据Wccftech报道，有网友使用了Pixel 9 Pro XL进行测试，显示在长时间高负载下，Tensor G4性能表现不佳，最多能降低至峰值性能的50%。在持续100%负载下，仅过了三分钟，CPU就开始降频了，四分钟后频率降幅最大，接近60%。

已经有用户开始抨击Pixel 9系列的性能表现，认为作为1099美元（Pixel 9 Pro XL）起售价格的旗舰产品，这样的性能输出实在不应该。过往Tensor系列芯片就因性能平庸及热控制问题被诟病，而随着Pixel系列的定位变高、价格攀升，Tensor G4的表现自然更受大家关注。

Tensor G4采用了三星的4nm工艺制造，CPU部分采用了1+3+4的三丛架构，共8个核心，其中超大核将继续沿用Cortex-X4核心，频率为3.1 GHz，另外还有四个Cortex-A720性能核心，频率为2.6 GHz，剩下是四个Cortex-A520能效核心，频率为1.95 GHz。GPU部分与Tensor G3一样，都是Mali-G715，不过频率从890 MHz提高到940 MHz。另外Tensor G4搭配的5G调制解调器是Exynos 5400，支持卫星连接。

不少人认为，Tensor系列芯片一直以来较为糟糕的表现与谷歌选择三星的制造工艺有关。幸运的是，明年谷歌将改变策略，在用于Pixel 10系列的Tensor G5上更换代工厂，变成了台积电，让其SoC至少在制造工艺上能与高通和联发科的产品处于同一水平线。

此外，报道文章还提到Ola目前已展示了Bodhi系列的部分AI芯片，以及Sarv-1云原生CPU和Ojas边缘AI芯片。其中Bodhi-1 AI芯片预计在2026年发布，Ola称该产品专为大规模LLM设计，将拥有“最佳的能效表现”，旨在迎合AI细分市场的大部分需求，同时其下一代产品Bodhi-2预计在2028年发布，Ola称Bodhi-2是一款对标业界顶级芯片推出的产品，性能和能效方面将优于英伟达的GPU（未表明具体型号），额定功耗为200W。而Sarv-1芯片，Ola称这是一款云原生芯片，是一款基于ARM指令集的通用服务器CPU，专为云计算领域设计，同时有传闻称Sarv-1可能采用ARM Neoverse N3内核。至于Ojas边缘AI芯片，Ola则表示该系列芯片将集成到自家的下一代电动汽车中，以帮助其运行充电、ADAS等系统，同时该系列芯片将通过AI原生架构支持广泛的生态系统。

至于在代工方面，Bhavish Aggarwal声称他们将与全球一级或二级代工厂合作，可能是台积电或三星。

华硕官方微博近日放出了消息，表示会在8月20日发布X870E/X870系列主板，从图上的可以看出，系你地主板包括TUF系列、ProArt系列、ROG Crosshair系列和ROG Strix系列。华硕大概率是要在德国科隆游戏展上举办发布会发布新的主板，需要注意的是，发布并不代表开卖，X870E/X870系列主板大概率要9月份才上市。

预览var tableid = settableid(); document.write(''); Table().init({ title: "AMD 800/600系列芯片规格对比", header: ["型号", "X870E","X870","B850","B840","X670E","X670","B650E","B650"], width: '100%', columnWidth: 'MAX', //表格单元宽度，"MAX"最大内容显示，""AVG"平均，"AUTO"自动，指定[15,16,20] id: tableid, data: [ ["CPU超频","√","√","√","×","√","√","√","√"], ["内存超频","√","√","√","√","√","√","√","√"], ["PCIe x16","Gen 5","Gen 5","可选Gen 5","Gen 4","Gen 5","Gen 4","Gen 5","Gen 4"], ["CPU M.2口","Gen 5","Gen 5","Gen 5","Gen 4","Gen 5","Gen 4","Gen 5","Gen 5"], ["板载USB4","√","√","可选","×","可选","可选","可选","可选"], ["通用PCIe","Gen 4 x12Gen 3 x8","Gen 4 x8Gen 3 x4","Gen 4 x8Gen 3 x4","Gen 3 x8","Gen 4 x12Gen 3 x8","Gen 4 x12Gen 3 x8","Gen 4 x8Gen 3 x4","Gen 4 x8Gen 3 x4"], ["芯片","2*Prom21","Prom21","Prom21","Prom19","2*Prom21","2*Prom21","Prom21","Prom21",] ] });根据此前的消息，AMD 800系主板并没有采用新的南桥芯片，X870E配了两颗Prom21南桥，整体规格其实就是配了USB4接口的X670E，由CPU提供的PCIe x16和M.2接口均支持PCIe 5.0，主板提供了12条PCIe 4.0和8条PCIe 3.0。而X870只有一颗Prom21南桥，在PCIe 5.0和USB4的支持上与X870E是相同的，但通用PCIe缩减到只有8条PCIe 4.0和4条PCIe 3.0，这规格看起来就是配了USB4接口的B650E。

此外更便宜的B850主板要更晚才会出现，可能要等到2025初，主板使用的芯片其实和B650是一样的，都是Prom21。AMD不会像X870那样要求B850一定得支持USB4，规格和B650接近，CPU M.2口强制支持PCIe 5.0，而显卡的x16接口是否支持由板厂决定。至于B840，定位更低，不支持PCIe 5.0，它用的是Prom19芯片，不支持PCIe 4.0，只可提供8条PCIe 3.0，不支持CPU超频，支持内存超频，它的定位实际上更加接近现在的A620。

据TrendForce报道，三星在存储器领域的主要竞争对手SK海力士，计划2026年引入ASML下一代TWINSCAN EXE:5200系统，并增加内部High-NA EUV开发人员的数量。与三星打算将High-NA EUV首先用于晶圆代工业务不同，SK海力士有望用在最先进的DRAM产品上。SK海力士暂时没有透露具体的计划内容，比如安装设备的工厂或者追加投资方向等。

传闻ASML手上已经收到了十多台High-NA EUV光刻机的订单，包括英特尔、台积电、三星、SK海力士、以及美光。ASML至今生产了8台TWINSCAN EXE:5000，这是第一代High-NA EUV产品，其中英特尔占据了最多的数量，近期将迎来第二台High-NA EUV光刻机，而三星则是ASML首批产品的最后一个订单客户。

由于High-NA EUV光刻机定价较高，有传言称达到了3.8亿美元，使得半导体制造商在投资上普遍都比较谨慎。

G27H4显示器搭载了27英寸Fast IPS面板，色深为10bit（8bit+FRC），可视角度为178度（水平和垂直），分辨率为2560 x 1440，刷新率为300Hz，响应时间为1ms，SDR/HDR亮度分别为350/400尼特，静态对比度为1000:1，DCI-P3色域为98%，sRGB色域为130%，Adobe RGB色域为100%，色准Delta≤2。HKC表示，经过专业电竞色彩调校，可以精准还原游戏战场，让玩家畅享更逼真沉浸的视觉盛宴。

HKC为G27H4显示器提供了定制游戏模式，拥有丰富的游戏辅助功能，玩家可以随心调整自定义功能选项。其中游戏模式包括游戏准星、暗场亮效、游戏计时和Free Sync，职业情景预设包括暗部均衡（FPS）、色彩均衡（RPG）、以及色彩增益（RCG）。同时搭载了HKC自研DIC5动态模糊消除技术，可以显著增强运动画面清晰度，减少拖影。另外还支持DC调光和硬件低蓝光，有效减少高能短波蓝光，减轻画面闪烁现象，还能延缓视觉疲劳。

G27H4显示器拥有两个HDMI 2.1接口、一个DisplayPort 1.4接口、以及音频插孔，采用了六边形小巧底座，稳固支撑的同时占地面积更小，配备的人体工学支架，具有0至100mm升降高度、-5°至20°的前倾后仰、以及-90°至90°的垂直旋转。此外，该款显示器还支持75 x 75 mm规格的VESA壁挂，便于玩家使用悬臂支架。

AOC Q27G12ZE采用黑红相配的外观设计，机身背面有两抹红色花纹，使显示器外观电竞风十足。支架底座采用X型设计，支架上留有红色的理线孔，方便用户整理线材，保持桌面整洁。显示器采用一块27英寸的Fast IPS面板，2560×1440分辨率，标准刷新率达到240Hz，AOC官方还表示刷新率可超频至260Hz，但需要接DP接口。显示器最高亮度300nits，拥有135% sRGB和101% DCI-P3色域容积，支持10bit色深，平均Delta E＜2。

另外，Q27G12ZE还拥有多种显示模式，预设了多媒体、网络、办公室和阅读等4个低蓝光情景模式，能够根据使用需求轻松切换，有效减少有害短波蓝光，且拥有不闪屏技术，缓解长时间观看带来的眼疲劳。显示器还支持PBP/PIP功能，并配备G-MENU调节软件，可以为每一款游戏定制色彩和亮度，打造更舒适的游戏体验。接口方面，Q27G12ZE配备了1个HDMI 2.0接口、2个DP 1.4接口和1个音频接口，并用按键来调整显示器菜单。

《黑神话：悟空》联名版TiPlus7100 SSD 1TB，售价629元，京东地址：点此前往>>>《黑神话：悟空》联名版TiPlus7100 SSD 2TB，售价1069元，京东地址：点此前往>>>

在外观上，《黑神话：悟空》联名版TiPlus7100 SSD在外包装和正面贴纸上加入了《黑神话：悟空》的logo及相关的文字元素，并且内附限定天命人专属ID身份卡，有着对应的编号，不过是随机掉落。

致态《黑神话：悟空》联名版TiPlus7100 SSD选择的也是无缓存的存储设计方案，采用了PCIe 4.0 x4接口，外形为M.2 2280规格，搭载了长江存储原厂的TLC颗粒，支持HMB、SLC Cache技术以及NVMe 1.4规范。其连续读写速度可达7000 MB/s和6000 MB/s，最大随机读取900K IOPS，最大随机写入700K IOPS，适配PS5、笔记本、台式机灯设备，为用户带来沉浸式的游戏体验。质保方面，1TB和2TB版本对应的写入量分别为600TBW和1200TBW，年限均为5年。

据Benchlife报道，英特尔已敲定了酷睿Ultra 200系列台式机处理器的发布时间，将于10月10日发布。其P-Core和E-Core都分别升级至新的Lion Cove和Skymont架构，并启用新的LGA 1851插座，取代现有的LGA 1700插座。

酷睿Ultra 200系列台式机处理器不再支持超线程技术和DDR4内存，将搭配800系列芯片组，支持DDR5-6400内存。厂商出货的主板在BIOS仅能设定为“Intel Default Settings”，用户可以在BIOS自行选择“Performance”或者“Extreme（可能只有Core Ultra 9出现）”等其他选项。

预览var tableid = settableid(); document.write(''); Table().init({ title: "酷睿Ultra 200系列台式机处理器 规格参数对比", header: ["型号", "Cores", "基础频率 (GHz)", "睿频 (GHz)", "TVB", "TBMT 3.0", "TBT 2.0", "TDP"], width: '92%', columnWidth: 'MAX', //表格单元宽度，"MAX"最大内容显示，""AVG"平均，"AUTO"自动，指定[15,16,20] id: tableid, data: [ ["Core Ultra 9 285K", "8P+16E (24)", "3.7 / 3.2", "5.4 / 4.6", "5.7GHz", "5.6GHz", "5.5GHz", "125W"], ["Core Ultra 9 285", "8P+16E (24)", "2.5 / 1.9", "5.3 / 4.6", "5.6GHz", "5.5GHz", "5.4GHz", "65W"], ["Core Ultra 9 285T", "8P+16E (24)", "1.4 / 1.2", "4.7 / 4.5", "-", "5.4GHz", "5.3GHz", "35W"], ["Core Ultra 7 265K", "8P+12E (20)", "3.9 / 3.9", "5.2 / 4.6", "-", "5.5GHz", "5.4GHz", "125W"], ["Core Ultra 7 265KF", "8P+12E (20)", "3.3 / 3.3", "5.2 / 4.6", "-", "5.5GHz", "5.4GHz", "125W"], ["Core Ultra 7 265", "8P+12E (20)", "2.4 / 1.8", "5.1 / 4.6", "-", "5.3GHz", "5.2GHz", "65W"], ["Core Ultra 7 265F", "8P+12E (20)", "2.4 / 1.8", "5.0 / 4.5", "-", "5.3GHz", "5.2GHz", "65W"], ["Core Ultra 7 265T", "8P+12E (20)", "1.5 / 1.2", "4.6 / 4.5", "-", "5.3GHz", "5.2GHz", "35W"], ["Core Ultra 5 245K", "6P+8E (14)", "4.2 / 3.6", "5.0 / 4.6", "-", "-", "5.4GHz", "125W"], ["Core Ultra 5 245KF", "6P+8E (14)", "4.2 / 3.6", "5.0 / 4.6", "-", "-", "5.4GHz", "125W"], ["Core Ultra 5 245", "6P+8E (14)", "TBC", "TBC", "-", "-", "TBC", "65W"], ["Core Ultra 5 235", "6P+8E (14)", "TBC", "TBC", "-", "-", "TBC", "65W"], ["Core Ultra 5 225", "6P+4E (10)", "3.3 / 2.7", "4.7 / 4.4", "-", "-", "4.9GHz", "65W"], ["Core Ultra 5 225F", "6P+4E (10)", "3.3 / 2.7", "4.7 / 4.4", "-", "-", "4.9GHz", "65W"], ] });TVB = Thermal Velocity Boost，1-Core；TBMT 3.0 = Turbo Boost Max Technology 3.0，2-Core；TBT 2.0 = Turbo Boost Technology 2.0，1-Core。

已修复的问题

英特尔Arc显卡产品：

《三角洲行动》（DX12）在游戏过程中可能会出现应用程序不稳定的问题。

《古墓丽影：暗影》（DX11）在游戏过程中可能会出现树叶损坏溃。

内置英特尔Arc核显的酷睿Ultra产品：

《战舰游戏》（DX11）在游戏过程中可能会出现纹理损坏。

已知的问题

英特尔Arc显卡产品：

《毁灭战士：永恒》（VK）在游戏菜单和游戏过程中，可能会出现间歇性闪烁的画面损坏。

Blender在渲染某些基准场景时可能会出现应用程序崩溃。

Topaz Video AI在使用某些模型进行视频增强时可能会遇到错误。

英特尔酷睿Ultra处理器产品：

《雾锁王国》（VK）在游戏过程中可能会出现应用程序崩溃。

《神秘海域：盗贼遗产合集》（DX12）加载到游戏后可能会出现应用程序崩溃。

《暗黑破坏神4》（DX12）在启用光线追踪设置时可能会遇到应用程序崩溃。

在执行Alt + Tab操作时，《堡垒之夜》可能会崩溃。

使用HQAO设置时，CATIA文本可能不会出现在质量工具栏中。

SPECworkstation3.1在运行某些工作负载时可能会出现系统不稳定。

Procyon AI在使用Float32精度运行基准测试时可能会遇到应用程序崩溃。

该版驱动程序适用于酷睿Ultra（Meteor Lake）和第11/12/13/14代酷睿处理器的核显（Tiger Lake / Rocket Lake / Alder Lake / Raptor Lake / Raptor Lake Refresh），以及DG1 / 锐炫A310 / A380 / A580 / A750 / A770 / A770M / A730M / A570M / A550M /A530M /A370M / A350M独立显卡。英特尔Arc显卡驱动程序31.0.101.5971 beta（Windows 10/11 64-bit），下载地址：点此前往>>>

据TrendForce报道，台积电一方面兴建新的封装设施，另一方面寻求收购，以尽快提升先进封装产能。最近台积电就与群创光电签订了合同，购买后者位于中国台湾南部科技园的工厂和相关设施。其建筑总面积超过9.6万平方米，成交金额为171.4亿新台币（约合人民币38.21亿元），远低于传闻里200多亿新台币的市场价。虽然台积电在同一园区也有其他工厂，但是台积电的厂房位于园区西北处，而这次购入的群创光电厂房位于西南处，两者并不相邻。

最近一段时间里，群创光电开始积极探索转型的可能性，于2023年底关闭了5.5代生产线，不过最终选择将工厂出售给台积电。有业内人士表示，这笔交易对双方都有利，群创光电通过出售闲置厂房资产获得了额外收入的支持，而台积电则快速获得现有厂房，将有助于加速扩充先进封装产能，减缓产能吃紧的状况。

此前台积电还从瀚宇彩晶购入了一家工厂，最后选择拆除重建，因为原来的规划与台积电的需求不相符。此外，台积电在南部科技园还购买了益通光能的一处工厂，目前被作用智能仓库。

TomsHardware表示，英特尔通过客户端产品阵容的实力和多样性，继续主导客户端PC市场，而AMD在服务器市场上继续推进，创下了市场份额新高。AMD在服务器市场以24.1%的市场份额赚取了28亿美元，而英特尔75.9%的市场份额获得的收入是30亿美元，这意味着AMD销售的是更为高端的产品，平均售价远高于竞争对手。

与去年同期相比，AMD在桌面、移动、以及服务器的市场份额分别同比增长了3.6%、3.8%和5.6%。

预览var tableid = settableid(); document.write(''); Table().init({ title: "2024年第一季度桌面处理器市场份额", header: ["季度","2Q24","1Q24","4Q23","3Q23","2Q23","1Q23","4Q22","3Q22","2Q22","1Q22","4Q21","3Q21","2Q21","1Q21","4Q20","3Q20","2Q20","1Q20","4Q19","3Q19","2Q19","1Q19","4Q18","3Q18","2Q18","1Q18"], width: '92%', columnWidth: 'MAX', //表格单元宽度，"MAX"最大内容显示，""AVG"平均，"AUTO"自动，指定[15,16,20] id: tableid, data: [ ["AMD桌面市场份额","23.0%","23.9%","19.8%","19.2%","19.4%","19.2%","18.6%","13.9%","20.6%","18.3%","16.2%","17.0%","17.1%","19.3%","19.3%","20.1%","19.2%","18.6%","18.3%","18%","17.1%","17.1%","15.8%","13%","12.3%","12.2%"], ["季度增长","-1.0%","+4.1%","+0.6%","-0.2%","+0.2%","+0.6%","+4.7%","-6.6%","+2.3%","+2.1%","-0.8%","-0.1%","-2.3%","+0.1%","-0.8%","+0.9%","+0.6%","+0.3%","+0.3%","+0.9%","0%","+1.3%","+2.8%","+0.7%","+0.1%","+0.2%"], ["年度增长","+3.6%","+4.7%","+1.2%","+5.3%","-1.2%","+0.9%","+2.4%","-3.1%","-3.1%","+3.5%","-1.0%","-3.1%","-3.1%","-2.1%","+0.7%","+2.4%","+5%","+4.8%","+4.9%","+3.8%","+2.1%","+1.2%","+0.8%","+2.1%","+1.8%","-"], ] });预览var tableid = settableid(); document.write(''); Table().init({ title: "2024年第二季度移动处理器市场总份额", header: ["季度","2Q24","1Q24","4Q23","3Q23","2Q23","1Q23","4Q22","3Q22","2Q22","1Q22","4Q21","3Q21","2Q21","1Q21","4Q20","3Q20","2Q20","1Q20","4Q19","3Q19","2Q19","1Q19","4Q18","3Q18","2Q18"], width: '92%', columnWidth: 'MAX', //表格单元宽度，"MAX"最大内容显示，""AVG"平均，"AUTO"自动，指定[15,16,20] id: tableid, data: [ ["AMD移动市场份额","20.3%","19.3%","20.3%","19.5%","16.5%","16.2%","16.4%","15.7%","24.8%","22.5%","21.6%","22.0%","20.0%","18.0%","19%","20.2%","19.9%","17.1%","16.2%","14.7%","14.1%","13.1%","12.2%","10.9%","8.8%"], ["季度增长","+1.0%","-1.0%","+0.8%","+2.9%","+0.3%","-0.2%","+0.7%","-9.1%","+2.3%","+0.9%","-0.4%","+2.0%","+1.9%","-1.0%","-1.2%","+0.3%","+2.9%","+0.9%","+1.5%","+0.7%","+1%","+0.9%","-","-","-"], ["年度增长","+3.8%","+3.1%","+3.9%","+3.8%","-8.3%","-6.3%","-5.2%","-6.3%","-0.4%","+4.8%","+4.4%","+2.6%","+1.8%","+0.01%","+1.1%","+2.8%","+5.5%","+5.8%","+3.2%","+4%","+3.8%","+5.3%","-","-","-"], ] });预览var tableid = settableid(); document.write(''); Table().init({ title: "2024年第二季度服务器处理器市场总份额", header: ["季度","2Q24","1Q24","4Q23","3Q23","2Q23","1Q23","4Q22","3Q22","2Q22","1Q22","4Q21","3Q21","2Q21","1Q21","4Q20","3Q20","2Q20","1Q20","4Q19","3Q19","2Q19","1Q19","4Q18","3Q18","2Q18"], width: '92%', columnWidth: 'MAX', //表格单元宽度，"MAX"最大内容显示，""AVG"平均，"AUTO"自动，指定[15,16,20] id: tableid, data: [ ["AMD服务器市场份额","24.1%","23.6%","23.1%","23.3%","18.6%","18.0%","17.6%","17.5%","13.9%","11.6%","10.7%","10.2%","9.5%","8.9%","7.1%","6.6%","5.8%","5.1%","4.5%","4.3%","3.4%","2.9%","3.2%","1.6%","1.4%"], ["季度增长","+0.5%","+0.5%","-0.2%","+4.7%","+0.6%","+0.4%","+0.1%","+3.6%","+2.3%","+0.9%","+0.5%","+0.7%","+0.6%","+1.8%","+0.5%","+0.8%","+0.7%","+0.6%","+0.2%","+0.9%","+0.5%","-0.3%","-1.6%","+0.2%","-"], ["年度增长","+5.6%","+5.6%","+5.5%","+5.8%","+4.7%","+6.3%","+6.9%","+7.3%","+4.4%","+2.7%","+3.6%","+3.3%","+3.7%","+3.8%","+2.6%","+2.3%","+2.4%","+2.2%","+1.4%","+2.7%","+2%","-","+2.4%","-","-"], ] });

if (IsMobile()) { jQuery("iframe").attr("width", "100%"); jQuery("iframe").attr("height", jQuery("#post_body").width() * 9 / 16); } else jQuery("iframe").attr("height", jQuery("#post_body").width() * 9 / 16 + 120);手机通道：哔哩哔哩

在PSN港服上，《黑神话：悟空》标准版售价为港币298元、数字豪华版售价为港币368元。标准版包含了《黑神话：悟空》完整游戏，数字豪华版在标准版基础上增加了兵器（铜云棒）、披挂（百戏傩面/百戏衬钱衣/百戏护手/百戏吊腿）和珍玩（风铎）等。

目前《黑神话：悟空》全球游戏媒体的评分已经放出，像IGN中国给出了10分的评价，海外的IGN给的是8分评价，另外PC Gamer给的是8.7分、Wccftech是8.5分、GameSpot给出了8分等，而Metacritic的评分暂时为82分。

目前索尼PlayStation在京东自营官方旗舰店开启了夏日特惠活动，PS5轻薄版游戏机（国行）迎来限时降价200元。其中光驱标准版优惠价3399元（原价3599元），数字版优惠价2799元（原价2999元）。此外，PS VR2头显也迎来了优惠，国行优惠价为3299元（原价4499元）。

S790MAX SSD 4TB采用了PCIe 4.0 x4接口，M.2 2280规格，双面PCB布局设计，搭载了英仞科技IG5236主控，低耗高效，搭配原厂优选3D TLC NAND闪存芯片，支持NVMe 1.4规范、4KLDPC、AES加密、E2E数据保护、硬件RAID 5/6数据保护等，配备了最2GB的独立缓存。其最大顺序读取速度为7400 MB/s，最大顺序写入速度为6400 MB/s，最大TBW为2560TB。

梵想为S790MAX SSD 4TB经过协同特殊设计的复合架构优化，搭配石墨铝箔散热贴片，实现有效降温，让SSD持久稳定。此外，梵想还引入了Auto-自适应温度调节和电源能效管理两大智能管理系统，可以更好地调节SSD的温度，有效控制功耗，从而进一步降低SSD的发热量，减轻设备的负担，并延长笔记本电脑等设备的续航时间。

除了常规的Windows和Linux操作系统，S790MAX SSD 4TB还支持安装到索尼的PlayStation 5游戏主机，提升存储容量，加快应用程序加载性能，让游戏运行更顺畅。

据Globes报道，Arm在以色列的办事处正在针对游戏应用开发GPU，雇佣了100多名芯片和软件工程师。消息人士透露，Arm正致力于游戏图形应用，同时也会兼顾人工智能（AI）方面的处理，未来或许会与英伟达竞争。此外，该开发团队还负责为包括Vulkan和OpenGL在内的外部图形引擎开发人员开发软件接口。

目前Arm为SoC提供了旗舰级的Immortalis GPU，另外还有更为实惠的主流级Mali GPU，主要是为智能手机等移动设备设计，不过其中也提供了一些桌面芯片的功能，比如光线追踪和名为Arm ASR（Arm Accuracy Super Resolution）的游戏画面超分采样技术。

有业内人士推测，Arm可能会有推出独立显卡使用的GPU，或许是针对Windows on Arm设计的GPU模块。现在绝大多数独立显卡使用的GPU仅适用于x86处理器，这不是一个简单的硬件设计和驱动程序安装工作，需要耐心和大量的维护和优化。

随着高通发布了骁龙X系列处理器，Arm架构扩大了在笔记本电脑领域的应用，下一步可能会出现在台式机。对于图形性能的要求也变得越来越高，与Arm处理器适配的独立显卡在市场需求方面也变得更加迫切。

其实上个月Sapphire Rapids Refresh就被发现已获得了Ubuntu认证，而联想即将推出的ThinkStation P5就搭载了至强W7-2595X处理器。其拥有26核心52线程，睿频可达4.8GHz，TDP为250W，是至强WX-2500系列最高规格的一款产品。

预览var tableid = settableid(); document.write(''); Table().init({ title: "英特尔至强Sapphire Rapids Refresh工作站处理器参数", header: ["型号", "核心/线程数量", "基础频率", "加速频率", "L3缓存", "TDP"], width: '92%', columnWidth: 'MAX', //表格单元宽度，"MAX"最大内容显示，""AVG"平均，"AUTO"自动，指定[15,16,20] id: tableid, data: [ ["Xeon W9-3595X", "60C/120T", "TBC", "4.6 GHz", "112MB", "350W"], ["Xeon W9-3575X", "44C/88T", "2.2 GHz", "TBC", "97.5MB", "340W"], ["Xeon W7-3565X", "32C/64T", "2.5 GHz", "TBC", "82.5MB", "335W"], ["Xeon W7-3555", "28C/56T", "2.7 GHz", "TBC", "75MB", "325W"], ["Xeon W7-3545", "24C/48T", "2.7 GHz", "TBC", "67.5MB", "310W"], ["Xeon W5-3535X", "20C/40T", "2.9 GHz", "TBC", "62.5MB", "300W"], ["Xeon W5-3525", "16C/32T", "2.9 GHz", "TBC", "45MB", "290W"], ["Xeon W7-2595X", "26C/52T", "2.8 GHz", "4.8 GHz", "48.75MB", "250W"], ["Xeon W7-2575X", "22C/44T", "3.0 GHz", "4.8 GHz", "45MB", "250W"], ["Xeon W5-2565X", "16C/32T", "3.2 GHz", "4.8 GHz", "37.5MB", "240W"], ["Xeon W5-2555X", "14C/28T", "3.3 GHz", "4.8 GHz", "33.75MB", "210W"], ["Xeon W5-2545", "12C/24T", "3.5 GHz", "4.7 GHz", "30MB", "210W"], ["Xeon W5-2535", "10C/20T", "3.5 GHz", "4.6 GHz", "26.25MB", "185W"], ["Xeon W3-2525", "8C/16T", "3.5 GHz", "4.5 GHz", "22.5MB", "175W"], ] });至强WX-3500系列最多配备60个核心，共120线程，支持八通道DDR5-4800内存，而至强WX-2500系列的定位会稍微低一些，最多配备26核心56线程，支持四通道DDR5-4800内存。