| 1 | 1/1 | 返回列表 |
| 查看: 922 | 回復: 0 | |||
yexuqing木蟲之王 (文學泰斗)
太陽系系主任
|
[交流]
融合憶阻器和數(shù)字內(nèi)存計算處理助力高效邊緣計算
|
|
融合憶阻器和數(shù)字內(nèi)存計算處理助力高效邊緣計算 ▲ 作者:TAI-HAO WEN, JE-MIN HUNG, WEI-HSING HUANG, CHUAN-JIA JHANG, YUN-CHEN LO, HUNG-HSI HSU, ET AL. ▲ 鏈接: https://www.science.org/doi/10.1126/science.adf5538 ▲ 摘要: 人工智能(AI)邊緣設備更傾向于采用高容量非易失性內(nèi)存計算(CIM)來實現(xiàn)高能效和足夠準確的快速喚醒響應。大多數(shù)先前的工作要么依據(jù)基于憶阻器的CIM,但因其耐用性有限而遭受精度損失且不支持訓練;要么依據(jù)數(shù)字靜態(tài)隨機存取存儲器(SRAM)的CIM,但受限于大面積制造需求和易失性存儲。 研究組報道了一種使用憶阻器-SRAM CIM融合方案的AI邊緣處理器,可同時利用數(shù)字SRAM CIM的高精度和電阻式隨機存取存儲器憶阻器CIM的高能效和存儲密度。這也使自適應本地訓練能夠適應個性化特征和用戶環(huán)境。 該融合處理器實現(xiàn)了高CIM容量、短喚醒-響應延遲(392微秒)、高峰值能效(77.64 TOPS/W)和穩(wěn)健的精度(精度損失<0.5%)。這項工作表明,憶阻器技術已經(jīng)超越了實驗室開發(fā)階段,現(xiàn)已具備用于AI邊緣處理器的可制造性。 ▲ Abstract: Artificial intelligence (AI) edge devices prefer employing high-capacity nonvolatile compute-in-memory (CIM) to achieve high energy efficiency and rapid wakeup-to-response with sufficient accuracy. Most previous works are based on either memristor-based CIMs, which suffer from accuracy loss and do not support training as a result of limited endurance, or digital static random-access memory (SRAM)–based CIMs, which suffer from large area requirements and volatile storage. We report an AI edge processor that uses a memristor-SRAM CIM-fusion scheme to simultaneously exploit the high accuracy of the digital SRAM CIM and the high energy-efficiency and storage density of the resistive random-access memory memristor CIM. This also enables adaptive local training to accommodate personalized characterization and user environment. The fusion processor achieved high CIM capacity, short wakeup-to-response latency (392 microseconds), high peak energy efficiency (77.64 teraoperations per second per watt), and robust accuracy (<0.5% accuracy loss). This work demonstrates that memristor technology has moved beyond in-lab development stages and now has manufacturability for AI edge processors. |

| 1 | 1/1 | 返回列表 |
| 最具人氣熱帖推薦 [查看全部] | 作者 | 回/看 | 最后發(fā)表 | |
|---|---|---|---|---|
|
[考研] 0703 東華大學 理學化學均過 a 線,總分 281 求調(diào)劑 +6 | 香香 Lu 2026-03-03 | 11/550 |
|
|---|---|---|---|---|
|
[考研] 08工科求調(diào)劑 +3 | 隆LLL 2026-03-06 | 4/200 |
|
|
[考研] 中科大材料299求調(diào)劑 +7 | DAIjiayo 2026-03-05 | 10/500 |
|
|
[考研] 306求調(diào)劑 +6 | Bahati 2026-03-05 | 6/300 |
|
|
[考研] 304求調(diào)劑 +3 | 52hz~~ 2026-03-05 | 4/200 |
|
|
[考研] 282求調(diào)劑 +7 | 夕~日 2026-03-05 | 8/400 |
|
|
[考研] 085602 293分求調(diào)劑 +3 | SivanNano. 2026-03-05 | 3/150 |
|
|
[考研] 歡迎采礦、地質(zhì)、巖土、計算機、人工智能等專業(yè)的同學報考 +10 | pin8023 2026-02-28 | 13/650 |
|
|
[考研] 江西師范大學王昆-盧章輝課題組招化學方向碩士研究生兩名 +3 | 哈哈大俠 2026-02-28 | 3/150 |
|
|
[考研] 085600材料與化工 298 調(diào)劑 +9 | 小西笑嘻嘻 2026-03-03 | 9/450 |
|
|
[考研] 材料學碩080500復試調(diào)劑294 +3 | 四葉zjz 2026-03-04 | 3/150 |
|
|
[考研] 0857求調(diào)劑 +4 | CDDD1013 2026-03-04 | 4/200 |
|
|
[考博] 26申博 +5 | north, 2026-02-28 | 5/250 |
|
|
[考研] 一志愿西工大材料學,英一數(shù)二總分321分,求調(diào)劑。 +3 | zz05zz 2026-03-02 | 5/250 |
|
|
[考研] 化工270求調(diào)劑 +10 | 什么名字qwq 2026-03-02 | 10/500 |
|
|
[考研] 1 +7 | 黑!在干嘛 2026-02-28 | 8/400 |
|
|
[考研] 0856調(diào)劑 +10 | 劉夢微 2026-02-28 | 10/500 |
|
|
[考研] 一志愿中石油(華東)本科齊魯工業(yè)大學 +3 | 石能偉 2026-03-02 | 3/150 |
|
|
[考研] 0856化工專碩求調(diào)劑 +15 | 董boxing 2026-03-01 | 15/750 |
|
|
[考研] 279求調(diào)劑 +3 | dua1 2026-03-01 | 4/200 |
|