当前位置:网站首页>The performance of arm's new CPU has been improved by 22%, up to 12 cores can be combined, and the GPU is first equipped with hardware optical tracking. Netizen: the gap with apple is growing
The performance of arm's new CPU has been improved by 22%, up to 12 cores can be combined, and the GPU is first equipped with hardware optical tracking. Netizen: the gap with apple is growing
2022-06-30 10:30:00 【QbitAl】
Abundant color From the Aofei temple
qubits | official account QbitAI
Arm v9 Here comes the new product of Architecture .
This is the Arm last year 3 A brand new product will be released in January v9 framework 、5 The first generation was released in January based on v9 The second update after product architecture .
This time, , A total of Two new CPU, Full turn 64 position That kind of , as well as Three new GPU, One of them supports... For the first time Hardware light chasing .

No accident , The fastest At the end of this year or early next year , These products can be used by Qualcomm 、 The latest processors of Samsung and MediaTek have been installed , For example, Xiaolong 8 G2~
that , How much has their performance and power consumption improved ?
Let's see... One by one .
Second generation v9 framework CPU Full turn 64 position
First look CPU.

Arm The first new product released this year CPU by Cortex-X3.
Cortex-X3 As a super core , Key performance improvements :
And the last generation running on the latest Android flagship Cortex-X2 comparison , Promoted 22%;
And the latest mainstream notebook processor (Intel Core i7 1260p 28w, Instead of Cortex-X1 or X2) comparison , Performance is expected to improve 34%.
Besides , With the same process Cortex-X2 comparison ,Cortex-X3 Of IPC Improved 11%.
Specific indicators ,Cortex-X3 The decoder instructions per cycle from 5 Upgrade to 6 individual , Out of order execution window from 288 Upgrade to 320 individual , Integers ALU Unit from 4 Upgrade to 6 individual ,L2 The cache capacity also increases from 512KB Promoted to 1MB.
Power consumption , From the table below ,Cortex-X3 In most cases than Cortex-X2 Less consumption , But as performance climbs ,Cortex-X3 It certainly doesn't look like “ Fuel efficient lights ”.

Section 2 NEW CPU by Cortex-A715, What this midcore pays attention to is the balance between performance and power consumption .
As Cortex-A710 Successor , It has the same power level and manufacturing process , Power savings 20%, Performance only improved 5%, Than before Cortex-A710 The promotion is much less (Cortex-A710 comparison Cortex-A78 Up 10%).
however ,Arm Express , This level can also Worth 2020 The heavyweight product of Cortex-X1 了 .
This bodes well for mid tier processor manufacturers , It can improve performance without increasing power consumption .
It is important to mention that ,Cortex-A715 and Cortex-X3 equally , Have given up aarch32 Instruction set , Only support 64 Yes. .
In addition to these two new CPU,Arm This time, we will also review the Cortex-A510 The small core has made a little update , The power consumption is reduced 5%, Most support 12 Nuclear architecture , And the only support 32 Bit compatible CPU.
in addition , This generation of Cortex The processor is for smart phones 、 Tablet, notebook and other platforms , according to Arm said , The new large and small core scheduling architecture can do at most 8+4+0, That is to say 12 nucleus .
GPU Support hardware optical tracing for the first time
Let's look at the new GPU.

The first one is brand new Immortalis-G715(Immortalis Yes “ immortal ” It means ).
Its biggest feature is that it supports hardware optical tracing on the mobile phone for the first time .
Arm call , This kind of GPU The core area of the shader used for light tracing on the 4%, Compared with light tracing based on software acceleration , Performance improved 300%.
Compared to the previous generation Mali-G710,Immortalis-G715 Performance improvement 15%, Energy saving 15%.
Besides , Its performance in machine learning has also been improved in two aspects , Performance on tasks such as computational enhancement and image enhancement has also improved .
There are two new models left GPU Respectively Mali-G715 and Mali-G615.
Mali-G715 For medium and high-end products , It's essentially a Immortalis-G715, But there is no light chasing , Only 7-9 Shader cores (Immortalis-G715 Yes 10-16 individual ).
Mali-G615 The number of shader cores continues to decrease , by 1-6 individual , But it still provides the advantages of variable coloring rate technology, performance and power consumption .

On the whole ,Arm The products released this year are not as heavy as when the architecture was replaced last year , but Arm Full turn 64 The determination of the bit instruction set has been formed .
Many netizens said , A bit disappointed ,Arm The gap with apple is getting bigger and bigger .

Others say :“Cortex-A710 after ,Arm Don't believe anything you say ”.

What do you think ?
Reference link :
https://www.androidauthority.com/arm-2023-cpu-gpu-smartphones-3178096/
— End —
「 qubits · viewpoint 」 Live registration
Why? “GPU+” It is more suitable for high performance and versatility AI Computational conundrum ? The public attention “GPU+” Where is the core competitiveness of architecture ? Use “GPU+” How powerful can the architecture chip be ?
6 month 30 On Thursday , Participate in the live broadcast , Answer for you ~

Focus on me here , Remember to mark the star ~
边栏推荐
- Right click to open CMD (command line)
- Basic MySQL operation commands of database
- “昆明城市咖啡地图”活动再度开启
- unable to convert expression into double array
- 移植完整版RT-Thread到GD32F4XX(详细)
- GeoffreyHinton:我的五十年深度学习生涯与研究心法
- What is the real performance of CK5, the king machine of CKB?
- mysql数据库基础:视图、变量
- 从0使用keil5软件仿真调试GD32F305
- Go -- maximum heap and minimum heap
猜你喜欢

Yixian e - commerce publie un rapport trimestriel: adhérer à la R & D et à l’investissement de la marque, réaliser un développement durable et de haute qualité

Detailed explanation of SolidWorks mass characteristics (inertia tensor, moment of inertia, inertia spindle)

Eth is not connected to the ore pool

Getting started with X86 - take over bare metal control

Xinguan has no lover, and all the people benefit from loving deeds to warm the world -- donation to the public welfare action of Shangqiu children's welfare home

Oracle creates a stored procedure successfully, but the compilation fails

Deployment of efficient and versatile clusters lvs+kept highly available clusters

Magnetic levitation 3D lamp

MySQL index, transaction and storage engine of database (2)

采坑:Didn‘t receive robot state (joint angles) with recent timestamp within 1 seconds.
随机推荐
Rider does not prompt after opening unity script
Node environment configuration
逸仙電商發布一季報:堅持研發及品牌投入,實現可持續高質量發展
mysql数据库基础:约束、标识列
MIT-6874-Deep Learning in the Life Sciences Week6
GD32 RT-Thread DAC驱动函数
MySQL index, transaction and storage engine of database (2)
Great Wall digital art digital collection platform releases the creation Badge
基于强化学习的股票量化交易Automated-Stock-Trading-Ensemble-Strategy
技能梳理[email protected]+adxl345+电机震动+串口输出
长城数艺数字藏品平台发布创世徽章
L'activité "Kunming City coffee map" a rouvert
CVPR 2022 | 清华&字节&京东提出BrT:用于视觉和点云3D目标检测的桥接Transformer
Eth is not connected to the ore pool
keras ‘InputLayer‘ object is not iterable
AttributeError: ‘Version‘ object has no attribute ‘major‘
机器人系统动力学——惯性参数
“昆明城市咖啡地图”活动再度开启
R语言plotly可视化:使用plotly可视化多分类模型的预测置信度、模型在2D网格中每个数据点预测的置信度、置信度定义为在某一点上最高分与其他类别得分之和之间的差值
最新SCI影响因子公布:国产期刊最高破46分!网友:算是把IF玩明白了