·HSK 5

AMD (semiconductor company)
AMD
AMD
and / together with
yǔ
与
Intel (semiconductor company)
Intel
Intel
jointly / together
liánhé
联合
to release / to publish
fābù
发布
ACE (AI Compute Extensions)
ACE
ACE
white paper
báipíshū
白皮书
，
to claim / to assert
chēngchēng
声称
AI (artificial intelligence)
AI
AI
performance
xìngnéng
性能
to improve / to enhance
tíshēng
提升
16 times over
16 bèi
16倍
。

AMD (semiconductor company)

AMD

and / together with

yǔ

与

Intel (semiconductor company)

Intel

on (a date) / at

yú

于

April 15

4 yuè 15 rì

4月15日

jointly

liánhé

联合

to publish / to release

fābù

发布

ACE (AI Compute Extensions)

ACE

instruction set

zhǐlìng jí

指令集

white paper

báipíshū

白皮书

，

to state / claiming

chēng

称

new instructions

xīn zhǐlìng

新指令

can / are able to

kě jiāng

可将

x86 (processor architecture)

x86

processor / CPU

chǔlǐqì

处理器

structural particle

的

compute density

jìsuàn mìdù

计算密度

increase by 16 times

tíshēng 16 bèi

提升16倍

。

May 5, 2026 at 8:10 AM

AMD (semiconductor company)

AMD

and / together with

yǔ

与

Intel (semiconductor company)

Intel

on (a date) / at

yú

于

April 15

4 yuè 15 rì

4月15日

through / via

tōngguò

通过

x86 (processor architecture)

x86

Ecosystem Advisory Group

shēngtài xìtǒng gùwèn zǔ

生态系统顾问组

jointly

liánhé

联合

to release / to publish

fābù

发布

ACE (AI Compute Extensions — AI computing extension instructions)

ACE

white paper

báipíshū

白皮书

。

ACE (AI Compute Extensions)

ACE

full name / stands for

quánchēng

全称

AI Compute Extensions instruction set

réngōng zhìnéng jìsuàn kuòzhǎn zhǐlìng jí

人工智能计算扩展指令集

，

aimed at / designed to

zhìzài

旨在

to significantly improve

dàfú tíshēng

大幅提升

x86 (processor architecture)

x86

chip / processor

xīnpiàn

芯片

in / for

zài

在

matrix multiplication

jùzhèn chéngfǎ

矩阵乘法

computation / operations

yùnsuàn

运算

in terms of

shàng de

上的

throughput

tūntǔliàng

吞吐量

。

white paper

báipíshū

白皮书

to state / to claim

chēng

称

，

ACE

compared to / relative to

xiāngjìayú

相较于

existing / current

xiànyǒu de

现有的

AVX10 (Advanced Vector Extensions 10)

AVX10

instructions

zhǐlìng

指令

，

can / is able to

kě jiāng

可将

compute density

jìsuàn mìdù

计算密度

to improve / to increase

tíshēng

提升

16 times over

16 bèi

16倍

。

this is

zhè shì

这是

two (companies)

liǎng jiā

两家

long-term competitors

chángqī jìngzhēng

长期竞争

structural particle

的

chip

xīnpiàn

芯片

giant / major player

jùtóu

巨头

zài

在

instruction set

zhǐlìng jí

指令集

standardization

biāozhǔnhuà

标准化

field / domain

lǐngyù

领域

achieved / reached

dáchéng de

达成的

rare collaboration

hǎnjiàn hézuò

罕见合作

。

currently / at present

mùqián

目前

deep learning

shēndù xuéxí

深度学习

and

hé

和

high-performance computing

gāo xìngnéng jìsuàn

高性能计算

library (software)

kù

库

have begun to

yǐ kāishǐ

已开始

to integrate / to incorporate

zhěnghé

整合

ACE

acceleration primitives

jiāsù yuányǔ

加速原语

，

including

bāokuò

包括

NumPy (Python numerical computing library)

NumPy

and

yǔ

与

SciPy (Python scientific computing library)

SciPy

and so on / etc.

děng

等

Python (programming language)

Python

scientific computing library

kēxué jìsuàn kù

科学计算库

。

regarding hardware

jiù yìngjiàn

就硬件

in terms of / as for

éryán

而言

，

AMD

next-generation

xià yīdài

下一代

Zen 6 (AMD processor architecture)

Zen 6

architecture

jiàgòu

架构

EPYC (AMD server processor line)

EPYC

processor / CPU

chǔlǐqì

处理器

is expected to

yùjì

预计

in / during

yú

于

the second half of 2026

2026 nián xià bànnián

2026年下半年

to launch / to release

tuīchū

推出

。

analysts / observers

fēnxī rénshì

分析人士

to believe / to consider

rènwéi

认为

，

this collaboration

cǐ cì hézuò

此次合作

is conducive to / helps

yǒuzhùyú

有助于

x86

platform

píngtái

平台

zài

在

inference

tuīlǐ

推理

market

shìchǎng

市场

with

yǔ

与

ARM (processor architecture)

ARM

and

hé

和

dedicated accelerators

zhuānyòng jiāsùqì

专用加速器

to compete / to mount competition

zhǎnkāi jìngzhēng

展开竞争

。

New Words

liánhé

联合

jointly / together

fābù

发布

to release / to publish

báipíshū

白皮书

white paper

xìngnéng

性能

performance

tíshēng

提升

to improve / to enhance

zhǐlìng jí

指令集

instruction set

jiàgòu

架构

architecture

jìngzhēng

竞争

to compete / competition

hézuò

合作

collaboration / to cooperate

tuīlǐ

推理

inference

shìchǎng

市场

market

yùjì

预计

is expected to

News in English

Science & Technology

AMD and Intel Jointly Publish ACE White Paper Claiming 16x AI Performance Leap on x86

AMD and Intel jointly published the ACE instruction set white paper on April 15, claiming the new instructions can increase the AI compute density of x86 processors by 16 times.

AMD and Intel jointly released the ACE (AI Compute Extensions) white paper through the x86 Ecosystem Advisory Group on April 15, claiming to improve AI performance by 16 times.

ACE, which stands for "AI Compute Extensions instruction set," is designed to significantly improve x86 chip throughput in matrix multiplication operations. The white paper states that compared to the existing AVX10 instructions, ACE can improve compute density by 16 times.

This represents a rare collaboration between two long-term chip industry competitors in the field of instruction set standardization. Deep learning and high-performance computing libraries have begun integrating ACE acceleration primitives, including Python scientific computing libraries such as NumPy and SciPy.

Regarding hardware, AMD's next-generation Zen 6 architecture EPYC processors are expected to be released in the second half of 2026. Analysts believe this collaboration will help the x86 platform compete with ARM and dedicated accelerators in the AI inference market.

New Words

News in English

AMD and Intel Jointly Publish ACE White Paper Claiming 16x AI Performance Leap on x86

More in Science & Technology