Pada si Iroyin
Ile-iṣẹAI Understanding finifini

MK ṣe ijabọ Nvidia ngbero atilẹyin RISC-V fun awọn olupin AI ati iṣafihan gbangba CUDA akọkọ

Iwe iroyin Iṣowo Maeil ṣe ijabọ pe Nvidia n faagun ilana CPU olupin AI-server lati pẹlu RISC-V, lẹgbẹẹ x86 ati Arm, ati pe SiFive yoo ṣafihan CUDA ni gbangba ti nṣiṣẹ lori RISC-V-orisun Sipiyu.

5 min readRead the linked source
Source-provided image accompanying MK reports Nvidia plans RISC-V support for AI servers and a first public CUDA demonstration
itọkasi orisunOrisun ti o gbasilẹ
Olutẹwe
mk.co.kr
Orisun ọna asopọ
mk.co.krhttps://www.mk.co.kr/en/it/12134515
Orisun iru
Orisun ti o sopọ mọ - ipo orisun akọkọ ko ti fi idi mulẹ.
AtokọLoye eyi ni iṣẹju 60

Bẹrẹ nibi

Awọn ofin bọtini

Iranti (Iranti Aṣoju)
Ọgangan ipamọ ti o jẹ aṣoju AI nlo kọja awọn igbesẹ tabi awọn akoko lati mu ilọsiwaju sii.
Lairi
Awọn akoko laarin a firanṣẹ ìbéèrè ati gbigba awọn awoṣe ká o wu.
Ṣe idanwo fun ara rẹAwọn awoṣe AI ti ṣalaye adanwo

Kini o ṣẹlẹ

Iwe iroyin Iṣowo Maeil ṣe ijabọ pe Nvidia n ngbaradi atilẹyin fun awọn CPUs ti o da lori RISC-V ninu pẹpẹ olupin AI rẹ, n ṣafikun faaji Sipiyu miiran lẹgbẹẹ x86 ati Arm. Ijabọ naa sọ pe Nvidia ti pese awọn ibeere fun CUDA ati NVLink Fusion lati ṣiṣẹ pẹlu awọn eto RISC-V ati pe SiFive ti ṣeto lati ṣafihan CUDA lori Sipiyu orisun RISC-V ni Awọn Chips Gbona 2026.

Iwe iroyin Iṣowo Maeil ṣe ijabọ pe Nvidia ṣafihan ilana kan lati faagun awọn faaji Sipiyu ti o ni atilẹyin nipasẹ pẹpẹ AI-server rẹ si RISC-V. Nkan naa gbe gbigbe naa lẹgbẹẹ atilẹyin Nvidia ti o wa tẹlẹ fun x86 ati Arm, eyiti o ṣapejuwe bi awọn faaji Sipiyu ti o jẹ oludari ni awọn olupin AI. RISC-V jẹ apejuwe ninu ijabọ bi ọna apẹrẹ Sipiyu ṣiṣi ti o le ṣee lo nipasẹ awọn ile-iṣẹ lọpọlọpọ.

Ijabọ naa ṣe ifitonileti naa si Nvidia ẹlẹrọ agba Frans Sistermans ni Gbona Chips 2026 apejọ semikondokito ni Ile-ẹkọ giga Stanford ni Oṣu Kẹjọ Ọjọ 23 ni akoko agbegbe. Gẹgẹbi MK, Arabinrin sọ pe RISC-V yoo di aṣayan olupin-CPU tuntun lẹhin x86 ati Arm ati pe Nvidia yoo faagun atilẹyin fun awọn ile-iṣẹ Sipiyu diẹ sii nipa lilo awọn iru ẹrọ AI rẹ. Awọn alaye wọnyi jẹ ijabọ nipasẹ MK ati pe wọn ko jẹrisi ni ominira nibi.

MK ṣe ijabọ pe Nvidia ti pese awọn ibeere fun CUDA, pẹpẹ idagbasoke GPU rẹ, ati NVLink Fusion, imọ-ẹrọ asopọ olupin AI-olupin, lati ṣiṣẹ lori awọn CPUs ti o da lori RISC-V. Ijabọ naa ṣe agbekalẹ eyi bi igbiyanju lati jẹ ki pẹpẹ AI gbooro ti Nvidia wa kọja awọn apẹrẹ ero isise ti o gbooro. Ko pese awọn alaye imọ-ẹrọ fun awọn ibeere tabi ṣe alaye boya wọn lo deede si gbogbo awọn imuse RISC-V.

Nkan naa tun ṣe ijabọ pe Nvidia n pọ si ifowosowopo pẹlu SiFive, ti n ṣapejuwe ile-iṣẹ bi alabaṣiṣẹpọ NVLink Fusion lẹgbẹẹ Arm, Intel, ati Fujitsu. MK sọ pe SiFive yoo ṣe ifihan gbangba ni Awọn Chips Gbona ti n ṣafihan CUDA nṣiṣẹ lori Sipiyu orisun RISC-V fun igba akọkọ. Orisun naa ko sọ ọjọ ifihan gangan, iṣeto eto, iṣẹ ṣiṣe, awọn abajade, tabi boya sọfitiwia naa yoo tu silẹ fun lilo gbogbogbo.

Awọn alaye orisun: mk.co.kr ↗

Kini idi ti o ṣe pataki

Gbigbe ti a royin le gbooro si sakani ti awọn CPUs ti o le ṣe so pọ pẹlu Nvidia's amayederun AI ati fun awọn apẹẹrẹ-chip olupin ni yiyan ayaworan diẹ sii. O tun ṣe idanwo boya sọfitiwia Nvidia ati ilolupo isọpọ le fa kọja x86 ti o jẹ gaba lori ati awọn iru ẹrọ Arm. Ijabọ naa ko ni ominira ṣe agbekalẹ iṣẹ ifihan, wiwa iṣowo, tabi nọmba awọn ọja olupin RISC-V iwaju.

Yiyan Sipiyu ṣe pataki nitori awọn olupin AI darapọ awọn ero isise pẹlu awọn accelerators, iranti, Nẹtiwọki, ati sọfitiwia. Atilẹyin RISC-V ti a royin Nvidia yoo koju Layer ero isise lakoko ti o tọju CUDA ati NVLink Fusion aringbungbun si eto naa. Ti a ba ṣe imuse ni gbooro, iyẹn le jẹ ki awọn apẹẹrẹ chirún diẹ sii kọ awọn CPUs ti a pinnu lati ṣiṣẹ pẹlu awọn amayederun AI ti o da lori Nvidia laisi gbigbekele iyasọtọ lori x86 tabi iwe-aṣẹ Arm ati awọn ilolupo.

Iṣe pataki da lori iye atilẹyin Nvidia ti o pese kọja ipele ibamu ipilẹ kan. Syeed olupin ti o wulo yoo nilo atilẹyin eto iṣẹ-ṣiṣe ti o gbẹkẹle, awọn akopọ, awọn ile-ikawe, awọn irinṣẹ n ṣatunṣe aṣiṣe, iṣakoso eto, ati ibaraenisepo pẹlu iyoku akopọ aarin data. MK ṣe ijabọ wiwa awọn ibeere fun CUDA ati NVLink Fusion, ṣugbọn ko ṣe agbekalẹ iwọn, idagbasoke, tabi pipe atilẹyin yẹn.

Fun Nvidia, ilolupo ilolupo Sipiyu ti o gbooro le ṣe alekun nọmba awọn ile-iṣẹ ti o ni anfani lati kopa ninu awọn eto ti a ṣe ni ayika awọn accelerators AI rẹ ati imọ-ẹrọ Nẹtiwọọki. Fun awọn ile-iṣẹ RISC-V, ibamu pẹlu CUDA le mu iraye si awọn aṣa olupin AI-ẹrọ nibiti sọfitiwia Nvidia wa pataki. Awọn iṣeeṣe wọnyẹn jẹ awọn ipa ti ilana ti a royin, kii ṣe awọn abajade ti a fihan nipasẹ orisun.

Idagbasoke naa le tun kan ifọkansi ati agbara idunadura ni awọn amayederun AI. Awọn aṣayan Sipiyu diẹ sii le ṣẹda yara fun agbegbe, amọja, tabi awọn olutọsọna olupin aṣa, ṣugbọn pẹpẹ Nvidia jẹ olutọju ẹnu-ọna ti o pọju ti iraye si CUDA tabi NVLink Fusion ni iṣakoso nipasẹ awọn ibeere imọ-ẹrọ tirẹ. Orisun naa ko pese alaye nipa idiyele, iwe-aṣẹ, iṣẹ ṣiṣe, awọn adehun alabara, tabi ipin ọja ti a nireti, nitorinaa ipa ifigagbaga naa ko ni idaniloju.

Interactive Mechanism

Ibaraẹnisọrọ Mechanism: Bii O Ṣe Nṣiṣẹ Lootọ

Ṣawari imọ-ẹrọ abẹlẹ lẹhin idagbasoke yii ni ibaraenisọrọ.

Document Size:128K tokens
Needle Placement Depth (Location in document):50% into text
Attention Context Buffer Map:
Target Fact (50%)
Equivalent Pages~320Standard book pages
Retrieval Accuracy99.9%Needle recall score
RAM / KV Cache5.1 GBMemory overhead
Prompt CachingActive~80% discount on reuse
Core takeaway: Million-token context windows allow querying whole codebases or legal archives in one prompt. However, KV cache memory scales with context length, making prompt caching crucial for real-time production.
Ibanisọrọ Erongba Ṣayẹwo+10 Points
AI Models Explained Quiz

Which component of an AI application is the machine-learning model itself?

Kini lati wo tókàn

Ojuami lẹsẹkẹsẹ lati wo ni boya iṣafihan ikede SiFive n ṣiṣẹ ni aṣeyọri ati iṣẹ wo ni o bo. Igba pipẹ, awọn alabara ati awọn ile-iṣẹ chirún yoo nilo ẹri nipa ibaramu, iṣẹ ṣiṣe, atilẹyin sọfitiwia, aabo, ipese, ati awọn akoko imuṣiṣẹ. Iṣe Nvidia ni ṣiṣe tabi idinku wiwọle si CUDA ati NVLink Fusion yoo tun ṣe apẹrẹ boya RISC-V di aṣayan ti o wulo fun awọn olupin AI.

Ojuami ijerisi akọkọ jẹ ifihan SiFive ti a ṣalaye nipasẹ MK. Awọn alafojusi yẹ ki o wa ero isise RISC-V kan pato ti a lo, awọn paati CUDA ti o ṣiṣẹ, idanwo fifuye iṣẹ, ati eyikeyi awọn abajade atunda ominira. Afihan kan yoo ṣe afihan iṣeeṣe imọ-ẹrọ, ṣugbọn kii yoo funrararẹ ṣe agbekalẹ imurasilẹ iṣelọpọ tabi iṣẹ ṣiṣe pẹlu x86- tabi awọn eto orisun-apa.

Ọrọ atẹle jẹ wiwa. Nvidia ati awọn alabaṣiṣẹpọ yoo nilo lati ṣalaye boya atilẹyin RISC-V jẹ idanwo, wa si awọn alabaṣiṣẹpọ ti a yan, tabi ṣetan fun awọn ọja olupin iṣowo. Orisun naa ko pese iṣeto idasilẹ, atokọ ọja, atokọ alabara, tabi ifaramo imuṣiṣẹ. O tun ko sọ boya awọn ohun elo CUDA ti o wa tẹlẹ le ṣiṣẹ laisi iyipada.

System-level performance will matter more than a successful software launch. Relevant evidence would include accelerator utilization, , power consumption, memory and I/O behavior, reliability, and performance on real AI workloads. None of those measurements is supplied in the MK report, so claims that RISC-V will improve efficiency or reduce costs would require additional evidence.

Finally, watch how Nvidia defines its relationships with SiFive, Arm, Intel, and Fujitsu and whether other RISC-V companies join the ecosystem. The key questions are whether support is open enough to encourage multiple suppliers, whether server manufacturers adopt the architecture, and whether customers see a reason to choose it. At present, MK reports an announced direction and a planned public demonstration, not a completed industry transition.

Awọn itọsọna ti o jọmọ & awọn ibeere

Awọn awoṣe AI ti ṣalayeAI IkẹkọỌjọ́ Iwájú AIṢe idanwo ohun ti o mọ — gbiyanju idanwo AI ọfẹ kanWa ọrọ AI kan ninu iwe-itumọ waTẹle olutọpa igbeowosile AI
Ṣe eyi wulo?