作者: 真衰鬼 時間: 2013-6-8 01:55 標題: Haswell比Ivy更熱情!
先講i7-4770K 超頻情況,以1.257V超至4.7GHz,用H100i散熱,溫度是……
98°C
垃圾,貴成咁,成100度,煲水都得了!
Overclocking
Turning our attention to our shiny new Core i7-4770K, we started by cranking up the multiplier to see how far we got without touching any other settings. We eventually hit a wall at 4.4GHz with the PC throwing a BSOD at us at 4.5GHz as soon as we entered Windows. However it seemed perfectly stable at 4.4GHz, which is very similar to what you'd expect from an Ivy Bridge CPU.
From here, though, the temperatures increased rapidly and we reached our limit of 4.7GHz using 1.257V and a scorching temperature of 98°C, and that's using a Corsair H100i, which we know to offer the best cooling short of a custom water-cooling kit. The jury is out on why this might be - the 22nm manufacturing process is clearly going to add a lot of heat density, as we saw with Ivy Bridge, but like Ivy Bridge, there's the issue of thermal paste vs solder too. We're not about to take a razor blade to our only test sample just yet but reports from the community on this will likely surface soon as to whether de-lidding is on the cards again.
http://www.bit-tech.net/hardware ... -4770k-cpu-review/8
為何Haswell 較Ivy更熱情?
因為……
Intel Core i7-3770K was 77W TDP and the Intel Core i7-4770K is now 84W TDP due to this. Intel also still does not solder the heatspreader (lid) to the top of the processor, so this will be interesting.
http://www.legitreviews.com/article/2203/14/
When it comes to load temperatures we fired up Prime 95 27.7 64-bit and saw the CPU core temperature rise to 86C with a CPU total TDP of around 75 Watts. The CPU temperature rose quickly, but eventually the fan profile kicked in and the fan speed increased enough to compensate for the heavy workload. No CPU Throttling was noted with the stock CPU cooler, but with temperatures hitting 86C, we highly advise aftermarket cooling if you plan on doing any overclocking as these new 22nm processors do run warm.

作者: 唐賓南紀念中學 時間: 2013-6-8 08:59
Intel also still does not solder the heatspreader (lid) to the top of the processor
ivy係咁, haswell都係咁. intel係唔係有心玩野?
作者: tofuon 時間: 2013-6-8 09:05
無計啦,這些年,INTEL高端市場都無競爭.
不過Haswell比Ivy令人更失望咁
作者: 車道士 時間: 2013-6-8 11:44
提示: 作者被禁止或刪除 內容自動屏蔽
作者: 真衰鬼 時間: 2013-6-8 11:47
唔超頻full load都80幾度,垃圾!

作者: sonymobile 時間: 2013-6-8 11:49
提示: 作者被禁止或刪除 內容自動屏蔽
作者: 美奈子廚 時間: 2013-6-8 12:06
根本吾係cpu熱,係吾夠位傳熱
cpu50度hs先30度
作者: lcska02nkc 時間: 2013-6-8 12:19
提示: 作者被禁止或刪除 內容自動屏蔽
作者: ms2e 時間: 2013-6-8 12:19
本帖最後由 ms2e 於 2013-6-8 12:24 編輯
之前intel不是說保證haswell會很省電嗎?怎74tdp 變84?
看來這家公司的橙信實在有問題
現在有點想anti intel 了。
作者: EyeOfMagic 時間: 2013-6-8 12:42
咁熱...
o係香港夏天唔開冷氣咪死得, 會唔會成部機燒左

都係睇定d先買......
作者: 真衰鬼 時間: 2013-6-8 14:51
四系TDP全部84W
http://ark.intel.com/products/family/75024
但Haswell只比Ivy效能高約10%
i5-3570K vs i5-4670k
i7-3770k vs i7-4770K
大家同是22nm製程,不禁令人聯想……廠家加電壓,出廠已OC的「Ivy」?

作者: win150 時間: 2013-6-8 16:17
本帖最後由 win150 於 2013-6-8 16:19 編輯
10%只係CPU, GPU高左廿幾%
張粒Ivy Bridge加下壓, O下C會唔會勁左廿幾%


作者: win150 時間: 2013-6-8 16:32
本帖最後由 win150 於 2013-6-8 18:11 編輯
佢應該係指mobile platform, For example 17W for Ivy Bridge U series CPU + 3W for QS77 chipset vs 15W for Haswell U series CPU+chipset
(15-20)/20 = 25% power savings 唔知係咪咁計
作者: 00006 時間: 2013-6-8 16:49
省電同TDP有d分別, TDP只係講粒CPU最高發放既熱力, 呢個數值係比散熱設計作為參考, 用電量係另一回事
用電量亦唔可以淨係睇粒CPU, 要連埋chipset成個平台計
Haswell用左更多芯片面積做GPU, die size大左, 自然發熱都多左, 但E3 1230v3呢d冇內顯既CPU應該TDP唔會提高?
可能因為整合左voltage regulator, 唔知呢種元件發熱多唔多
作者: jasonchan424 時間: 2013-6-8 16:55
其實100%對比75%係升左33.3%..
作者: Kundera 時間: 2013-6-8 17:35
本帖最後由 Kundera 於 2013-6-8 17:38 編輯
"There’s more fine grained power gating, lower chipset power and the CPU cores can transition between power states about 25% quicker than in Ivy Bridge"
http://www.anandtech.com/show/70 ... 0k-i54560k-tested/2
Thermal Design Power concerns the heat rather than power. It is a requirement for cooling system to assure the MAXIMUM amount of power to be dissipated in the worst loading scenario. By the same 22nm process, Haswell has a bigger GPU than Ivy Bridge, it makes full sense to an increase of TDP. But in real world scenario it will be less likely to have both CPU and GPU full throttle all the time, the power gating and voltage control trick described in reviews are more efficient than simply forcing the TDP down but having all components wild awake. The benefit is best illustrated in the drop in idle power consumption
作者: alfredxp 時間: 2013-6-8 17:51
用返IVY 都OK了.
作者: 00006 時間: 2013-6-8 20:37
"There’s more fine grained power gating, lower chipset power and the CPU cores can transition bet ...
Kundera 發表於 2013-6-8 17:35
正解! TDP唔等於用電量
不過都係解釋唔倒點解E3 1230v3 TDP會高過1230v2
既然大家disable左GPU, 咁應該GPU增加既晶片面積唔會影響TDP?
v3 TDP 80W, v2 TDP 69W
作者: 真衰鬼 時間: 2013-6-8 22:58
4670貴3570成200蚊是為了GPU+20%?

作者: 真衰鬼 時間: 2013-6-8 23:00
正解! TDP唔等於用電量
不過都係解釋唔倒點解E3 1230v3 TDP會高過1230v2
既然大家disable左GPU, 咁應該GPU增加既晶片面積唔會影響TDP?
v3 TDP 80W, v2 TDP 69W
00006 發表於 2013-6-8 20:37
我懷疑Intel 自家OC 炮製
作者: ak8 時間: 2013-6-8 23:12
估到
風冷輕鬆上5GHz作者: 真衰鬼 時間: 2013-6-8 23:14
On the face of it, you are right.
But I wrote research article on this topic and published on HKEPC
There is good correlation between power consumption and TDP of Intel product.
Look:

Source: http://cnbeta.com/articles/239553.htm
作者: 真衰鬼 時間: 2013-6-8 23:17
Haswell.... really rubbish
搭载GT3e的Core i7 4770R处理器
功 耗和温度方面,不得不说Haswell表现很糟糕,性能提升微弱的情况下,功耗却比Ivy Bridge得到了不少的提升,尽管引入了待机800MHz的更低运行频率以及更为高效的FIVR调节模块,但是收效非常甚微。另外备受DIY用户关注的 超频性能让人感慨:一代不如一代!
作者: 真衰鬼 時間: 2013-6-8 23:19
Indeed

作者: paultch 時間: 2013-6-8 23:43
對Haswell非常失望

作者: wilsonkf 時間: 2013-6-9 00:01
提示: 作者被禁止或刪除 內容自動屏蔽
作者: 00006 時間: 2013-6-9 00:17
人地講緊GPU

況且中階點同頂級比
作者: wing1100 時間: 2013-6-9 00:23
唔係掛,,,仲諗住換機
作者: wilsonkf 時間: 2013-6-9 00:24
提示: 作者被禁止或刪除 內容自動屏蔽
作者: 真衰鬼 時間: 2013-6-9 00:27
i5-4670 減價了

http://www.hkepc.com/forum/viewt ... ;page=1#pid30058045
之前i5-4670 $1,827
岩岩海鮮舫 $1,777
作者: qcmadness 時間: 2013-6-9 00:31
都係炒價
官價係US$213
作者: 真衰鬼 時間: 2013-6-9 00:36
炒唔起,即係……

作者: qcmadness 時間: 2013-6-9 00:41
1. Intel因為chipset bug唔會供大貨
2. 根本PC market已經唔係靠效能就推得郁
作者: aeolian2009 時間: 2013-6-9 00:56
這是新產品 煮飯CPU

作者: thomaswai 時間: 2013-6-9 00:57
咁熱情
搭單問下~2600K 同4770K 比..效能差多?
作者: Geek2012 時間: 2013-6-9 00:59
提示: 作者被禁止或刪除 內容自動屏蔽
作者: 飛起潛水員 時間: 2013-6-9 01:12
http://www.anandtech.com/bench/Product/836?vs=287好地地無必要轉
作者: 最強幻象 時間: 2013-6-9 01:23
仲用緊 i5 2500K, 幾時先值得我轉 >.<
作者: qcmadness 時間: 2013-6-9 01:26
有排都唔會
因為single-thread performance再升都無乜意思
作者: 最強幻象 時間: 2013-6-9 01:31
咁就要等 game 同 M$ 去搞大個 loading 先得了
作者: sancianyuanna 時間: 2013-6-9 01:40
但還在用i7-920的我想入手了
作者: albert_ma 時間: 2013-6-9 01:42
我覺得呢個係真相
如果IB-E用番solder,就證明到呢點
作者: qcmadness 時間: 2013-6-9 01:49
講真丫, client processor係好賺過server processor
作者: dom 時間: 2013-6-9 02:05
乜 Intel U 唔係會製冷既咩

作者: dom 時間: 2013-6-9 02:09
講真
Intel 比你班DIY 友超超超 超死完拿去 Claim 保你估真係唔駛cost

佢做商用客 好過理你 DIY 啦
作者: Kundera 時間: 2013-6-9 02:29
It just a reference for cooling design
Actual power consumption is to be measured
作者: 林洋2 時間: 2013-6-9 03:17
我諗住用SANDY成世,但驚到時壞板,冇晒1155板賣

作者: bobocar 時間: 2013-6-9 03:58
AMD 唔爭氣...Intel 出垃圾都要焗買

作者: MikuMeow 時間: 2013-6-9 04:50
本帖最後由 MikuMeow 於 2013-6-9 05:08 編輯
Haswell 呢代重點放喺iGPU TDP高咗就無可厚非 但喺你超頻都只喺超CPU 超完full load燒機都只喺CPU 同頻下竟然比ivy高溫咁多!?
10% performance increase 要咁熟?呢幾年intel啲TickTock真喺冇咩動力令人去升級 除咗要SSD/USB3.0 CPU performance喺Core之上改嚟改去都改到極限 再加上AMD Faildozer 唔爭氣 intel都好難開發新架構
好啦 完全確定1156又可以用多半年 等Kerveri
作者: fireeye 時間: 2013-6-9 10:17
打手文可以不理,實際無禁差。
作者: 真衰鬼 時間: 2013-6-9 14:49
咁,有冇D唱好文先?
作者: kcstudiohk 時間: 2013-6-11 16:35
我套 i7-920 可以玩到佢成個日先換機了...
反正都係咁熱... orz... ( now 3.6Ghz fullload 88度.. )
作者: ee9861 時間: 2013-6-11 16:57
但INTEL好似係計核心溫度

