+2 (SMIC domestic)N+2 (SMIC domestic)
Programming ModelSIMD + SIMT dual modelSIMD + SIMT dual model
HBM TypeHiBL 1.0 (in-house, cost-optimized)HiZQ 2.0 (in-house, bandwidth-optimized)
HBM Capacity128 GB144 GB
HBM Bandwidth1.6 TB/s4 TB/s
FP8 Compute1 PFLOPS (HiF8 format)1 PFLOPS (HiF8 format)
FP4 Compute2 PFLOPS2 PFLOPS
BF16/FP16 Compute~500 TFLOPS~500 TFLOPS
INT8 Compute~2,000 TOPS~2,000 TOPS
TDP~400 W~500 W
PCIeGen 5 ×16Gen 5 ×16
InterconnectLingQu 2 TB/sLingQu 2 TB/s
Price¥70,000 per card (1/3 of the H200)~¥120,000-150,000 per card (estimated)
Launch2026 H12026 H1

Differentiated Positioning: 950PR vs 950DT​

Dimension950PR950DT
Target ScenarioInference Prefill (first-token generation)Inference Decode (token-by-token generation) + training
HBM Capacity128 GB (smaller; Prefill has low memory demand)144 GB (larger; training needs more memory)
HBM Bandwidth1.6 TB/s (moderate bandwidth demand for Prefill)4 TB/s (high bandwidth demand for Decode and training)
Typical ApplicationsVideo recommendation, real-time interaction, searchDialogue generation, text continuation, model training
PricingLower (cost-optimized)Higher (high performance)

Key Technical Breakthroughs​

1. In-House HBM (HiBL 1.0 / HiZQ 2.0)​

  • HiBL 1.0 (950PR): low-cost solution that resolves the HBM supply chokepoint problem
  • HiZQ 2.0 (950DT): high-bandwidth solution, 4 TB/s rivaling NVIDIA HBM3e
  • Fills a domestic gap, completely removing dependence on SK Hynix / Samsung

2. HiF8 Format — A Precision Revolution​

  • Huawei's in-house FP8 variant with precision close to FP16 and 2x the compute of FP16
  • Resolves the "low precision → accuracy loss" trade-off in large-model training
  • Combined with the FP4 format, overall compute utilization improves by 30%+

3. SIMD + SIMT Dual Programming Model​

  • SIMD: efficient vector computation (continuing the Da Vinci core strength of the 910C)
  • SIMT: newly added model supporting flexible scheduling
  • Memory access granularity goes from 512 bytes → 128 bytes
  • Discrete memory access efficiency improves 4x
  • Offers twin ASIC (Ascend Core) and GPGPU variants

4. CloudMatrix 384 System​

  • 384 950 chips form a supernode
  • Total compute: 384 × 1 PFLOPS FP8 ≈ 384 PFLOPS
  • AI cluster performance surpasses the NVIDIA GB300 NVL72

Comparison with Previous-Generation 910C​

MetricAscend 910CAscend 950Improvement
ArchitectureDa Vinci v4 (SIMD)Da Vinci v5 (SIMD + SIMT)Dual model
Process7nm-classN+2More advanced
HBMHBM2e (externally sourced)In-house HiBL / HiZQIndependent and controllable
HBM Capacity128 GB (dual chip)128-144 GBComparable
FP8/BF16 Compute780 TFLOPS BF161 PFLOPS FP8~2.5x
FP4 Support❌ Not supported✅ SupportedNew
TDP~310 W~400-500 W+29-61%
Programming ModelProprietary SIMDSIMD + SIMT + ASIC/GPGPU twinsOpen
In-House HBM❌ Externally sourced✅ In-houseMilestone

Vendor Information​

ParameterDetails
ManufacturerHuawei Technologies Co., Ltd. (HiSilicon)
Official Websitehttps://www.hiascend.com
CANNNhttps://www.hiascend.com/en/software/cann
Launch2026 H1 (mass production)
Pricing (estimated)950PR ~¥80,000-100,000, 950DT ~¥120,000-150,000

Use Cases​

  • ✅ Large-model training (950DT, 1 PFLOPS FP8)
  • ✅ Inference Prefill (950PR, video recommendation / real-time interaction)
  • ✅ Inference Decode (950DT, dialogue generation / text continuation)
  • ✅ Domestic AI clusters (CloudMatrix 384)
  • ✅ Government and enterprise AI infrastructure (in-house HBM, supply chain security)
  • ❌ CUDA ecosystem lock-in (migration to CANN / MindSpore required)
  • ❌ FP4 training (the 950 supports FP4 for inference; training remains primarily FP8/BF16)