Thank you for the English subtitles! You guys do a GREAT work going thoroughly through the minutia of each chip and always gives us a detailed view, specially on power efficiency of these new portable chips.
Hello. Thank you for another great & thorough review! Could you please test LPE vs E cores power & efficiency curve? This way we could somewhat compare TSMC N6 and intel4.
he said that later on in the video that the lpe cores don't get used unless in idle situations. Also wikichip fuse has a comparison about the intel4 node. Node is pretty good, intel's implementation not so much. Maybe it'll be fixed by a firmware update but I'm losing hope
I'm holding out hope that intel is having firmware troubles with meteor lake. AMD has been getting too comfortable in the laptop space lately, I can't find find many good amd laptops for me. Intel needs this to succeed. Good review nonetheless. Can't wait for the English version.
I don't know if subtitles weren't available at the time of your comment, but now they are and the video is very informative. Also, on what you said about firmware having problems, I'm pretty sure that some of it is actually the case for Meteor Lake with its disaggregated tiled design.
Maybe Intel is misusing the LPE-cores while waiting to make the SoC tile in house on Lunar or Arrow Lake with its 18A process and being able to say it has improved LOADS instead of using TSMC (as Meteor Lake is using at the moment). Just regular corporate business.
Intel says its scheduling strategy used to send all tasks through the performance core and if it was small, it would be sent to the efficiency core, although it would only make sense if it was the other way around. Now they are saying it is using primarily the Low-power efficiency core on the SoC tile to decide if the task needs the CPU tile, but obviously it's just as shady as before, seeing that you guys noticed that the LPE-core is misused on low power scenarios. That's kinda stupid, and the performance and efficiency on low level usage could be so much better.
It's kinda sad now that my best alternative to Anandtech barely reviewing timely tech on time anymore is now reading subs on youtube lol. Geekerwan is great, but free in depth articles are really decimated these days as youtube takes over, when most of them are just flashy benchmark channels. Techreport is like a weird shopping blog now and articles by Jeff Kampman were retroactively changed to be authored by whoever the heck that deals blogger is. They completely killed that site after the buyout and killed the community. Anyways I agree with his conclusions, when it works it really seems to work, there's lots of potential in here. It might need some more Windows updates for more optimal core parking so it can turn the compute tile off more. Battery life tests range from not improved much to very improved, I guess depending on how much of the test can park on the SoC tile.
But now I'm really pessimistic about optimization because the scheduling of the 13th generation CPUs is still bad. I don't know how much longer we have to wait
一樣類似的內容,但每次極客灣的圖說跟解說順序總是最能讓我們這種外行人最快理解的,點讚。
Thank you for the English subtitles! You guys do a GREAT work going thoroughly through the minutia of each chip and always gives us a detailed view, specially on power efficiency of these new portable chips.
終於等到了,未看先推
You guys do an amazing job reviewing these products with such depth and accuracy, please revive the English channel as well :P
Hope to see you guys' review of the Snapdragon X Elite next year.
终于更新了
來催個更 大狸子送的steam deck OLED的影片.......
希望早日能見
但有新影片就已經很欣慰了
加油 辛苦了
希望15代桌面端能讓I3 I5帶核顯型號用上這種等級的核顯 多加一點錢也很值
明显是调度有问题,部分oem昨天更新了一批补丁,同功耗下r23提升了10%左右,后续明年第二批机型表现预计会好不少
Hello. Thank you for another great & thorough review! Could you please test LPE vs E cores power & efficiency curve? This way we could somewhat compare TSMC N6 and intel4.
he said that later on in the video that the lpe cores don't get used unless in idle situations. Also wikichip fuse has a comparison about the intel4 node. Node is pretty good, intel's implementation not so much. Maybe it'll be fixed by a firmware update but I'm losing hope
@@nidalspam509 At 9:40 they managed to measure SPEC2017 scores of LPE cores so it should be possible to get at least some data on power too.
Great idea!
Will we get a review of google tensor g3? Awesome videos btw
Windows下另一個問題,Office 365還是間中出現大小錯配被後卡卡卡。
老一點的軟體你基本上還是要手動調核。
Windows Update還是單線。
效能再好Windows 10升11那堆動畫就把吃回去。
當然上面上的所有問題都可以透過調整修改更改Windows的設定或者使用第三方軟體修改。
但是我就是需要一台買回來爽快省電好用的手提電腦。
rog自带兼容模式。开启后,就是禁用小核心
我不相信短时间Microsoft能把Windows调好,兼容性太强就是难以解决这些问题。所以直接准备买全大核的AMD了😂
同疑惑视频中图表标题下方的编译参数,指的是 aarch 环境下交叉编译得到 x86 产物?但是 -mcpu=native 又说不通,如果没有完整开启对应的指令集加速,测试速度可能会差很多。有条件可以考虑放出测试用的完整 cmake 和源码,这样所有人都能复现结果
期待牙膏廠和農廠能更好的讓對方廠減價
如果都用台积电工艺,感觉台积电会给涨价
极客湾的续航测试模型有没有公开版呀,觉得这个才是靠谱的测试办法.PCMark8/10这些东西都是看着好看,用起来两码事.导致我们还是习惯按10Wh/小时算.大佬上架卖也行哇
專業
感覺 7840hs 性價比很高啊
Ultra 5 其實也不錯 就是調度不夠聰明
Meteor Lake looks great. Will buy new laptop with it.
Intel also is updating the firmware to make it even better.
Thank you!
I'm holding out hope that intel is having firmware troubles with meteor lake. AMD has been getting too comfortable in the laptop space lately, I can't find find many good amd laptops for me. Intel needs this to succeed. Good review nonetheless. Can't wait for the English version.
Yes it has softwares issues.
Acer swift go got 13 hours and 19 minutes in PC mark battery test.
It will get better even more.
I don't know if subtitles weren't available at the time of your comment, but now they are and the video is very informative. Also, on what you said about firmware having problems, I'm pretty sure that some of it is actually the case for Meteor Lake with its disaggregated tiled design.
第一 Ultra真的很有潛力
期待看到桌面版
这种新产品 一般第三代才是完全体 除非搞颠覆来的
能有15% 提升已經擠爆牙膏了😂😂😂
@@gmjacky以前都是5%的😂
Soc你要他上pc喔
真不知道你们这些营销小号🖕有没有看视频,Intel4都开始倒吸牙膏了居然还期待上桌面。怎么?你们这些卖货的原来都是用核显打游戏?最搞笑的是核显还是用tsmc N5,你要觉得这个进步大去买amd和Nvidia的显卡啊
Lunar lake 真的好像M2/M3 soc, looking forward to it !!
继承了发哥曾经的优良传统:大核跑不停,小核围观。
一核有難,多核圍觀
有个疑惑,为什么spec2017的编译环境是aarch64而不是x86_64? 跑性能测试的时候如果要过一层从x86到arm的转译感觉不能体现原本的性能,除非这是有什么缘由的,有哪位大佬可以解释一下吗
一个字,专业!
还是极客湾的视频最有价值,这是目前唯一一个详细对比了上代和A家的视频。
intel的思路是對的,在移動時代,與其一昧追求高性能,還不如在維持效能的同時改善功耗比
說壞的話,可以說是AMD的堆核戰術而導致主流平台的核心線程數目得到劇烈程度增長,那會想到在流動領域也可以達成32核線程的怪獸級平台🙉
和amd 7840u比,能效和性能双输。。。
笑死 能耗比打過amd了嗎 pc跟筆電一樣輸慘慘 難怪amd不急先搞一個馬甲 價格打不過人家 效能也只能跟別人打平 請問除了品牌黨誰要買
zen4 desroyed this crap. how about next year's Zen5? huh? + RDNA4 with enhanced Ray tracing. AMD will own PC world
Some tested the ultra 7 vs 7840hs, this is my first seeing the ultra 5 vs amd's and wow they still beat it in iGPU performance o.o
哇,刚发了几秒钟我就刷到了
想看NPU兩廠,算力比較
很高興 Intel 推出 4 製程的產品
期待未來推出桌上型產品
Intel 4菜的口交,除非大幅削减密度,不然还是继续改进intel7提升性能吧
@@jameswoolf1440 我覺得 X86-S 推出會幫助更大,只是目前只存在於白皮書 🙃
@@jameswoolf1440 10nm在怎麼提升還是10nm繼續打磨只會重演當年,14nm++++
英特尔现在工艺制程方面研发成绩相当出色,但新工艺频率性能上不去的老毛病一下还难以去掉。@@ABABBAB
恭喜百大
低耗電核在那些工作下會調用?
Maybe Intel is misusing the LPE-cores while waiting to make the SoC tile in house on Lunar or Arrow Lake with its 18A process and being able to say it has improved LOADS instead of using TSMC (as Meteor Lake is using at the moment). Just regular corporate business.
所谓NPU就是可以快速执行低精度,和显卡的那些一样,只不过单独拿出来堆切大规模
還是APU厲害
感觉这次ultra lake 性能不及预期啊😢
专业
這代煙頭的內顯效能,不難看到是得益內存頻寬,如果換轉是改用D4 3200的話,分數應該就沒甚麼好看了🙊
不過這點對於AMD應該能感受到一點壓力,尤其是內顯方面,就看下代8000系列的整體表現如何🙊🙊
看之後改用DDR5-8000以上,內顯效能才有機會摸到GTX1650
想到公司沒幾個月前幫同事剛換的筆記本電腦,開會demo卡成狗,後來想想不太對,請他把電源接上就好了...一台用電池的筆記本電腦連開個網頁都卡... 基本上大家跑分測試都是插著電源在測的,用電池之後大概成績都很難看... Macbook Pro不管插不插電跑分或跑軟件測時成績都一樣。
毕竟apple家是自己的封闭生态,所有环节都可以为产品做出优化。这是x86+windows很难跟得上,众多硬件厂商的众多硬件,需要做到像macbook那样,需要花费的成本谁都不愿意负担的,好比,微软大投入紧跟红蓝厂的硬件变化做出无死角的优化,最终最大的获利者还是硬件厂商,微软赚个吆喝而已,很难去协同的。你提出的这个现象至少未来几年还是不会有什么改善的。
感覺這代intel把心力放比較多在能耗控制上,但調度策略顯然沒寫好
但是第一次看到核顯可以光追,性能也算不錯,算是文書機黨的福音呀
文书笔电以后也可以玩玩游戏了喜喜
@@yorkerW是阿 我的mx150能跑個cs2就偷笑了
第一次不至于吧,AMD的680M和780M也可以光追的
@@erasersc5509 原來 我只知道680m很不錯
只是台灣這邊很少看到標註680m的筆電就沒再關注了
@@mentlegenluo2911你不关注 RDNA2、3光追就没了? 视频里又不是没提。而且不止一次,甚至测了Path Tracing。都当没有。
用mcm封裝是能理解
但既然都用了Base tile 怎麼不考慮把dram 也封在一起呢
這不是更能提升記憶體頻寬跟大幅度降低延遲嗎
現在x86最大效能瓶頸不就在記憶體這塊
頻寬拉大對gpu來講更是利多
intel要弄革命性產品
東西就不要只弄一半
這樣只會變電子垃圾
消費者又須要等個兩三代才成熟
m1一出來令人驚艷
coreultra這樣我還是頭上三條線
Sir LP E cores are working for playing Netflix video playback by turning off CPU and iGPU Tile?
看完想買AMD,然而發覺 amd 7x40系列最大的問題大概是幾乎找不到幾台”不“搭載獨顯的高性價比筆電。不知道在搞啥
18:31 相比amd,牙膏廠在温度功耗上大勝。
核顯終於飛躍, 但intel 4 工藝不見得很大幫助
核顯就TSMC N5阿
核顯主要還是台積電加持的緣故
如果有超小型核能,所有的能源方案可以迎刃而解
代表intel 4根本沒進步吧,還不是靠台積電的GPU
Intel says its scheduling strategy used to send all tasks through the performance core and if it was small, it would be sent to the efficiency core, although it would only make sense if it was the other way around.
Now they are saying it is using primarily the Low-power efficiency core on the SoC tile to decide if the task needs the CPU tile, but obviously it's just as shady as before, seeing that you guys noticed that the LPE-core is misused on low power scenarios. That's kinda stupid, and the performance and efficiency on low level usage could be so much better.
ipc性能差不多的话,游戏本应该可以忽略这代处理器提升去买13代更有性价比
不是說要測試核外功耗嗎?
怎麼沒看到?還是我看走眼了?
只担心一个问题就是发热不一致会不会造成晶圆被散热片压坏
讚一個
其实真正的性能瓶颈,应该在windows😅
期待極客灣的ultra 7的評測,看起來效能似乎跟上了m2,續航也不差(dave2d那邊的評測
一眼假
@@jameswoolf1440 ??
所以 架構大改到認不出來 然後CPU基本沒提升?
等一手同款处理器的 LG gram 调教
我開始考慮amd 因為我看到你的衣服上
9w功耗的134u和164u的本还没出,到时候看9w续航如何,准备出手了
感觉这种模块化设计还不如恢复北桥,把真正负责通用计算的部分:核心、Ring、IMC封装在CPU上,而杂项如核显、NPU、显示输出、编解码器、雷电控制器、PCIE控制器之类的放在北桥,这样也不会有以前IMC在北桥导致的延迟
別天真了,imc在cpu,那你分離出來的北橋的內顯要使用記憶體還得透過cpu存取?這樣性能只會暴死
@@hazanoko 核显放北桥会增加延迟带宽基本不变,而核显需要的是带宽、对延迟不敏感 所以核显性能应该不受影响吧
这样的思路是
把好处优先给真正的“CPU”--通用计算的部分,
把影响不大的坏处
摊给能用就行的核显、各种控制器之类的杂项
你真的太天真了,現在的架構中,跟南橋連接的頻寬只有pcie gen4 x4,請問你pcie控制器、NPU、內顯都弄在傳統北橋的話,那cpu跟北橋中間的通訊佈線要怎麼佈?光顯卡就要吃一個gen 4 x 16了,一個雷電口吃個x4,再來個x4的M2,其他雜項加一加也x4,原本cpu對南橋只要x4的佈線,恢復北橋至少有7倍的複雜度,光想想就頭皮發麻,主板價格大概也突破天際。
正是因為cpu對北橋的頻寬要求越來越高,所以近代的cpu才會逐漸把北橋甚至部份南橋的功能併入。
我自己的總結是:
這代的CPU更傾向於創作者,一般使用者維持13代就好,因為在日常使用上會用到的核心差別其實不太大
毕竟这个价格而还没有应用在桌面端?感觉如果应用到桌面端之后性能可能会有个很夸张的提升
桌面端用处不大的,14代不是已经出了,这个设计成本巨高,而且大部分的人还是会使用cpu+gpu的方式吧,这个个人认为完全是在为降低功耗去考虑,桌面端基本没有低功耗的使用场景吧,感觉只有在minipc领域会有一些作用,这个感觉就像苹果设计m芯片的感觉差不多@@hjonson8721
主要還是續航力 筆電看的就是這個 這次提昇續航有感
適合做遊戲掌機
原来CPU基本没有性能提升呀。。。
问你一个问题,我的台式机现在有时候会突然卡主3秒左右,主机好像有个异常的声音,卡的时候,屏幕也是不动,鼠标键盘也不能动。但是几秒钟后又会恢复,并不会死机。cpu是8700
一个多月没更新了。。。
B站有
@@Vasfcxdff4027如果只想在B站更新不想在油管更新那就销号啊!区别对待是什么意思?
小中國蛙再50年也追不上😂😂
在墙内更新了,主要上个月的评测都是墙内产品
@@Vasfcxdff4027墙内也没更新一个月
真让我好等啊😂
所以调度问题,有没有微软的锅?
英特尔U9(悠久)处理器--在历史的时光流转中体验性能。 英特尔U9(悠久)处理器--在处理中感受时代变迁。
Intel Core Ultra 先進封裝
CPU intel4 7nm. GPU tsmc n5. Soc tsmc n6
AI的能耗比測試做不了嗎?
好奇這台是不是用低溫錫😂
比起遥遥领先怎么样?
問題是i家會否加入pc掌機的戰局
aarch64??intel 买了arm?
试一下把跑分窗口最小化
base tile這個製程是模仿台積的cowos,是fowlp的一種,也可能intel就是包給台積用cowos做的,intel自身的類似做法是emib,但emib和cowos差別極大,一眼就看出來不同
不是,base tile是Intel 22FFL,封装技术是Intel自己的foveros,base tile是active的不是passive的,封装是在Intel的马来西亚工厂做的,极客湾去参观过。emib的bridge是埋进封装的substrate基板里的,但foveros是直接在base tile的wafer里面光刻电路,要较真的话你可以管这个技术叫co-emib,Intel自己也是这么叫的,目测明年的lunar lake会用上这个,因为会把DRAM也封装进去,就是完整意义的co-emib了。
It's kinda sad now that my best alternative to Anandtech barely reviewing timely tech on time anymore is now reading subs on youtube lol. Geekerwan is great, but free in depth articles are really decimated these days as youtube takes over, when most of them are just flashy benchmark channels. Techreport is like a weird shopping blog now and articles by Jeff Kampman were retroactively changed to be authored by whoever the heck that deals blogger is. They completely killed that site after the buyout and killed the community.
Anyways I agree with his conclusions, when it works it really seems to work, there's lots of potential in here. It might need some more Windows updates for more optimal core parking so it can turn the compute tile off more. Battery life tests range from not improved much to very improved, I guess depending on how much of the test can park on the SoC tile.
But now I'm really pessimistic about optimization because the scheduling of the 13th generation CPUs is still bad. I don't know how much longer we have to wait
there seems to be that there is already an update out, which improves CPU peroformance and efficiency
酷睿Ultra目前拿不出手到桌面端处理器,不然不可能出14代依托大X拿出来卖。
15代 自带的NPU和 Movidius Myriad X 比 高级多少
下一代連架構製程一起改就有機會和高通打下蘋果的江山了
蘋果的市佔就那麼點,也算不上什麼江山吧!比較像是M1橫空出世,各家廠商的激烈競爭帶來的一些效應
ICU 終於來了
還是看一下zen 5會不會有32core 實在是不喜歡大小核 windows調度太烙賽
大核就是CPU 小核的話...沒CPU功能 所以就大核跑
如果是這樣的話 乾脆把大核再切開成四個小核吧 到時閒置時也只需開啟一個核心就好
電量用那麼高 根本不是好事...
怎么说呢,英特尔在这代酷睿的宣发PPT上面,拿自家的H后缀标压跟AMD的U后缀低压CPU对比,然后说强了12%还是多少,就很欺负小白。
实际上换算过来大概就是:这代酷睿移动端性能追上Ryzen 7000移动端。
所以我个人建议是等Windows 12出来以后再上新平台也不迟,毕竟微软自己说的下代Windows会以AI PC为基础进行开发,到时候NPU的利用率才会有明显提升。现阶段除了硬性需求以外,盲目上新平台实在是没必要。看似英特尔好像很成熟,实际上它也很难适应新的架构变化。包括这代酷睿,以及刚开始的Arc独显,都是更新了很多版驱动以后才变得完善的。以前之所以觉得英特尔稳定,不用操心驱动和BIOS等等问题,主要还是因为它反复沿用已经成熟的老事物(从1代酷睿i一直到6代、7代到11代一直没本质变化,那可不稳定嘛)。然而一旦他们自己下狠心换了新东西,就连他们自己都难以驾驭。
不止,AMD CPU的硬件质量问题很多。而且你说的挤牙膏问题,是CPU行业的潜规则,从还没有每年有规律地更新时就开始了,AMD也不例外,因为半导体没有那么多花活儿可以玩,所以对手的底牌不掏出来自己就可以跟着挤好几年,英特尔AMD英伟达高通皆是如此。
这代的ultra5/7的TDP是28W,和AMD7000系移动端U系列TDP是一样的
微软:累死我了。阿三:卧槽,不干了爱咋滴咋滴
说起这个 aochua ...就让我想起了以前犹如诈骗的 超级本...英特尔那几年 真是不当人 炒概念
怎么里面的彩蛋是猪王?
聯發科砍掉手機的小核,intel加入小核,現在是要黃金交叉嗎
應該說是 趨同演化 兩邊都意識到有個兼具續行與能效的甜蜜點 對未來大多用戶來說可能桌面與移動端沒這麼分明了
聯發科砍小核單純是arm提供的公版a510太爛 a710都能用更低功耗提供同樣性能 a510沒有存在的意義
感覺intel現在已經把重心放在AI上了,不知道到時候會不會兩頭燒
既然剪片那么强,不如和苹果打打试试?
下一步,进程,内存硬盘到上面。
win笔记本好多年没关注,那么多年过去了,续航也没多大改变,应该还是伴随着轰鸣声工作吧.......
这个新构架还是有搞头的,下一代估计会疯狂堆核。期待内存通道翻倍,做出超大规模GPU。
都倒吸牙膏了有个屁搞头,营销号能先看视频再来comment吗
@@jameswoolf1440根本沒倒吸,不同產品線的CPU是在比什麼,45W比28W,不贏才奇怪咧
@@Nasen6239 你是不是不知道片头测的IPC是什么?
@@jameswoolf1440請問您說的是14900k定頻那段還是?
功耗调度的评论有点想当然了,功耗调度这块儿,到底是让强力U更快的完成手上的任务更省,还是让稍弱的U慢慢的完成,一直是有争论的。
是不是linux驱动没优化好?
好强啊现在的笔记本😮
我的玩原神也只能最低画质60帧
不上4纳米以内的制程是不会买的
其實最想看的是對比m2和m3
幾個知名的做筆電評測的外國人已經有做了
硬件很強大 但不及不計成本的APPLE,調度還需要優化。太多奇怪的調度了。
Dude post something on your English channel ..it has been 3 months ..still waiting for 8 gen 3 vs A17 PRO
欧耶,来了