Halayen Samfur

Bukatar FPGAs ta AI accelerator katunan

Saboda sassauƙar su da ƙarfin kwamfuta mai sauri, FPGAs ana amfani da su sosai a cikin katunan haɓaka AI.Idan aka kwatanta da GPUs, FPGAs suna da fa'idodin ingantaccen kuzari;idan aka kwatanta da ASICs, FPGAs suna da mafi girman sassauci don dacewa da saurin juyin halittar hanyoyin sadarwa na AI da kuma ci gaba da sabunta abubuwan algorithms.Fa'ida daga fa'idar haɓakar haɓakar haƙƙin ɗan adam, buƙatar FPGAs don aikace-aikacen AI za ta ci gaba da haɓaka a nan gaba.Dangane da SemicoResearch, girman kasuwa na FPGAs a cikin yanayin aikace-aikacen AI zai ninka sau uku a cikin 19-23 don isa dalar Amurka biliyan 5.2.Idan aka kwatanta da kasuwar FPGA dala biliyan 8.3 a cikin '21, yuwuwar aikace-aikace a cikin AI ba za a iya yin la'akari da shi ba.

Mafi kyawun kasuwa ga FPGAs shine cibiyar bayanai

Cibiyoyin bayanai ɗaya ne daga cikin kasuwannin aikace-aikacen da suka kunno kai don kwakwalwan kwamfuta na FPGA, tare da ƙarancin latency + babban kayan aiki wanda ke shimfida mahimman ƙarfin FPGAs.Ana amfani da FPGAs na cibiyar bayanai galibi don haɓaka kayan masarufi kuma suna iya samun ci gaba mai mahimmanci yayin sarrafa algorithms na al'ada idan aka kwatanta da mafita na CPU na gargajiya: alal misali, aikin Microsoft Catapult ya yi amfani da FPGA maimakon mafita na CPU a cikin cibiyar bayanai don aiwatar da algorithms na al'ada na Bing sau 40 cikin sauri, tare da gagarumin hanzari effects.Sakamakon haka, an tura masu haɓaka FPGA akan sabobin a cikin Microsoft Azure, Amazon AWS, da AliCloud don haɓaka haɓakar lissafi tun daga 2016. A cikin yanayin cutar da ke haɓaka canjin dijital na duniya, abubuwan buƙatun cibiyar bayanai na gaba don aikin guntu zai ƙara haɓaka, kuma ƙarin cibiyoyin bayanai za su ɗauki mafitacin guntu na FPGA, wanda kuma zai ƙara ƙimar ƙimar kwakwalwan FPGA a cikin guntuwar cibiyar bayanai.

Babban tallace-tallace na tuki mai cin gashin kansa yana haɓaka buƙatun samar da FPGA mai yawa

Kamar yadda masana'antar kera ke ci gaba da haɓakawa daga ADAS zuwa cikakkiyar tuƙi mai cin gashin kai, dandamali na lissafin nau'ikan ta amfani da FPGAs na iya ɗaukar fashewar bayanan da ke haifar da karuwar yawan na'urori masu auna firikwensin, rage yawan lokacin amsawar tsarin gabaɗaya ta hanyar aiki tare da haɗa na'urori masu auna firikwensin da yawa, da haɓaka sassauci scalability, kunna scalability daga na'urori masu auna firikwensin zuwa masu kula da yanki, yayin da ke ba da damar sake tsarawa mai ƙarfi, rage farashin tsarin, da asara.Bugu da kari, FPGAs na iya samar da sassauƙa, mai rahusa, mafita mai inganci don buƙatun haɓaka da sauri na aikace-aikacen lantarki iri-iri.tsakiyar 20 ga Yuni, shugaban FPGA Xilinx yana da kusan guntuwar mota miliyan 70 da ake amfani da su a ADAS.

Samun AMD na yarjejeniyar Xilinx ya jinkirta zuwa kammala 22Q1

Bayan siyan Intel na FPGA Dragon II Altera a cikin 2015, AMD ta sanar a cikin Oktoba 2020 cewa tana shirin kashe dalar Amurka biliyan 35 (a cikin sigar hannun jari) don siyan manyan FPGA Xilinx a yunƙurin faɗaɗa TAM ta hanyar shiga kasuwar FPGA yayin haɓaka samfuran ta. layi don samar da cikakken tsarin ƙididdiga mai girma tare da na'urori masu sarrafa CPU na yanzu, katunan zane-zane na GPU, da ƙarar katunan kwamfuta.A cewar sabon labarai a ranar 31 ga Disamba 21, ana sa ran kammala sayan a cikin 22Q1, jinkiri daga jadawalin da aka sa ran farko, saboda ba a sami duk yarda ba.

A nan gaba, motsi ta hanyar 5G, ana sa ran FPGAs za su sami hauhawar girma da farashi, yayin da shugaban FPGA Xilinx kuma zai ci gaba da cin gajiyar sakin da ake nema a kasuwannin aikace-aikacen FPGA kamar AI, cibiyoyin bayanai, da tuƙi mai cin gashin kai. .