https://www.dnzjds.com/https://www.dnzjds.com/zhuangji/3265.html<div class="paragraph " id="yvPpVtTVlY9YtzbJQLLfCyK1e234s2" style="white-space: pre-wrap; margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-indent: 1.6rem; font-size: 0.8rem; color: rgb(34, 34, 34); line-height: 36.8px; word-break: break-all; overflow-wrap: normal; overflow: hidden; font-family: &quot;Microsoft Yahei&quot;; background-color: rgb(255, 255, 255);"><span style="font-size: 18px;">10月26日，技嘉推出面向AMD线程撕裂者PRO处理器的WRX90 AI TOP主板。该主板采用AMD WRX90芯片组与sTR5插槽，兼容线程撕裂者PRO 9000WX和7000WX处理器，搭载八通道DDR5 RDIMM插槽最高支持2TB ECC内存，最多可接驳九张GPU。主板配备主动散热模块，最高可将内存温度降低43℃，采用32+3+3+3相供电，搭载多条PCIe插槽、多类存储接口、双万兆网卡及远程管理芯片，支持技嘉AI TOP Utility本地AI计算工具，目前该主板的价格与上市时间均尚未公布。</span></div><div class="paragraph " id="0MFUeoad9dgzXlkHheoDvgs7Feqcfc" style="white-space: pre-wrap; margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-indent: 1.6rem; font-size: 0.8rem; color: rgb(34, 34, 34); line-height: 36.8px; word-break: break-all; overflow-wrap: normal; overflow: hidden; font-family: &quot;Microsoft Yahei&quot;; background-color: rgb(255, 255, 255);"><span style="font-size: 18px;">快科技10月26日消息，技嘉推出了一款新的WRX90 AI TOP主板，面向AMD线程撕裂者PRO处理器，支持最高2TB ECC DDR5内存和最多九张GPU。</span></div><div class="paragraph " id="hywanpjVvd8wpnJnUW4ym38puqYtUW" style="white-space: pre-wrap; margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-indent: 1.6rem; font-size: 0.8rem; color: rgb(34, 34, 34); line-height: 36.8px; word-break: break-all; overflow-wrap: normal; overflow: hidden; font-family: &quot;Microsoft Yahei&quot;; background-color: rgb(255, 255, 255);"><span style="font-size: 18px;">WRX90 AI TOP采用AMD WRX90芯片组和sTR5插槽，兼容线程撕裂者PRO 9000WX和7000WX处理器。</span></div><div class="paragraph " id="mnXR6StSB7ff27hl1gH3uxVs4rm5Wv" style="white-space: pre-wrap; margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-indent: 1.6rem; font-size: 0.8rem; color: rgb(34, 34, 34); line-height: 36.8px; word-break: break-all; overflow-wrap: normal; overflow: hidden; font-family: &quot;Microsoft Yahei&quot;; background-color: rgb(255, 255, 255);"><span style="font-size: 18px;">八条DDR5 RDIMM插槽运行在八通道模式下，支持最高2TB ECC内存，面向高负载本地AI运算等内存密集型应用。</span></div><div class="pic" style="text-wrap-mode: wrap; margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-align: center; position: relative; color: rgb(116, 116, 116); font-family: &quot;Microsoft Yahei&quot;; background-color: rgb(255, 255, 255);"><img class="img loaded" src="https://www.dnzjds.com/zb_users/upload/2026/10/20261026152348_39447.jpg" alt="技嘉推新款主板：风冷八通道支持2TB内存、还能塞进九块GPU！" data-was-processed="true" style="border: 0px; padding: 0px; max-width: 100%; margin: 20px auto; display: block; min-width: 100px; min-height: 102px; background: transparent; border-radius: 0px; position: relative;"/><div class="description" style="margin: 12px 0px 0px; padding: 0px; box-sizing: border-box; color: rgb(172, 172, 172); font-size: 0.6rem;"><span style="font-size: 18px;">技嘉推新款主板：风冷八通道支持2TB内存、还能塞进九块GPU！</span></div></div><div class="paragraph " id="9zzfU8f23qcEpdXYiynhuTGmVRqAip" style="white-space: pre-wrap; margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-indent: 1.6rem; font-size: 0.8rem; color: rgb(34, 34, 34); line-height: 36.8px; word-break: break-all; overflow-wrap: normal; overflow: hidden; font-family: &quot;Microsoft Yahei&quot;; background-color: rgb(255, 255, 255);"><span style="font-size: 18px;">八条内存插槽密集排列带来的散热压力自然不小，因此技嘉为这块板配备了主动散热，双风扇直接向DIMM插槽吹风，官方称可将内存温度降低最多43℃，VRM和芯片组也配有主动散热，保证长时间高负载下的稳定性。</span></div><div class="pic" style="text-wrap-mode: wrap; margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-align: center; position: relative; color: rgb(116, 116, 116); font-family: &quot;Microsoft Yahei&quot;; background-color: rgb(255, 255, 255);"><img class="img loaded" src="https://www.dnzjds.com/zb_users/upload/2026/10/20261026152348_43472.png" alt="技嘉推新款主板：风冷八通道支持2TB内存、还能塞进九块GPU！" data-was-processed="true" style="border: 0px; padding: 0px; max-width: 100%; margin: 20px auto; display: block; min-width: 100px; min-height: 102px; background: transparent; border-radius: 0px; position: relative;"/><div class="description" style="margin: 12px 0px 0px; padding: 0px; box-sizing: border-box; color: rgb(172, 172, 172); font-size: 0.6rem;"><span style="font-size: 18px;">技嘉推新款主板：风冷八通道支持2TB内存、还能塞进九块GPU！</span></div></div><div class="paragraph " id="QCFXgLPdXEHH6hykPhuTxTaNJcWyjE" style="white-space: pre-wrap; margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-indent: 1.6rem; font-size: 0.8rem; color: rgb(34, 34, 34); line-height: 36.8px; word-break: break-all; overflow-wrap: normal; overflow: hidden; font-family: &quot;Microsoft Yahei&quot;; background-color: rgb(255, 255, 255);"><span style="font-size: 18px;">扩展性方面，技嘉WRX90 AI TOP拥有七条全长PCIe 5.0插槽直连CPU，其中六条支持x16带宽，第七条为x8。</span></div><div class="pic" style="text-wrap-mode: wrap; margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-align: center; position: relative; color: rgb(116, 116, 116); font-family: &quot;Microsoft Yahei&quot;; background-color: rgb(255, 255, 255);"><img class="img loaded" src="https://www.dnzjds.com/zb_users/upload/2026/10/20261026152348_45327.png" alt="技嘉推新款主板：风冷八通道支持2TB内存、还能塞进九块GPU！" data-was-processed="true" style="border: 0px; padding: 0px; max-width: 100%; margin: 20px auto; display: block; min-width: 100px; min-height: 102px; background: transparent; border-radius: 0px; position: relative;"/><div class="description" style="margin: 12px 0px 0px; padding: 0px; box-sizing: border-box; color: rgb(172, 172, 172); font-size: 0.6rem;"><span style="font-size: 18px;">技嘉推新款主板：风冷八通道支持2TB内存、还能塞进九块GPU！</span></div></div><div class="paragraph " id="X9thdokCxtNTuavvaQu3KIDcedIr2c" style="white-space: pre-wrap; margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-indent: 1.6rem; font-size: 0.8rem; color: rgb(34, 34, 34); line-height: 36.8px; word-break: break-all; overflow-wrap: normal; overflow: hidden; font-family: &quot;Microsoft Yahei&quot;; background-color: rgb(255, 255, 255);"><span style="font-size: 18px;">供电采用32+3+3+3相设计，配备110A供电模块和14层PCB，还有两个MCIO 8i接口。存储方面提供两条PCIe 5.0 M.2插槽、一条PCIe 4.0 M.2插槽和四个SATA接口，其中一个MCIO接口与两条PCIe 5.0 M.2插槽共享带宽，不可同时使用。</span></div><div class="pic" style="text-wrap-mode: wrap; margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-align: center; position: relative; color: rgb(116, 116, 116); font-family: &quot;Microsoft Yahei&quot;; background-color: rgb(255, 255, 255);"><img class="img loaded" src="https://www.dnzjds.com/zb_users/upload/2026/10/20261026152348_16871.png" alt="技嘉推新款主板：风冷八通道支持2TB内存、还能塞进九块GPU！" data-was-processed="true" style="border: 0px; padding: 0px; max-width: 100%; margin: 20px auto; display: block; min-width: 100px; min-height: 102px; background: transparent; border-radius: 0px; position: relative;"/><div class="description" style="margin: 12px 0px 0px; padding: 0px; box-sizing: border-box; color: rgb(172, 172, 172); font-size: 0.6rem;"><span style="font-size: 18px;">技嘉推新款主板：风冷八通道支持2TB内存、还能塞进九块GPU！</span></div></div><div class="paragraph " id="JGg8PFzRYuGsD9wllDGqnuO7FTxjbc" style="white-space: pre-wrap; margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-indent: 1.6rem; font-size: 0.8rem; color: rgb(34, 34, 34); line-height: 36.8px; word-break: break-all; overflow-wrap: normal; overflow: hidden; font-family: &quot;Microsoft Yahei&quot;; background-color: rgb(255, 255, 255);"><span style="font-size: 18px;">其他配置包括双Intel X710万兆网卡、两个USB4 Type-C接口、六个后置USB 3.2 Gen 2接口，以及ASPEED AST2600远程管理芯片支持IPMI。</span></div><div class="pic" style="text-wrap-mode: wrap; margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-align: center; position: relative; color: rgb(116, 116, 116); font-family: &quot;Microsoft Yahei&quot;; background-color: rgb(255, 255, 255);"><img class="img loaded" src="https://www.dnzjds.com/zb_users/upload/2026/10/20261026152348_43236.jpg" alt="技嘉推新款主板：风冷八通道支持2TB内存、还能塞进九块GPU！" data-was-processed="true" style="border: 0px; padding: 0px; max-width: 100%; margin: 20px auto; display: block; min-width: 100px; min-height: 102px; background: transparent; border-radius: 0px; position: relative;"/><div class="description" style="margin: 12px 0px 0px; padding: 0px; box-sizing: border-box; color: rgb(172, 172, 172); font-size: 0.6rem;"><span style="font-size: 18px;">技嘉推新款主板：风冷八通道支持2TB内存、还能塞进九块GPU！</span></div></div><div class="paragraph " id="XB1luojD4HxjFshaW6KFPJtwOFmkkF" style="white-space: pre-wrap; margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-indent: 1.6rem; font-size: 0.8rem; color: rgb(34, 34, 34); line-height: 36.8px; word-break: break-all; overflow-wrap: normal; overflow: hidden; font-family: &quot;Microsoft Yahei&quot;; background-color: rgb(255, 255, 255);"><span style="font-size: 18px;">主板尺寸305×330mm，E-ATX板型，支持技嘉AI TOP Utility本地AI计算工具，价格和上市时间尚未公布。</span></div><div class="pic" style="text-wrap-mode: wrap; margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-align: center; position: relative; color: rgb(116, 116, 116); font-family: &quot;Microsoft Yahei&quot;; background-color: rgb(255, 255, 255);"><img class="img loaded" src="https://www.dnzjds.com/zb_users/upload/2026/10/20261026152348_45541.png" alt="技嘉推新款主板：风冷八通道支持2TB内存、还能塞进九块GPU！" data-was-processed="true" style="border: 0px; padding: 0px; max-width: 100%; margin: 20px auto; display: block; min-width: 100px; min-height: 102px; background: transparent; border-radius: 0px; position: relative;"/><div class="description" style="margin: 12px 0px 0px; padding: 0px; box-sizing: border-box; color: rgb(172, 172, 172); font-size: 0.6rem;"><span style="font-size: 18px;">技嘉推新款主板：风冷八通道支持2TB内存、还能塞进九块GPU！</span></div></div><div class="paragraph " id="AV5Jtz6Ed68My1Lg5cMw15lCLCzAXP" style="white-space: pre-wrap; margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-indent: 1.6rem; font-size: 0.8rem; color: rgb(34, 34, 34); line-height: 36.8px; word-break: break-all; overflow-wrap: normal; overflow: hidden; font-family: &quot;Microsoft Yahei&quot;; background-color: rgb(255, 255, 255);"><span style="font-size: 18px;">【本文结束】出处：快科技</span></div><p><br/></p>Mon, 26 Oct 2026 15:20:33 +0800https://www.dnzjds.com/zhuangji/3264.html<div class="paragraph " id="3n1EvtpOUcMECGUqbvqBK8Tk8TeeOX" style="margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-indent: 1.6rem; font-size: 0.8rem; color: rgb(34, 34, 34); line-height: 36.8px; word-break: break-all; overflow-wrap: normal; white-space: pre-wrap; overflow: hidden; font-family: &quot;Microsoft Yahei&quot;; background-color: rgb(255, 255, 255);"><span style="font-size: 18px;">快科技10月26日消息，OptiScaler模组社区近日取得重大技术突破，成功将AMD FSR 4.1.2神经超分辨率技术适配至英特尔锐炫显卡，并实现了对XMX硬件单元的深度调用。</span><span style="font-size: 18px; text-indent: 1.6rem;">根据社区公开的测试数据，在最新的英特尔锐炫B580显卡上，运行《暗夜旅程：远征33》时，FSR 4.1.2 Lite版本的超分计算耗时仅为0.59ms，这与英特尔原生XeSS技术的0.58ms几乎持平。</span><span style="font-size: 18px; text-indent: 1.6rem;">相比之下，未经过优化的FSR 4.1.2完整版耗时为0.85ms，而老款的4.0.2版本则需要0.87ms。</span></div><div class="pic" style="margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-align: center; position: relative; color: rgb(116, 116, 116); font-family: &quot;Microsoft Yahei&quot;; text-wrap-mode: wrap; background-color: rgb(255, 255, 255);"><img class="img loaded" src="https://www.dnzjds.com/zb_users/upload/2026/10/20261026152018_49236.png" alt="白嫖AMD黑科技！英特尔显卡也能跑FSR 4.1.2：画质还比自家的更好" data-was-processed="true" style="border: 0px; padding: 0px; max-width: 100%; margin: 20px auto; display: block; min-width: 100px; min-height: 102px; background: transparent; border-radius: 0px; position: relative;"/><div class="description" style="margin: 12px 0px 0px; padding: 0px; box-sizing: border-box; color: rgb(172, 172, 172); font-size: 0.6rem;"><span style="font-size: 18px;">白嫖AMD黑科技！英特尔显卡也能跑FSR 4.1.2：画质还比自家的更好</span></div></div><div class="paragraph " id="cwkrNe6VLBHyT4zlmYzEdtafhYenhL" style="margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-indent: 1.6rem; font-size: 0.8rem; color: rgb(34, 34, 34); line-height: 36.8px; word-break: break-all; overflow-wrap: normal; white-space: pre-wrap; overflow: hidden; font-family: &quot;Microsoft Yahei&quot;; background-color: rgb(255, 255, 255);"><span style="font-size: 18px;">技术原理层面，本次MOD修改了FSR 4的INT8动态链接库文件，强制其绕过通用的DP4a指令路径，转而直接利用英特尔显卡内部专为AI运算设计的XMX核心。</span><span style="font-size: 18px; text-indent: 1.6rem;">这种借鸡生蛋的做法，让AMD的神经网络模型在对手的硬件上跑出了原生级别的执行效率。</span><span style="font-size: 18px; text-indent: 1.6rem;">在针对锐炫A770显卡的《巫师3：狂猎》重制版测试中，FSR 4.1.2 Lite在质量模式下达到了101.5 FPS的平均帧率，甚至微弱领先于原生XeSS的101.1 FPS。</span></div><div class="paragraph " id="ueRp04G21neVsQP4uJcqfggANCVk3h" style="margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-indent: 1.6rem; font-size: 0.8rem; color: rgb(34, 34, 34); line-height: 36.8px; word-break: break-all; overflow-wrap: normal; white-space: pre-wrap; overflow: hidden; font-family: &quot;Microsoft Yahei&quot;; background-color: rgb(255, 255, 255);"><span style="font-size: 18px;">这一结果证明，英特尔的硬件底座完全有能力高效驱动AMD的最新AI模型。</span><span style="font-size: 18px; text-indent: 1.6rem;">甚至不少社区测试者反馈，FSR 4.1.2在画质重建方面表现出了比原生XeSS更好的稳定性，尤其是在边缘抗锯齿和细节保留上，即使是精简后的Lite模型，其画面感官也并未出现明显的缩水。</span><span style="font-size: 18px; text-indent: 1.6rem;">不过需要注意的是，目前这项技术仅支持DirectX 12 API环境。由于AMD与英特尔官方均未对该配置提供正式支持，未来的显卡驱动更新可能会对该MOD的兼容性产生影响，追求稳定的玩家仍需观望。</span></div><div class="pic" style="margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-align: center; position: relative; color: rgb(116, 116, 116); font-family: &quot;Microsoft Yahei&quot;; text-wrap-mode: wrap; background-color: rgb(255, 255, 255);"><img class="img loaded" src="https://www.dnzjds.com/zb_users/upload/2026/10/20261026152018_90216.jpg" alt="白嫖AMD黑科技！英特尔显卡也能跑FSR 4.1.2：画质还比自家的更好" data-was-processed="true" style="border: 0px; padding: 0px; max-width: 100%; margin: 20px auto; display: block; min-width: 100px; min-height: 102px; background: transparent; border-radius: 0px; position: relative;"/><div class="description" style="margin: 12px 0px 0px; padding: 0px; box-sizing: border-box; color: rgb(172, 172, 172); font-size: 0.6rem;"><span style="font-size: 18px;">白嫖AMD黑科技！英特尔显卡也能跑FSR 4.1.2：画质还比自家的更好</span></div></div><div class="paragraph " id="TXAmQEZKQLFVSSypeCzdWTS8kUjMcy" style="margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-indent: 1.6rem; font-size: 0.8rem; color: rgb(34, 34, 34); line-height: 36.8px; word-break: break-all; overflow-wrap: normal; white-space: pre-wrap; overflow: hidden; font-family: &quot;Microsoft Yahei&quot;; background-color: rgb(255, 255, 255);"><span style="font-size: 18px;">【本文结束】出处：快科技</span></div>Mon, 26 Oct 2026 15:01:36 +0800https://www.dnzjds.com/zhuangji/3263.html<div class="paragraph " id="8GAcVbz8rAhGIxhVNCpinzgBB1ckM1" style="margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-indent: 1.6rem; font-size: 0.8rem; color: rgb(34, 34, 34); line-height: 36.8px; word-break: break-all; overflow-wrap: normal; white-space: pre-wrap; overflow: hidden; font-family: &quot;Microsoft Yahei&quot;; background-color: rgb(255, 255, 255);"><span style="font-size: 18px;">快科技10月26日消息，知名硬件媒体Gamers Nexus近日发布的一项对比测试显示，英伟达上一代神卡GTX 1060在特定游戏中的性能表现竟然反超了PS5游戏主机，领先幅度超过30%。</span><span style="font-size: 18px; text-indent: 1.6rem;">在本次测试中，GN选用了一台搭载四核处理器锐龙3 3300X的入门级PC，分别配备了GTX 1000系列的几款老显卡，所有游戏均关闭了光线追踪功能，并调整PC画质选项以匹配PS5的预设参数。</span></div><div class="pic" style="margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-align: center; position: relative; color: rgb(116, 116, 116); font-family: &quot;Microsoft Yahei&quot;; text-wrap-mode: wrap; background-color: rgb(255, 255, 255);"><img class="img loaded" src="https://www.dnzjds.com/zb_users/upload/2026/10/20261026145734_19572.jpg" alt="10系老将反杀次世代！GTX 1060实测PC平均142帧：领先PS5主机30%" data-was-processed="true" style="border: 0px; padding: 0px; max-width: 100%; margin: 20px auto; display: block; min-width: 100px; min-height: 102px; background: transparent; border-radius: 0px; position: relative;"/><div class="description" style="margin: 12px 0px 0px; padding: 0px; box-sizing: border-box; color: rgb(172, 172, 172); font-size: 0.6rem;"><span style="font-size: 18px;">10系老将反杀次世代！GTX 1060实测PC平均142帧：领先PS5主机30%</span></div></div><div class="paragraph " id="fiYpMaYJGPUqZlKJos6fz8AjvW3DQf" style="margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-indent: 1.6rem; font-size: 0.8rem; color: rgb(34, 34, 34); line-height: 36.8px; word-break: break-all; overflow-wrap: normal; white-space: pre-wrap; overflow: hidden; font-family: &quot;Microsoft Yahei&quot;; background-color: rgb(255, 255, 255);"><span style="font-size: 18px;">实测数据显示，在《鬼泣5》的120Hz高帧率模式测试中，配备GTX 1060 6GB显卡的PC跑出了142.1 FPS的平均帧率。</span><span style="font-size: 18px; text-indent: 1.6rem;">作为对比，PS5在相同场景下的平均帧率仅为108.7 FPS，PC端的领先幅度达到了30.7%。</span><span style="font-size: 18px; text-indent: 1.6rem;">而且相比平均帧率的提升，两者在帧率稳定性上的差距更为惊人：数据显示，PC端的1%最低帧高达117.1 FPS，而PS5在同一指标下仅录得39 FPS。</span></div><div class="paragraph " id="4Dl7vW32JMLMbyKGwNAuHVOSC3ruob" style="margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-indent: 1.6rem; font-size: 0.8rem; color: rgb(34, 34, 34); line-height: 36.8px; word-break: break-all; overflow-wrap: normal; white-space: pre-wrap; overflow: hidden; font-family: &quot;Microsoft Yahei&quot;; background-color: rgb(255, 255, 255);"><span style="font-size: 18px;">这意味着在激烈的战斗场景中，PS5会出现肉眼可见的卡顿，而老旧的GTX 1060却能保持极高的画面平滑度。</span><span style="font-size: 18px; text-indent: 1.6rem;">不过这一结果并不能简单推广至所有现代游戏，测试者指出，要实现在其他大作中对标PS5，显卡档次需要略微上移。</span><span style="font-size: 18px; text-indent: 1.6rem;">例如在《无主之地3》中需要GTX 1070 Ti 8GB，而在《尘埃5》中则需要GTX 1080 8GB才能达到与主机类似的帧率水平。</span></div><div class="pic" style="margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-align: center; position: relative; color: rgb(116, 116, 116); font-family: &quot;Microsoft Yahei&quot;; text-wrap-mode: wrap; background-color: rgb(255, 255, 255);"><img class="img loaded" src="https://www.dnzjds.com/zb_users/upload/2026/10/20261026145735_70201.png" alt="10系老将反杀次世代！GTX 1060实测PC平均142帧：领先PS5主机30%" data-was-processed="true" style="border: 0px; padding: 0px; max-width: 100%; margin: 20px auto; display: block; min-width: 100px; min-height: 102px; background: transparent; border-radius: 0px; position: relative;"/><div class="description" style="margin: 12px 0px 0px; padding: 0px; box-sizing: border-box; color: rgb(172, 172, 172); font-size: 0.6rem;"><span style="font-size: 18px;">10系老将反杀次世代！GTX 1060实测PC平均142帧：领先PS5主机30%</span></div></div><div class="paragraph " id="Qw8khe9aMDyDsfAgzZpsA2TjXfsjSg" style="margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-indent: 1.6rem; font-size: 0.8rem; color: rgb(34, 34, 34); line-height: 36.8px; word-break: break-all; overflow-wrap: normal; white-space: pre-wrap; overflow: hidden; font-family: &quot;Microsoft Yahei&quot;; background-color: rgb(255, 255, 255);"><span style="font-size: 18px;">【本文结束】出处：快科技</span></div><p><br/></p>Mon, 26 Oct 2026 14:48:35 +0800https://www.dnzjds.com/zhuangji/3262.html<div class="paragraph " id="vPqA0jQTVFJrXOVzAp3ms35pgkhhjI" style="margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-indent: 1.6rem; font-size: 0.8rem; color: rgb(34, 34, 34); line-height: 36.8px; word-break: break-all; overflow-wrap: normal; white-space: pre-wrap; overflow: hidden; font-family: &quot;Microsoft Yahei&quot;; background-color: rgb(255, 255, 255);"><span style="font-size: 18px;">快科技10月24日消息，开发者TheHiepDang近日正式发布了《血源诅咒》非官方PC移植版的最新分支项目BloodborneX 1.0.0版本。</span><span style="font-size: 18px; text-indent: 1.6rem;">本次更新最核心的技术突破在于原生集成了英伟达最新的DLSS 5神经渲染技术，这标志着这款原本运行在旧世代主机上的游戏，在PC平台上正式获得了人工智能重构画面的能力。</span></div><div class="pic" style="margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-align: center; position: relative; color: rgb(116, 116, 116); font-family: &quot;Microsoft Yahei&quot;; text-wrap-mode: wrap; background-color: rgb(255, 255, 255);"><img class="img loaded" src="https://www.dnzjds.com/zb_users/upload/2026/10/20261024212705_40062.png" alt="原生支持大力水手5！《血源诅咒》PC移植版大升级：移动版3070稳47帧" data-was-processed="true" style="border: 0px; padding: 0px; max-width: 100%; margin: 20px auto; display: block; min-width: 100px; min-height: 102px; background: transparent; border-radius: 0px; position: relative;"/><div class="description" style="margin: 12px 0px 0px; padding: 0px; box-sizing: border-box; color: rgb(172, 172, 172); font-size: 0.6rem;"><span style="font-size: 18px;">原生支持大力水手5！《血源诅咒》PC移植版大升级：移动版3070稳47帧</span></div></div><div class="paragraph " id="A55clnb8Wuo0T4yKXRLNh7GhJdc2H8" style="margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-indent: 1.6rem; font-size: 0.8rem; color: rgb(34, 34, 34); line-height: 36.8px; word-break: break-all; overflow-wrap: normal; white-space: pre-wrap; overflow: hidden; font-family: &quot;Microsoft Yahei&quot;; background-color: rgb(255, 255, 255);"><span style="font-size: 18px;">据开发者介绍，该项目基于Vulkan图形API运行，使用修改版的Lantern Engine 1.0.0库，面向64位Windows系统。</span><span style="font-size: 18px; text-indent: 1.6rem;">想要跑起来并正确初始化神经网络算法，用户还需要手动把nvgx_dlssnr.dll动态库放进可执行文件所在的目录。</span><span style="font-size: 18px; text-indent: 1.6rem;">修改版的图形菜单提供了相当细的调节粒度：除了选择NVIDIA DLSS超分和开关帧生成之外，还内置了以1280x720为基础渲染分辨率的缩放预设、带纹理mipmap调节的RCAS锐化滤镜，以及亚像素抖动抑制。</span></div><div class="paragraph " id="P1g8zHLtFzGjpPO8fH3VXN1gRoDgev" style="margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-indent: 1.6rem; font-size: 0.8rem; color: rgb(34, 34, 34); line-height: 36.8px; word-break: break-all; overflow-wrap: normal; white-space: pre-wrap; overflow: hidden; font-family: &quot;Microsoft Yahei&quot;; background-color: rgb(255, 255, 255);"><span style="font-size: 18px;">DLSS神经渲染部分可调的选项更多，包括输入重建分辨率、整体效果强度、渲染风格、神经网络架构预设、额外帧处理次数和局部场景几何保留等。</span><span style="font-size: 18px; text-indent: 1.6rem;">实测部分，作者用移动版GeForce RTX 3070 Laptop GPU（8GB显存）进行了演示。在开阔场景的游戏画面中，项目跑出了45到47帧的稳定帧率，帧生成时间在21.4至22.4毫秒之间，全程没有出现明显的性能骤降。</span></div><div class="paragraph " id="QkgmKVM0s0givMKfi2JOarWnLFdV4y" style="margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-indent: 1.6rem; font-size: 0.8rem; color: rgb(34, 34, 34); line-height: 36.8px; word-break: break-all; overflow-wrap: normal; white-space: pre-wrap; overflow: hidden; font-family: &quot;Microsoft Yahei&quot;; background-color: rgb(255, 255, 255);"><span style="font-size: 18px;">目前作者已在GitHub的公开仓库放出176MB的LanternEngine-nr-1.0.0-win64.zip整合包，并附上完整源代码。</span></div><div class="pic" style="margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-align: center; position: relative; color: rgb(116, 116, 116); font-family: &quot;Microsoft Yahei&quot;; text-wrap-mode: wrap; background-color: rgb(255, 255, 255);"><img class="img loaded" src="https://www.dnzjds.com/zb_users/upload/2026/10/20261024212705_54454.png" alt="原生支持大力水手5！《血源诅咒》PC移植版大升级：移动版3070稳47帧" data-was-processed="true" style="border: 0px; padding: 0px; max-width: 100%; margin: 20px auto; display: block; min-width: 100px; min-height: 102px; background: transparent; border-radius: 0px; position: relative;"/><div class="description" style="margin: 12px 0px 0px; padding: 0px; box-sizing: border-box; color: rgb(172, 172, 172); font-size: 0.6rem;"><span style="font-size: 18px;">原生支持大力水手5！《血源诅咒》PC移植版大升级：移动版3070稳47帧</span></div></div><div class="paragraph " id="G961QEk8YXjywHVo3U1cROOHvYPMId" style="margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-indent: 1.6rem; font-size: 0.8rem; color: rgb(34, 34, 34); line-height: 36.8px; word-break: break-all; overflow-wrap: normal; white-space: pre-wrap; overflow: hidden; font-family: &quot;Microsoft Yahei&quot;; background-color: rgb(255, 255, 255);"><span style="font-size: 18px;">【本文结束】出处：快科技</span></div>Sat, 24 Oct 2026 21:24:24 +0800https://www.dnzjds.com/zhuangji/3261.html<div class="paragraph " id="Lc1LQAaRX0vEOaWNZWeYAg9fzYB5xo" style="margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-indent: 1.6rem; font-size: 0.8rem; color: rgb(34, 34, 34); line-height: 36.8px; word-break: break-all; overflow-wrap: normal; white-space: pre-wrap; overflow: hidden; font-family: &quot;Microsoft Yahei&quot;; background-color: rgb(255, 255, 255);"><span style="font-size: 18px;">快科技10月24日消息，随着Agent智能体的爆发，CPU的重要性日益凸显，单核及多核性能都有更高要求，国产新一代C86处理器也要做到256核了，看齐Zen6水平。</span><span style="font-size: 18px; text-indent: 1.6rem;">这个处理器就是海光研发的第五代处理器——海光五号C86-5G，以及配套的AI加速器DCU深算四号，该公司日前在财报会议上回应称两款产品进展顺利。</span><span style="font-size: 18px; text-indent: 1.6rem;">海光表示，新一代通用CPU和DCU研发按既定路线推进，整体进度符合预期，算力密度、互联带宽等指标将升级。公司坚持“销售一代，验证一代，研发一代”的产品研发策略，具体发布时间和量产节点留待后续披露。</span></div><div class="paragraph " id="SatsaN7z5vVGntc8KhhFvsicHQFk0j" style="margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-indent: 1.6rem; font-size: 0.8rem; color: rgb(34, 34, 34); line-height: 36.8px; word-break: break-all; overflow-wrap: normal; white-space: pre-wrap; overflow: hidden; font-family: &quot;Microsoft Yahei&quot;; background-color: rgb(255, 255, 255);"><span style="font-size: 18px;">今年4月份财报会议上，海光公司提到公司自2016年获得AMD Zen 1架构授权后，已实现独立技术迭代，CPU产品线从第一代7100系列发展至今，已迭代至海光四号。</span><span style="font-size: 18px; text-indent: 1.6rem;">目前高端处理器7400系列已在各行业大规模应用，相比上一代产品，其多核性能提升52%，单核性能提升超过135%，最高达到128核256线程。</span></div><div class="paragraph " id="p63ghTX8lwBGPL3YNAbzjGUMd1TBMK" style="margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-indent: 1.6rem; font-size: 0.8rem; color: rgb(34, 34, 34); line-height: 36.8px; word-break: break-all; overflow-wrap: normal; white-space: pre-wrap; overflow: hidden; font-family: &quot;Microsoft Yahei&quot;; background-color: rgb(255, 255, 255);"><span style="font-size: 18px;">规划显示，海光公司2027年后将推出256核的新产品，性能将与海外最领先的x86服务器看齐，而且海光五号产品将支持自研的HSL互联协议。</span><span style="font-size: 18px; text-indent: 1.6rem;">256核的设计放在2027年依然是全球领先的水平，AMD最新的Zen6架构EPYC处理器Venice这一代也就是256核处理器，Intel最高能做到288核了，但CPU核心是E核架构，跟常规的全大核处理器不是一个路线。</span></div><div class="paragraph " id="mcrHv5wmWMpIsyJNKL4v9gXsmPv8KK" style="margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-indent: 1.6rem; font-size: 0.8rem; color: rgb(34, 34, 34); line-height: 36.8px; word-break: break-all; overflow-wrap: normal; white-space: pre-wrap; overflow: hidden; font-family: &quot;Microsoft Yahei&quot;; background-color: rgb(255, 255, 255);"><span style="font-size: 18px;">不过海光公司并没有具体公布海光五号处理器的规格，国海证券在7月份的研报中提到，海光五号C86-5G芯片是公司自研架构的增强版本，同时还支持AVX-512指令集。</span><span style="font-size: 18px; text-indent: 1.6rem;">只是这个报告对C86-5G 处理器的核心数描述不同，称其配备128个物理核心，支持四路同步多线程（SMT）技术，即每个核心可以处理四个线程，从而在128个核心的基础上实现了512线程，可大幅提升多任务处理能力。</span></div><div class="paragraph " id="HxZ8fEwSH2NWbi5oXMWlmemBsuJPLL" style="margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-indent: 1.6rem; font-size: 0.8rem; color: rgb(34, 34, 34); line-height: 36.8px; word-break: break-all; overflow-wrap: normal; white-space: pre-wrap; overflow: hidden; font-family: &quot;Microsoft Yahei&quot;; background-color: rgb(255, 255, 255);"><span style="font-size: 18px;">此外，C86-5G 微架构也进行了升级，每周期指令数（IPC）提升超过17%。</span></div><div class="paragraph " id="GPYAeLJQcpG4KEqpaEmUEAHYEVYD1q" style="margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-indent: 1.6rem; font-size: 0.8rem; color: rgb(34, 34, 34); line-height: 36.8px; word-break: break-all; overflow-wrap: normal; white-space: pre-wrap; overflow: hidden; font-family: &quot;Microsoft Yahei&quot;; background-color: rgb(255, 255, 255);"><span style="font-size: 18px;">内存方面，C86-5G 提供16通道 DDR5-5600接口，而当前的C86-4G配置为12通道 DDR5-4800。</span><span style="font-size: 18px; text-indent: 1.6rem;">连接性能方面，C86-5G支持 Compute Express Link 2.0（CXL2.0）标准，与AMD EPYC 9005 和 Intel第五代Xeon处理器保持同等水平。</span><span style="font-size: 18px; text-indent: 1.6rem;">PCIe5.0信道数量暂未披露，但C86-4G 已提供128条PCIe 5.0 信道，与AMD第四代EPYC 7004（Genoa）芯片相当。</span></div><div class="pic" style="margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-align: center; position: relative; color: rgb(116, 116, 116); font-family: &quot;Microsoft Yahei&quot;; text-wrap-mode: wrap; background-color: rgb(255, 255, 255);"><img class="img loaded" src="https://www.dnzjds.com/zb_users/upload/2026/10/20261024205212_23428.jpg" alt="国产C86处理器看齐Zen6：最多256核、IPC性能大涨17%" data-was-processed="true" style="border: 0px; padding: 0px; max-width: 100%; margin: 20px auto; display: block; min-width: 100px; min-height: 102px; background: transparent; border-radius: 0px; position: relative;"/><div class="description" style="margin: 12px 0px 0px; padding: 0px; box-sizing: border-box; color: rgb(172, 172, 172); font-size: 0.6rem;"><span style="font-size: 18px;">国产C86处理器看齐Zen6：最多256核、IPC性能大涨17%</span></div></div><div class="paragraph " id="GuRdZbzlpf6gPFGyzkR3SjmA152ToT" style="margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-indent: 1.6rem; font-size: 0.8rem; color: rgb(34, 34, 34); line-height: 36.8px; word-break: break-all; overflow-wrap: normal; white-space: pre-wrap; overflow: hidden; font-family: &quot;Microsoft Yahei&quot;; background-color: rgb(255, 255, 255);"><span style="font-size: 18px;">网上早前流传过海光的CPU路线图，显示海光五号最多128核512线程，但图中的海光四号说是最多64核，实际现在也做到了128核，因此下一代的海光五号做到256核依然是很有可能的。</span></div><div class="pic" style="margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-align: center; position: relative; color: rgb(116, 116, 116); font-family: &quot;Microsoft Yahei&quot;; text-wrap-mode: wrap; background-color: rgb(255, 255, 255);"><img class="img loaded" src="https://www.dnzjds.com/zb_users/upload/2026/10/20261024205212_12275.png" alt="国产C86处理器看齐Zen6：最多256核、IPC性能大涨17%" data-was-processed="true" style="border: 0px; padding: 0px; max-width: 100%; margin: 20px auto; display: block; min-width: 100px; min-height: 102px; background: transparent; border-radius: 0px; position: relative;"/><div class="description" style="margin: 12px 0px 0px; padding: 0px; box-sizing: border-box; color: rgb(172, 172, 172); font-size: 0.6rem;"><span style="font-size: 18px;">国产C86处理器看齐Zen6：最多256核、IPC性能大涨17%</span></div></div><div class="paragraph " id="WjIQAlfx3DIRfzexwyh8KUxhR5Ld3g" style="margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-indent: 1.6rem; font-size: 0.8rem; color: rgb(34, 34, 34); line-height: 36.8px; word-break: break-all; overflow-wrap: normal; white-space: pre-wrap; overflow: hidden; font-family: &quot;Microsoft Yahei&quot;; background-color: rgb(255, 255, 255);"><span style="font-size: 18px;">【本文结束】出处：快科技</span></div>Sat, 24 Oct 2026 20:51:31 +0800https://www.dnzjds.com/zhuangji/3260.html<div class="paragraph " id="eSNQ29Sh8bYokuAtS5ACXdfSXqC5j7" style="white-space: pre-wrap; margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-indent: 1.6rem; font-size: 0.8rem; color: rgb(34, 34, 34); line-height: 36.8px; word-break: break-all; overflow-wrap: normal; overflow: hidden; font-family: &quot;Microsoft Yahei&quot;; background-color: rgb(255, 255, 255);"><span style="font-size: 18px;">快科技10月24日消息，近日，市场传出消息称，英伟达计划在2027年1月停产RTX 5090和RTX 5080两款显卡。</span><span style="font-size: 18px; text-indent: 1.6rem;">针对该传闻，英伟达方面已经公开否认停产说法，确认产品会持续出货，但同时坦言供应链产能存在限制。尽管如此，RTX 50系列的停产传闻，还是搅动了华强北电脑配件市场。</span></div><div class="paragraph " id="cLWa3XTIV8DjfM6kl6WNUbKnNJu57Z" style="white-space: pre-wrap; margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-indent: 1.6rem; font-size: 0.8rem; color: rgb(34, 34, 34); line-height: 36.8px; word-break: break-all; overflow-wrap: normal; overflow: hidden; font-family: &quot;Microsoft Yahei&quot;; background-color: rgb(255, 255, 255);"><span style="font-size: 18px;">据媒体报道，自9月以来，深圳电脑零部件外贸数据明显走高，华强北显卡板块率先异动：RTX 50系列高端消费级显卡价格持续跳涨，渠道报价一日一变，单日、隔日价差最高可达数百元。</span><span style="font-size: 18px; text-indent: 1.6rem;">与此同时，来自印度、俄罗斯、法国、意大利、巴西等地区的海外客商纷纷到店询价，采购重点集中在显卡、内存、硬盘等散件。</span></div><div class="paragraph " id="VkbP1CtHxCADKzvsi8w81P80wFk34c" style="white-space: pre-wrap; margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-indent: 1.6rem; font-size: 0.8rem; color: rgb(34, 34, 34); line-height: 36.8px; word-break: break-all; overflow-wrap: normal; overflow: hidden; font-family: &quot;Microsoft Yahei&quot;; background-color: rgb(255, 255, 255);"><span style="font-size: 18px;">据报道，本轮行情最突出的特点就是价格暴涨。以RTX 50系列为例，RTX 5090官方首发零售价约1.5万元—1.6万元，当前渠道报价已涨到5.5万元—6万元；RTX 5080首发售价约8000元，现已涨到1.7万元。</span><span style="font-size: 18px; text-indent: 1.6rem;">就连一位在华强北经营16年的商铺经理都感慨：“半个月前RTX 5080渠道价约1.2万元，短期就上涨了4000元—5000元；RTX 5090此前约3万元，现在直接攀升到5万元级别。”</span></div><div class="paragraph " id="pM8RAW1e3sTuJ0pohuPaunjTDCcFdq" style="white-space: pre-wrap; margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-indent: 1.6rem; font-size: 0.8rem; color: rgb(34, 34, 34); line-height: 36.8px; word-break: break-all; overflow-wrap: normal; overflow: hidden; font-family: &quot;Microsoft Yahei&quot;; background-color: rgb(255, 255, 255);"><span style="font-size: 18px;">那么，本轮显卡涨价背后是什么原因？多位商户向媒体分析，核心是芯片产能分配调整。</span></div><div class="paragraph " id="DEogtDYPScDE4p2EQS52N9otQAxgh7" style="white-space: pre-wrap; margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-indent: 1.6rem; font-size: 0.8rem; color: rgb(34, 34, 34); line-height: 36.8px; word-break: break-all; overflow-wrap: normal; overflow: hidden; font-family: &quot;Microsoft Yahei&quot;; background-color: rgb(255, 255, 255);"><span style="font-size: 18px;">有商家表示，现在一卡难求，一方面是供给端产能结构变化，另一方面也受到AI 算力需求外溢带动，原本可以供给消费级显卡的芯片资源，被算力卡挤占。</span><span style="font-size: 18px; text-indent: 1.6rem;">“涨价的核心原因在于英伟达芯片产能优先向AI算力卡生产倾斜，消费级显卡芯片产能被压缩，供给不足支撑价格持续走高。”有商家工作人员分析。</span></div><div class="paragraph " id="HFl08M3TzildqJZMUaVaRuSvwsMUTT" style="white-space: pre-wrap; margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-indent: 1.6rem; font-size: 0.8rem; color: rgb(34, 34, 34); line-height: 36.8px; word-break: break-all; overflow-wrap: normal; overflow: hidden; font-family: &quot;Microsoft Yahei&quot;; background-color: rgb(255, 255, 255);"><span style="font-size: 18px;">实际上，过往公开报道显示，近年来全球科技企业纷纷加码AI算力，英伟达持续将晶圆、先进封装等稀缺产能向数据中心算力芯片倾斜，数据中心业务营收规模远超游戏业务，利润空间也更高。</span><span style="font-size: 18px; text-indent: 1.6rem;">受此影响，消费级显卡的芯片配额被持续压缩，叠加海外集中采购、渠道囤货炒作，进一步放大供给缺口，助推了价格上行。</span></div><div class="pic" style="text-wrap-mode: wrap; margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-align: center; position: relative; color: rgb(116, 116, 116); font-family: &quot;Microsoft Yahei&quot;; background-color: rgb(255, 255, 255);"><img class="img loaded" src="https://www.dnzjds.com/zb_users/upload/2026/10/20261024202503_72021.png" alt="英伟达RTX 50系列传闻要停产 华强北显卡卖疯 从1.5万炒到6万" data-was-processed="true" style="border: 0px; padding: 0px; max-width: 100%; margin: 20px auto; display: block; min-width: 100px; min-height: 102px; background: transparent; border-radius: 0px; position: relative;"/><div class="description" style="margin: 12px 0px 0px; padding: 0px; box-sizing: border-box; color: rgb(172, 172, 172); font-size: 0.6rem;"><span style="font-size: 18px;">英伟达RTX 50系列传闻要停产 华强北显卡卖疯 从1.5万炒到6万</span></div></div><div class="paragraph " id="IxCfcnhWuwf2NTnRIT4ewy8rUvXJX5" style="white-space: pre-wrap; margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-indent: 1.6rem; font-size: 0.8rem; color: rgb(34, 34, 34); line-height: 36.8px; word-break: break-all; overflow-wrap: normal; overflow: hidden; font-family: &quot;Microsoft Yahei&quot;; background-color: rgb(255, 255, 255);"><span style="font-size: 18px;">【本文结束】出处：快科技</span></div><p><br/></p>Sat, 24 Oct 2026 20:24:37 +0800https://www.dnzjds.com/zhuangji/3259.html<div class="paragraph " id="i4lKX1Gb90ynYqHiFelUMerpEmlXXL" style="margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-indent: 1.6rem; font-size: 0.8rem; color: rgb(34, 34, 34); line-height: 36.8px; word-break: break-all; overflow-wrap: normal; white-space: pre-wrap; overflow: hidden; font-family: &quot;Microsoft Yahei&quot;; background-color: rgb(255, 255, 255);"><span style="font-size: 18px;">快科技10月23日消息，据报道，Linux 7.4内核近日合入了x86/bugs代码合并请求，其中一项关键调整直接关系到AMD下一代Zen 6处理器的性能表现。内核将利用Zen 6新增的BTB CTX硬件隔离功能，正式关闭开销较大的SafeRET软件安全缓解措施。</span></div><div class="pic" style="margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-align: center; position: relative; color: rgb(116, 116, 116); font-family: &quot;Microsoft Yahei&quot;; text-wrap-mode: wrap; background-color: rgb(255, 255, 255);"><img class="img loaded" src="https://www.dnzjds.com/zb_users/upload/2026/10/20261023141956_59794.png" alt="不靠软件补丁防漏洞！AMD Zen 6用硬件搞定：性能白赚" data-was-processed="true" style="border: 0px; padding: 0px; max-width: 100%; margin: 20px auto; display: block; min-width: 100px; min-height: 102px; background: transparent; border-radius: 0px; position: relative;"/><div class="description" style="margin: 12px 0px 0px; padding: 0px; box-sizing: border-box; color: rgb(172, 172, 172); font-size: 0.6rem;"><span style="font-size: 18px;">不靠软件补丁防漏洞！AMD Zen 6用硬件搞定：性能白赚</span></div></div><div class="paragraph " id="VefY1Zfv5Ftb8Gg9H0sRPiuteHqJR9" style="margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-indent: 1.6rem; font-size: 0.8rem; color: rgb(34, 34, 34); line-height: 36.8px; word-break: break-all; overflow-wrap: normal; white-space: pre-wrap; overflow: hidden; font-family: &quot;Microsoft Yahei&quot;; background-color: rgb(255, 255, 255);"><span style="font-size: 18px;">BTB全称分支目标缓冲区，是CPU前端预测间接分支跳转方向的核心部件。在Zen 1到Zen 4上，用户态进程和内核态共享BTB条目，恶意进程可以污染预测条目，诱导内核跳转到错误地址，进而泄露敏感数据。这正是2023年曝光的SRSO漏洞，也称Inception。</span><span style="font-size: 18px; text-indent: 1.6rem;">Linux此前的应对方案是SafeRET，它通过重写返回指令序列，让攻击者无法瞄准分支预测器。这套机制确实有效，但代价是每一次函数返回都要执行额外的指令，长期拖累系统性能。</span></div><div class="paragraph " id="OeIAVNOTxUU7NGNSZae5gMP3e20OEf" style="margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-indent: 1.6rem; font-size: 0.8rem; color: rgb(34, 34, 34); line-height: 36.8px; word-break: break-all; overflow-wrap: normal; white-space: pre-wrap; overflow: hidden; font-family: &quot;Microsoft Yahei&quot;; background-color: rgb(255, 255, 255);"><span style="font-size: 18px;">Zen 6在硬件层面解决了这个问题，BTB条目按上下文打上标签，内核的预测查询永远不会命中用户进程植入的条目，特权级隔离在源头生效。</span><span style="font-size: 18px; text-indent: 1.6rem;">因此，Linux 7.4可以直接关闭SafeRET，把软件开销彻底抹掉。补丁由AMD工程师Borislav Petkov提交，已于2026年9月合入x86/bugs分支，预计随Linux 7.4正式发布。</span></div><div class="paragraph " id="Lo0Yyrju61dSrTqH4VORlbAvRMcB4I" style="margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-indent: 1.6rem; font-size: 0.8rem; color: rgb(34, 34, 34); line-height: 36.8px; word-break: break-all; overflow-wrap: normal; white-space: pre-wrap; overflow: hidden; font-family: &quot;Microsoft Yahei&quot;; background-color: rgb(255, 255, 255);"><span style="font-size: 18px;">不过需要注意，BTB隔离只覆盖用户态到内核态、客户机到宿主机两条路径。用户进程之间、虚拟机之间的攻击向量，仍需依赖Spectre v2防护和上下文切换时的IBPB屏障。对Zen 6用户而言，这是一次实打实的免费性能回归。</span></div><div class="pic" style="margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-align: center; position: relative; color: rgb(116, 116, 116); font-family: &quot;Microsoft Yahei&quot;; text-wrap-mode: wrap; background-color: rgb(255, 255, 255);"><img class="img loaded" src="https://www.dnzjds.com/zb_users/upload/2026/10/20261023141956_77995.jpg" alt="不靠软件补丁防漏洞！AMD Zen 6用硬件搞定：性能白赚" data-was-processed="true" style="border: 0px; padding: 0px; max-width: 100%; margin: 20px auto; display: block; min-width: 100px; min-height: 102px; background: transparent; border-radius: 0px; position: relative;"/><div class="description" style="margin: 12px 0px 0px; padding: 0px; box-sizing: border-box; color: rgb(172, 172, 172); font-size: 0.6rem;"><span style="font-size: 18px;">不靠软件补丁防漏洞！AMD Zen 6用硬件搞定：性能白赚</span></div></div><div class="paragraph " id="xGAF6CZjHYEsAyMKx8eXU9EHSnz5Fj" style="margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-indent: 1.6rem; font-size: 0.8rem; color: rgb(34, 34, 34); line-height: 36.8px; word-break: break-all; overflow-wrap: normal; white-space: pre-wrap; overflow: hidden; font-family: &quot;Microsoft Yahei&quot;; background-color: rgb(255, 255, 255);"><span style="font-size: 18px;">【本文结束】出处：快科技</span></div><p><br/></p>Fri, 23 Oct 2026 14:17:23 +0800https://www.dnzjds.com/zhuangji/3258.html<div class="paragraph " id="dAuukrmX7OMsJulAA34ymByQlACwiv" style="margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-indent: 1.6rem; font-size: 0.8rem; color: rgb(34, 34, 34); line-height: 36.8px; word-break: break-all; overflow-wrap: normal; white-space: pre-wrap; overflow: hidden; font-family: &quot;Microsoft Yahei&quot;; background-color: rgb(255, 255, 255);"><span style="font-size: 18px;">快科技10月23日消息，AMD目前正在全力开发代号为“FSR光线重建2.0”的次世代神经网络降噪系统，旨在彻底解决光线追踪技术在实际运行中的效率瓶颈。</span><span style="font-size: 18px; text-indent: 1.6rem;">据了解，该技术隶属于AMD Redstone SDK库文件包。在首代降噪算法中，光线追踪与路径追踪的计算过程往往会消耗显卡约10%至20%的原生性能，这种高昂的开销在很大程度上限制了重度光追游戏的普及。</span></div><div class="pic" style="margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-align: center; position: relative; color: rgb(116, 116, 116); font-family: &quot;Microsoft Yahei&quot;; text-wrap-mode: wrap; background-color: rgb(255, 255, 255);"><img class="img loaded" src="https://www.dnzjds.com/zb_users/upload/2026/10/20261023141717_81276.png" alt="光追拖影大幅减少！AMD推出全新降噪技术：算力开销直接砍半" data-was-processed="true" style="border: 0px; padding: 0px; max-width: 100%; margin: 20px auto; display: block; min-width: 100px; min-height: 102px; background: transparent; border-radius: 0px; position: relative;"/><div class="description" style="margin: 12px 0px 0px; padding: 0px; box-sizing: border-box; color: rgb(172, 172, 172); font-size: 0.6rem;"><span style="font-size: 18px;">光追拖影大幅减少！AMD推出全新降噪技术：算力开销直接砍半</span></div></div><div class="paragraph " id="qScETKuLizya1Z3s1c4uiRpFImcR1i" style="margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-indent: 1.6rem; font-size: 0.8rem; color: rgb(34, 34, 34); line-height: 36.8px; word-break: break-all; overflow-wrap: normal; white-space: pre-wrap; overflow: hidden; font-family: &quot;Microsoft Yahei&quot;; background-color: rgb(255, 255, 255);"><span style="font-size: 18px;">AMD工程师在2.0版本把重心放在神经网络权重优化上，新算法强化了和Radeon显卡硬件矩阵加速器的底层联动，尽可能压低计算开销，目标最高可减少50%，让开启光追后也能保住高帧率。</span><span style="font-size: 18px; text-indent: 1.6rem;">画质层面，FSR光线重建2.0重点解决高速动态场景里的画面瑕疵，新模型依托更大规模数据集训练，能够更精准还原几何阴影、微阴影以及复杂镜面反射。</span></div><div class="paragraph " id="Sm1oSc7Kn2F7Jl0Pq3FTkcvpXHNDaA" style="margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-indent: 1.6rem; font-size: 0.8rem; color: rgb(34, 34, 34); line-height: 36.8px; word-break: break-all; overflow-wrap: normal; white-space: pre-wrap; overflow: hidden; font-family: &quot;Microsoft Yahei&quot;; background-color: rgb(255, 255, 255);"><span style="font-size: 18px;">就算在每像素光线数量很少的低采样场景下，也可以有效抑制画面闪烁和拖影问题。</span><span style="font-size: 18px; text-indent: 1.6rem;">除此之外，这项新技术还在降低对原生渲染分辨率的依赖。简单来讲，开启高倍率缩放时，FSR 2.0可以更好留住环境细节与精细纹理，避免AI重构带来的画面发糊问题。</span><span style="font-size: 18px; text-indent: 1.6rem;">不过想要用上这项新功能也有门槛：显卡需要支持机器学习指令；系统必须是Windows 11，同时满足DirectX 12、Shader Model 6.6及以上标准。</span></div><div class="pic" style="margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-align: center; position: relative; color: rgb(116, 116, 116); font-family: &quot;Microsoft Yahei&quot;; text-wrap-mode: wrap; background-color: rgb(255, 255, 255);"><img class="img loaded" src="https://www.dnzjds.com/zb_users/upload/2026/10/20261023141717_70604.jpg" alt="光追拖影大幅减少！AMD推出全新降噪技术：算力开销直接砍半" data-was-processed="true" style="border: 0px; padding: 0px; max-width: 100%; margin: 20px auto; display: block; min-width: 100px; min-height: 102px; background: transparent; border-radius: 0px; position: relative;"/><div class="description" style="margin: 12px 0px 0px; padding: 0px; box-sizing: border-box; color: rgb(172, 172, 172); font-size: 0.6rem;"><span style="font-size: 18px;">光追拖影大幅减少！AMD推出全新降噪技术：算力开销直接砍半</span></div></div><div class="paragraph " id="wiumhlKtztgM4KOKEedMz0MmsGaIRY" style="margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-indent: 1.6rem; font-size: 0.8rem; color: rgb(34, 34, 34); line-height: 36.8px; word-break: break-all; overflow-wrap: normal; white-space: pre-wrap; overflow: hidden; font-family: &quot;Microsoft Yahei&quot;; background-color: rgb(255, 255, 255);"><span style="font-size: 18px;">【本文结束】出处：快科技</span></div><p><br/></p>Fri, 23 Oct 2026 14:16:38 +0800https://www.dnzjds.com/zhuangji/3257.html<div class="paragraph " id="rg61pmOeX5XU9RrJFMAinNo3aDR5Y9" style="margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-indent: 1.6rem; font-size: 0.8rem; color: rgb(34, 34, 34); line-height: 36.8px; word-break: break-all; overflow-wrap: normal; white-space: pre-wrap; overflow: hidden; font-family: &quot;Microsoft Yahei&quot;; background-color: rgb(255, 255, 255);"><span style="font-size: 18px;">快科技10月23日消息，TrendForce发布预测，认为服务器CPU出货有望在2026、2027两年分别增长31%和超40%，这得益于代理式人工智能工作负载对CPU与加速器需求比例的提升。</span><span style="font-size: 18px; text-indent: 1.6rem;">总体来看，基本全数基于Arm指令集的非x86阵营今明两年有望分别拿下23%和30%市占，份额显著提升；龙头英特尔则由于下代“Diamond Rapids”出货迟缓而出现市占下滑。</span></div><div class="paragraph " id="6gW9OBQWzRm73zwAapVmdHyxZPg43Z" style="margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-indent: 1.6rem; font-size: 0.8rem; color: rgb(34, 34, 34); line-height: 36.8px; word-break: break-all; overflow-wrap: normal; white-space: pre-wrap; overflow: hidden; font-family: &quot;Microsoft Yahei&quot;; background-color: rgb(255, 255, 255);"><span style="font-size: 18px;">CPU是计算机核心，也是程序的调用者和运行者，程序的每一条指令都要经过CPU解析和执行。从发展历史看，CPU的字长、晶体管数量、主频、核心数持续提升。</span><span style="font-size: 18px; text-indent: 1.6rem;">AI推理与智能体应用正重塑算力结构，CPU重回核心。中原证券认为，Muse上线22天下载量突破500万次，同期ChatGPT仅230万次。</span></div><div class="paragraph " id="M7i5KEDlYjij9ZOiQO5cV1nUDBNUSH" style="margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-indent: 1.6rem; font-size: 0.8rem; color: rgb(34, 34, 34); line-height: 36.8px; word-break: break-all; overflow-wrap: normal; white-space: pre-wrap; overflow: hidden; font-family: &quot;Microsoft Yahei&quot;; background-color: rgb(255, 255, 255);"><span style="font-size: 18px;">AI智能体需要自主规划、多步推理、调用外部工具、连接数据库、执行代码并持续交互，工作负载从计算密集转向控制密集、分支密集、IO密集。GPU负责计算，CPU决定能否持续高效计算，Agentic AI时代CPU重回AI数据中心核心。</span><span style="font-size: 18px; text-indent: 1.6rem;">AMD CEO苏姿丰表示，2030年服务器CPU市场规模将从此前1200亿美元上调至2200亿美元，复合增速将超50%。传统AI数据中心CPU与GPU比例为1:4至1:8，智能体正驱动其转向接近1:1。</span></div><div class="pic" style="margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-align: center; position: relative; color: rgb(116, 116, 116); font-family: &quot;Microsoft Yahei&quot;; text-wrap-mode: wrap; background-color: rgb(255, 255, 255);"><img class="img loaded" src="https://www.dnzjds.com/zb_users/upload/2026/10/20261023141408_60971.png" alt="2027服务器CPU出货有望大增40% Arm拿下三成市场" data-was-processed="true" style="border: 0px; padding: 0px; max-width: 100%; margin: 20px auto; display: block; min-width: 100px; min-height: 102px; background: transparent; border-radius: 0px; position: relative;"/><div class="description" style="margin: 12px 0px 0px; padding: 0px; box-sizing: border-box; color: rgb(172, 172, 172); font-size: 0.6rem;"><span style="font-size: 18px;">2027服务器CPU出货有望大增40% Arm拿下三成市场</span></div></div><div class="paragraph " id="d73affZcWUXMYwLcdp07HffoWfbD6o" style="margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-indent: 1.6rem; font-size: 0.8rem; color: rgb(34, 34, 34); line-height: 36.8px; word-break: break-all; overflow-wrap: normal; white-space: pre-wrap; overflow: hidden; font-family: &quot;Microsoft Yahei&quot;; background-color: rgb(255, 255, 255);"><span style="font-size: 18px;">【本文结束】出处：快科技</span></div>Fri, 23 Oct 2026 14:12:42 +0800https://www.dnzjds.com/zhuangji/3256.html<div class="paragraph " id="s81bQsTAeDtGLsB9uchKBpLnjvd28V" style="margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-indent: 1.6rem; font-size: 0.8rem; color: rgb(34, 34, 34); line-height: 36.8px; word-break: break-all; overflow-wrap: normal; white-space: pre-wrap; overflow: hidden; font-family: &quot;Microsoft Yahei&quot;; background-color: rgb(255, 255, 255);"><span style="font-size: 18px;">快科技10月23日消息，在传奇显卡3dfx Voodoo诞生30周年之际，其联合创始人罗斯·史密斯在接受采访时公开致谢id Software及其经典作品《雷神之锤》，直言这款游戏是3dfx取得全球性成功的头号功臣。</span><span style="font-size: 18px; text-indent: 1.6rem;">史密斯在回顾中提到，3dfx能够卖出数百万块显卡，很大程度上得益于与id Software 联合创始人，《雷神之锤》主引擎开发者约翰·卡马克的深度协作。</span></div><div class="pic" style="margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-align: center; position: relative; color: rgb(116, 116, 116); font-family: &quot;Microsoft Yahei&quot;; text-wrap-mode: wrap; background-color: rgb(255, 255, 255);"><img class="img loaded" src="https://www.dnzjds.com/zb_users/upload/2026/10/20261023141018_90623.png" alt="初代3D卡迎来30周年！创始人致敬《雷神之锤》：靠它卖出数百万显卡" data-was-processed="true" style="border: 0px; padding: 0px; max-width: 100%; margin: 20px auto; display: block; min-width: 100px; min-height: 102px; background: transparent; border-radius: 0px; position: relative;"/><div class="description" style="margin: 12px 0px 0px; padding: 0px; box-sizing: border-box; color: rgb(172, 172, 172); font-size: 0.6rem;"><span style="font-size: 18px;">初代3D卡迎来30周年！创始人致敬《雷神之锤》：靠它卖出数百万显卡</span></div></div><div class="paragraph " id="Spt3G8ThJEP1fGzyecJn9WcppeF6Hx" style="margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-indent: 1.6rem; font-size: 0.8rem; color: rgb(34, 34, 34); line-height: 36.8px; word-break: break-all; overflow-wrap: normal; white-space: pre-wrap; overflow: hidden; font-family: &quot;Microsoft Yahei&quot;; background-color: rgb(255, 255, 255);"><span style="font-size: 18px;">要知道《雷神之锤》于1996年6月发售，但直到1997年初才推出支持3D硬件加速的GLQuake版本。</span><span style="font-size: 18px; text-indent: 1.6rem;">据史密斯回忆，约翰·卡马克主动联系3dfx，双方随后合作开发了名为MiniGL的API，使3dfx显卡得以运行该游戏，并以更高分辨率、更丰富细节与更广色域呈现画面。</span><span style="text-indent: 1.6rem; font-size: 18px;">史密斯对此评价道：&quot;我一定要把这份功劳归给id、John和《雷神之锤》——我们可能正是因为那款游戏卖出了数百万块板卡，那真是太精彩了。那也是我最后一款打通关的游戏，玩完后我精疲力竭，它做得太精良、太美、太黑暗了。&quot;</span><span style="text-indent: 1.6rem; font-size: 18px;">3dfx曾在发展初期就主动与游戏开发商建立深度合作关系，这套合作思路在当年极具颠覆性。S3等竞品对开发者态度冷淡，只把对方当成不得不应付的麻烦，而3dfx则将游戏开发者视作核心命脉。</span></div><div class="paragraph " id="wLfZXXRZUsQ4oqkmwSjjQ8AxjYEels" style="margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-indent: 1.6rem; font-size: 0.8rem; color: rgb(34, 34, 34); line-height: 36.8px; word-break: break-all; overflow-wrap: normal; white-space: pre-wrap; overflow: hidden; font-family: &quot;Microsoft Yahei&quot;; background-color: rgb(255, 255, 255);"><span style="font-size: 18px;">他们会提前向开发者展示Voodoo的预渲染效果，这套以开发者为中心的策略，最终让Voodoo在分辨率、画面细节、色彩表现上，碾压传统软件渲染方案。</span><span style="font-size: 18px; text-indent: 1.6rem;">这种硬件厂商与游戏开发商直接协作的模式，如今在英伟达、英特尔、AMD身上仍能看到延续。</span></div><div class="pic" style="margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-align: center; position: relative; color: rgb(116, 116, 116); font-family: &quot;Microsoft Yahei&quot;; text-wrap-mode: wrap; background-color: rgb(255, 255, 255);"><img class="img loaded" src="https://www.dnzjds.com/zb_users/upload/2026/10/20261023141018_42802.png" alt="初代3D卡迎来30周年！创始人致敬《雷神之锤》：靠它卖出数百万显卡" data-was-processed="true" style="border: 0px; padding: 0px; max-width: 100%; margin: 20px auto; display: block; min-width: 100px; min-height: 102px; background: transparent; border-radius: 0px; position: relative;"/><div class="description" style="margin: 12px 0px 0px; padding: 0px; box-sizing: border-box; color: rgb(172, 172, 172); font-size: 0.6rem;"><span style="font-size: 18px;">初代3D卡迎来30周年！创始人致敬《雷神之锤》：靠它卖出数百万显卡</span></div></div><div class="paragraph " id="K4SIwuTqh1q8LQag7g9Kf8htCHMCxE" style="margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-indent: 1.6rem; font-size: 0.8rem; color: rgb(34, 34, 34); line-height: 36.8px; word-break: break-all; overflow-wrap: normal; white-space: pre-wrap; overflow: hidden; font-family: &quot;Microsoft Yahei&quot;; background-color: rgb(255, 255, 255);"><span style="font-size: 18px;">除此之外，3dfx的产品设计同样极具开创性。Voodoo并非独立显卡，而是一款协处理器，需要搭配2D显卡协同工作，专门承担3D加速任务，CPU则可以同时处理其他后台任务。</span></div><div class="paragraph " id="ZqtzjQL55Qd4LIECHd6vQEZNrLhYlU" style="margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-indent: 1.6rem; font-size: 0.8rem; color: rgb(34, 34, 34); line-height: 36.8px; word-break: break-all; overflow-wrap: normal; white-space: pre-wrap; overflow: hidden; font-family: &quot;Microsoft Yahei&quot;; background-color: rgb(255, 255, 255);"><span style="font-size: 18px;">史密斯直言，这个设计和它没有集成VGA输出直接相关，一方面是受限于当时的设计经验，另一方面也是出于利润、兼容性的权衡：“要兼容上百万款应用软件难度极高，但只适配200款游戏就简单很多。”</span></div><div class="paragraph " id="ah9JISPNQCzJpFG8HbXGIortmvdXS1" style="margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-indent: 1.6rem; font-size: 0.8rem; color: rgb(34, 34, 34); line-height: 36.8px; word-break: break-all; overflow-wrap: normal; white-space: pre-wrap; overflow: hidden; font-family: &quot;Microsoft Yahei&quot;; background-color: rgb(255, 255, 255);"><span style="font-size: 18px;">正是这套聚焦游戏的思路，让Voodoo在当年一众通用图形方案里杀出重围，也为后来游戏专用GPU的发展打下基础。</span></div><div class="pic" style="margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-align: center; position: relative; color: rgb(116, 116, 116); font-family: &quot;Microsoft Yahei&quot;; text-wrap-mode: wrap; background-color: rgb(255, 255, 255);"><img class="img loaded" src="https://www.dnzjds.com/zb_users/upload/2026/10/20261023141018_97521.png" alt="初代3D卡迎来30周年！创始人致敬《雷神之锤》：靠它卖出数百万显卡" data-was-processed="true" style="border: 0px; padding: 0px; max-width: 100%; margin: 20px auto; display: block; min-width: 100px; min-height: 102px; background: transparent; border-radius: 0px; position: relative;"/><div class="description" style="margin: 12px 0px 0px; padding: 0px; box-sizing: border-box; color: rgb(172, 172, 172); font-size: 0.6rem;"><span style="font-size: 18px;">初代3D卡迎来30周年！创始人致敬《雷神之锤》：靠它卖出数百万显卡</span></div></div><div class="paragraph " id="t3jmKlm9Fj0z80ffvBLCtsaXgDZkMA" style="margin: 32px 0px 0px; padding: 0px; box-sizing: border-box; text-indent: 1.6rem; font-size: 0.8rem; color: rgb(34, 34, 34); line-height: 36.8px; word-break: break-all; overflow-wrap: normal; white-space: pre-wrap; overflow: hidden; font-family: &quot;Microsoft Yahei&quot;; background-color: rgb(255, 255, 255);"><span style="font-size: 18px;">【本文结束】出处：快科技</span></div><p><br/></p>Fri, 23 Oct 2026 14:09:45 +0800