Hot Chips 33大會：AMD展示3D芯片封裝技術，IBM發佈7納米Telum處理器

{"type":"doc","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"美國時間 8 月 22 日，爲期 3 天的 Hot Chips 33 芯片大會正式召開。受新冠疫情影響，本次會議在線上播出，付費註冊才能收看。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"本月早些時候，Hot Chips 官方公佈了大會日程，從日程中可以看到，Intel、AMD、NVIDIA、IBM、臺積電等芯片巨頭都將登場，講解各家的最新芯片架構、封裝技術等。"}]},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","text":"AMD 展示 3D 芯片封裝技術"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"據外媒報道，在 Hot Chips 33 芯片大會上，AMD 進一步展示了"},{"type":"link","attrs":{"href":" https:\/\/allinfo.space\/2021\/08\/22\/hot-chips-33-amd-gives-further-insights-into-3d-packaging","title":"xxx","type":null},"content":[{"type":"text","text":"堆疊技術"}]},{"type":"text","text":"。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"事實上，在 6 月初的 Computex 大會上，AMD 就曾公佈過 3D V-Cache 技術的部分細節。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"AMD 在 Computex 上展示的"},{"type":"link","attrs":{"href":"https:\/\/www.hardwareluxx.de\/index.php\/news\/hardware\/prozessoren\/56917-hot-chips-33-amd-nennt-weitere-details-zum-3d-v-cache.html","title":"xxx","type":null},"content":[{"type":"text","text":" 3D V-Cache "}]},{"type":"text","text":"是在銳龍 9 5900X 上簡單地增加了一個額外的 L3 緩存。3D 堆疊排列使 AMD 能夠使用一種製造工藝，該工藝允許它的每一個計算芯片上都堆疊了64MB SRAM，可作爲額外的三級緩存使用，這樣加上處理器原本集成的 64MB，總的三級緩存容量就達到了 192MB。對比標準的銳龍 9 5900X 處理器，頻率都固定在 4GHz，3D V-Cache 緩存加入之後，遊戲性能平均提升了多達 15％。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"在 Hot Chips 33 上，AMD 透露了有關 3D V-Cache 的更多細節。根據介紹，AMD 實現了 3D V-Cache 的 TSV 之間的距離僅爲 9 μm。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.infoq.cn\/resource\/image\/9d\/1f\/9dca83462b62fd9a2cc13e26dyyc151f.jpeg","alt":null,"title":"","style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":"","fromPaste":false,"pastePass":false}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"此外，3D V-Cache 只是 AMD 3D 堆棧計劃的第一步。在下一步中，AMD 正在考慮將單個內核置於彼此之上，就像內核可以堆疊在處理器的非內核區域一樣。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"AMD 封裝高級研究員 Raja Swaminathan 表示，並非每個解決方案都適合所有產品。但是，未來無疑屬於模塊化設計與匹配協調封裝，以延續摩爾定律。業界對此已達成共識，知名廠商提供的各種解決方案就是明證。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.infoq.cn\/resource\/image\/41\/03\/41a2f4552f147feb5043d9d0306b9b03.jpeg","alt":null,"title":"","style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":"","fromPaste":false,"pastePass":false}},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","text":"IBM 發佈 Telum 處理器，主打 AI 加速"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"在 Hot Chips 33 芯片大會上， IBM 公佈了下一代 Z 系列企業級處理器 Telum。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"據 IBM 介紹，"},{"type":"link","attrs":{"href":"https:\/\/diy.zol.com.cn\/775\/7751732.html","title":"xxx","type":null},"content":[{"type":"text","text":"Telum "}]},{"type":"text","text":"採用三星 7nm 工藝製造，面積 530 平方毫米，集成多達 225 億個晶體管，擁有全新的分支預測、緩存、多芯片一致性互連，性能提升超過 40%。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.infoq.cn\/resource\/image\/74\/67\/74c20f0666dd7135591825a3a2175067.jpeg","alt":null,"title":"","style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":"","fromPaste":false,"pastePass":false}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"此外，Telum 還是 IBM 第一款結合 AI 加速推論功能的處理器芯片，其單芯片性能超過 6TFlops，內部矩陣陣列擁有 128 個單元，延遲超低且一致，支持 ML、RNN、CNN 各種模型，支持企業級內存虛擬化和保護，可通過硬件、固件更新進行拓展。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"Telum 可用於 AI 或深度學習模型推論。IBM 表示，金融企業可以將訓練完的 AI 或深度學習模型打包到採用 Telum 處理器的 IBM 系統上來部署與進行 AI 推論。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"另據瞭解，首款採用 Telum 處理器的 IBM 系統預計於 2022 年上半年推出。"}]}]}