broadsword
Brigadier
Haha, my Infiray P2 Pro, which arrived only four days ago, uses its self-developed ASIC chips.
big moment for China' AI inference chips
Horizon has now come out with Journey 6 SoC
Each chip supports 24 camera/Lidar/other sensors. Can handle full aspect NOA including all city driving scenarios
Customers include BYD, GAC, Bosch, Li Auto & VW/Cariad
560 TOPS & 350K DMIPS computation using BPU 3.0 (new generation AI processing)
He is wrong there btw. 6000 series use 12nm processChina's Premiere Chipmaker Accelerates to 7nm CPU Design Despite US Sanctions
Chinese CPU and GPU maker Loongson is beginning its transition to 7nm in 2024, which will apparently provide a 20% to 30% uplift in performance. Due to U.S. sanctions, the company's upcoming 7nm chips can only be fabbed in China — and although that might cost Loongson some performance and efficiency, it's still a win for China's native semiconductor industry...
@tokenanalyst you were right about SMIC advance process capacity. if this news is true
China's Premiere Chipmaker Accelerates to 7nm CPU Design Despite US Sanctions
Chinese CPU and GPU maker Loongson is beginning its transition to 7nm in 2024, which will apparently provide a 20% to 30% uplift in performance. Due to U.S. sanctions, the company's upcoming 7nm chips can only be fabbed in China — and although that might cost Loongson some performance and efficiency, it's still a win for China's native semiconductor industry...
@tokenanalyst you were right about SMIC advance process capacity. if this news is true
1|征程 6 核心架构
-BPU 3.0
-多核 BPU 算力 560TOPS
-CPU 350K+DMIPS(ARM Cortex A78 核)
-3D 可视化渲染GPU
-全功能 ASIL-D
-数据驱动优化(BEV Transformer 感知优化、 端到端的感知预测能力、多传感器融合能力)
-规则驱动优化(规划控制算法、城区复杂场景优化、ASIL-D 安全计算)
2| 征程 6 更像一个平台
-基础 L2 :ACC/LCC 等
-L2+ :高速 NOA 、HPA (6-7V)
-L2++ :高速 NOA 城区专线 AVP(7-11v)
-L2+++:全场景 NOA (9-11V 1-3L)
-从一体机到全场景高阶辅助驾驶能力是共用一套硬件架构、工具链以及软件栈。
-底软交付即量产
3|征程 6 旗舰
高集成度:CPU GPU BPU MCU 四芯合一
降低硬件架构的复杂度,提高性价比
电子电气从分布到集成预控,在软件部署难度上会有优化
4|高效 :优化 Transformer
-BPU 纳什架构
-优化并行浮点算力(支持 SIMT 架构 更自定义的算子)
-优化超越函数(算子的硬件加速能力)
-全新的存储系统(提供更好数据带宽,算的快的同时数据能够供的上)
-FPS :相比 CNN Transformer 有十倍提升
5|总线架构:
-自研总线:TB 级总线带宽(相比业内 2 倍的提升)
-全链路 ASIL-D (车规)
-延迟率 130ns(相比主流竞品降低40%)
6|画质优化
-图像处理带宽优化
-24bit动态
-最大支持 18MP(1800万像素前摄像头)
7|兼容多传感器方案
-支持24路摄像头(最高 1800万)
-AD Map
-GNSS
-激光雷达
-4D毫米波/毫米波
-超声波
-DMS /OMS /CMS/ ToF
8|集成 MCU
-集成 ASIL-D 全功能 MCU
-算力提升 3.5倍(主要解决过去规划算力不足问题)
-Main 域与 MCU 域强隔离(满足功能安全)
Intel had what everyone commonly refer to as 7nm for a long time. Intel 10nm is basically what everyone else called 7nm.And apart from TSMC and Samsung and now SMIC, no other company produce 7nm chips
So basically only China (inc Taiwan) and South Korea are able to manufacture 7nm chips. No US nor Japan nor EU
in term transistor density, yes Intel 10nm is comparable with 7nm ... but I don't think it is the same, otherwise why would Intel still call it 10nm ... thats not helping Intel imageIntel had what everyone commonly refer to as 7nm for a long time. Intel 10nm is basically what everyone else called 7nm.
transistor density is probably the only PPAcT metric we can determine without having actual chip to test performance, power consumption, and even cost. It IS the lone Apple to Apple benchmark of anyone’s given process capability.in term transistor density, yes Intel 10nm is comparable with 7nm ... but I don't think it is the same, otherwise why would Intel still call it 10nm ... thats not helping Intel image
and by the way, Intel already has a larger wafer capacity output of their Intel4 with EUV than SMIC capacity for N+2. So claiming they don’t even have 7nm capability seemed silly.
It’s long over due for this forum to stop the ridiculous narrative that Intel is somehow behind SMIC in capability. This kind of nonsense strips anyone promoting this false narrative of any credibility.