BTC $80,884.10 +1.62%
ETH $2,352.50 +1.89%
BNB $745.67 +1.31%
XRP $1.33 +2.41%
SOL $101.89 +1.90%
TRX $0.3308 +0.69%
DOGE $0.0796 +2.87%
ADA $0.2064 +1.32%
BCH $297.64 +6.21%
LINK $11.73 +3.52%
HYPE $81.40 +5.03%
AAVE $134.72 +1.78%
SUI $0.9378 +3.37%
XLM $0.1717 +2.47%
ZEC $1,313.31 +7.65%
AAPL $346.74 +0.95%
AMZN $254.37 +2.06%
GOOGL $324.87 +1.89%
MSFT $530.64 -0.17%
META $621.53 -1.92%
NVDA $218.67 +1.10%
TSLA $408.37 +3.89%
SNDK $1,359.67 +7.18%
INTC $95.02 +4.91%
SPCX $149.97 +1.50%
MU $966.25 +5.55%
AMD $568.44 +6.99%
BTC $80,884.10 +1.62%
ETH $2,352.50 +1.89%
BNB $745.67 +1.31%
XRP $1.33 +2.41%
SOL $101.89 +1.90%
TRX $0.3308 +0.69%
DOGE $0.0796 +2.87%
ADA $0.2064 +1.32%
BCH $297.64 +6.21%
LINK $11.73 +3.52%
HYPE $81.40 +5.03%
AAVE $134.72 +1.78%
SUI $0.9378 +3.37%
XLM $0.1717 +2.47%
ZEC $1,313.31 +7.65%
AAPL $346.74 +0.95%
AMZN $254.37 +2.06%
GOOGL $324.87 +1.89%
MSFT $530.64 -0.17%
META $621.53 -1.92%
NVDA $218.67 +1.10%
TSLA $408.37 +3.89%
SNDK $1,359.67 +7.18%
INTC $95.02 +4.91%
SPCX $149.97 +1.50%
MU $966.25 +5.55%
AMD $568.44 +6.99%

matrix

All
Article
Flash

Huawei Ascend 950PR's matrix computing speed is only 43% of H100

The semiconductor research organization SemiAnalysis has released the actual test results of Huawei's Ascend 950PR. The organization selected 1,055 sets of commonly used matrix calculations for large models. Huawei's own underlying program has allowed the 950PR to achieve 86% of its theoretical computing power, higher than NVIDIA's H100 at 75.5%; however, the actual computation speed is still only 43% of the H100, and only 21% compared to the relatively newer B300. This set of results has not yet utilized the newly ported operators from DeepSeek.DeepSeek is working on reducing data movement and waiting from the software side. In attention computation, after adjusting storage with FlashMLA, the two sets of data that originally needed to be read separately can now be retrieved at once; TileLang allows the chip to compute while preloading the next batch of data. SemiAnalysis measured that automatic scheduling has reduced the time for some small computing tasks by up to about 40%. DeepSeek has also ported DeepEP, which is responsible for MoE expert communication, but communication and computation will still compete for memory bandwidth, and Huawei plans to continue improving the firmware.Last year, DeepSeek attempted to train R2 using Huawei's Ascend chips, but even with Huawei engineers on-site for assistance, it could not be completed due to issues related to chip stability, inter-card communication, and software support. This year, both parties continue to collaborate, and by the end of September, DeepSeek focused on publicly releasing six sets of Ascend adaptation code. In October, it was reported that Huawei is adjusting the next generation of chip designs according to DeepSeek's model requirements. The 950PR tested this time is mainly aimed at inference input processing, while DeepSeek is preparing to use the 950DT for large-scale pre-training as early as the end of the year. Previously, both parties had conducted tests at the thousand-card level, but to enable tens of thousands of chips to complete a round of large model training continuously, more complex coordination and stability issues still need to be resolved.

HTX Genesis Hackathon opens in Hong Kong, HTX DAO ambassador Molly reveals the panoramic view of the ecological matrix

The HTX Genesis Hackathon, hosted by HTX DAO and B.AI, and co-initiated by TinTinLand and OpenCity, has opened in Hong Kong. The opening event was lively, with increasing frequency of exchanges and interaction density, creating a high-intensity dialogue atmosphere among developers, investors, and ecosystem participants.HTX DAO ambassador Molly shared on the theme "HTX DAO Ecosystem Development Plan: Launch of a Multi-Million Dollar Ecological Fund and Value Closed-Loop Strategy." She pointed out that the core vision of the HTX Genesis Hackathon is to build a free financial port based on the exchange's hinterland. To achieve this goal, five major ecological matrices covering the entire industry chain have been constructed, including: Huobi HTX as the core business engine and liquidity center; HTX DAO serving as the governance hub and centralized decision-making brain of the ecosystem; public chains and innovation layers providing payment network support through TRON, and building a solid DeFi infrastructure with JustLend, SunX, etc.; HTX Ventures focusing on industry frontiers, providing investment, incubation, and acceleration services for outstanding projects; and B.AI aiming at future trends, dedicated to creating financial infrastructure for the AI Agent era.Molly further emphasized that regarding future development, Huobi HTX has clearly proposed four core strategic keywords: "stability, transparency, institutionalization, AI empowerment," aiming to enhance the platform's long-term competitiveness in a complex market environment and provide ecosystem participants with more certain growth expectations.
app_icon
ChainCatcher Building the Web3 world with innovations.