• 瀏覽: 7,925
  • 回覆: 32
[隱藏]
引用:
原帖由 令狐椒 於 2019-8-17 01:33 PM 發表


AMD CPU 3年limited warranty. 3年後壞, 都可以叫 within spec, design.
In order to warrant 3 years of normal operation the chip must have been designed to work well past 3 years. You never target your design to the mean of the expected distribution. Chip designers are way more concerned about not having design flaws than software designers because of the huge cost of doing respins.



熱賣及精選
引用:
原帖由 hamsterman 於 2019-8-17 06:34 PM 發表

In order to warrant 3 years of normal operation the chip must have been designed to work well past 3 years. You never target your design to the mean of the expected distribution. Chip designers ar ...
我的火牛就係設計成十年保養.
哩啲先係 well past 三年設計.
不過, 火牛講明不應在80度工作.



引用:
原帖由 令狐椒 於 2019-8-17 07:33 PM 發表


我的火牛就係設計成十年保養.
哩啲先係 well past 三年設計.
不過, 火牛講明不應在80度工作.
You are comparing apples and oranges. Power supply wires are so thick compared to ICs that they don’t have EM issues. Chip design does not have  unlimited resources.To guard against potential EM violations in chip design one needs to use extra resources, resources that could otherwise be used to improve other aspects of the chip. Higher voltage, increased frequency, hotter temperature, and smaller feature size all contribute to worse EM conditions, but performance is what sells chips. There is no incentive to design for a 10 year life span when chip generation turn over every few years. Of all the different things that worry designers at chip tapeout, EM is not at the top of the totem pole. If one is concerned about cpu longevity, don’t overclock the thing.

[ 本帖最後由 hamsterman 於 2019-8-17 10:49 PM 編輯 ]



引用:
原帖由 hamsterman 於 2019-8-17 10:41 PM 發表

You are comparing apples and oranges. Power supply wires are so thick compared to ICs that they don’t have EM issues. Chip design does not have  unlimited resources.To guard against potentia ...
power supply 都有IC, 電路板, 電容. 全部都受高溫影響.
power supply 80度隨時都仲可以運作一段時間, 不過不是十年吧.



回覆 引用 TOP

[隱藏]
根據我既經驗如果部機係近乎長開,任何零件超過70C都係唔OK


CPU高熱燒既九成九唔係CPU,而係主板上既mosfet或電容



回覆 引用 TOP

引用:
原帖由 hamsterman 於 2019-8-17 08:24 AM 發表



Chips are typically designed for EM for the expected lifecycle of the product to minimize field failure. 80C is well within design specs, which typically guarantees normal operation between around ...
Talking to him is a waste of time. He's an AMD hater. 80 degrees Celsius is well within safe margins, and it's a full load temperature, so it's not even a regular use temperature.
Everyone around the world is shifting to AMD already. Even Google data centres are moving to AMD's Epyc chips. If that doesn't spell out the story of where the world is heading, then nothing will, but there are people like him that hold onto legacy beliefs from the early 2000s.



引用:
原帖由 channy911 於 2019-8-18 12:34 AM 發表



Talking to him is a waste of time. He's an AMD hater. 80 degrees Celsius is well within safe margins, and it's a full load temperature, so it's not even a regular use temperature.
Everyone around ...
I’m neither for nor against AMD, I actually worked for them a while back. I work in the chip design industry and many of my colleagues are present or ex AMDers.



引用:
原帖由 hamsterman 於 2019-8-18 12:58 AM 發表

I’m neither for nor against AMD, I actually worked for them a while back. I work in the chip design industry and many of my colleagues are present or ex AMDers.
師兄是有份設計ryzen cpu?
以你的專業見識, 新ryzen3000 的 RDRAND 指令bug是不是冇得用AGESA修理? 要等respin所以咁耐都未修好 RDRAND BUG?



引用:
原帖由 令狐椒 於 2019-8-18 02:59 AM 發表


師兄是有份設計ryzen cpu?
以你的專業見識, 新ryzen3000 的 RDRAND 指令bug是不是冇得用AGESA修理? 要等respin所以咁耐都未修好 RDRAND BUG?
Nope, I was completely not involved. My time with AMD was ancient history. I also don’t follow news on whatever bugs they’re having. However, I would say that as a general rule, chip makers will avoid redo-ing masks as much as possible because it is very costly. If they can find a way around the problem in software they would. Also, if they have to respin, they would wait to collect and resolve as many issues as possible and fix them in the same rev, because, again, each spin cost a lot of money. Actually some fixes may cost more than others to implement and that comes into the decision making as well. They will evaluate the trade offs between fixing something and the cost of fixing them, just like any other business would. 



[隱藏]
唔超頻及空氣流通的情況,所有原廠風扇都夠力!


回覆 引用 TOP

引用:
原帖由 hamsterman 於 2019-8-18 04:39 AM 發表



Nope, I was completely not involved. My time with AMD was ancient history. I also don’t follow news on whatever bugs they’re having. However, I would say that as a general rule, chip m ...
多謝師兄專業解釋.
似乎, 等一個stepping先買比較理想. 不過唔知等到幾時.



回覆 引用 TOP

原裝散熱唔開PBO 夠用


Intel 個散熱一樣咁廢, 9400 用電仲比 3600 多少少, 都夠用

CPU 溫度呢樣野, 同一款先有得比較
同一粒CPU die 唔同位置溫度都唔一樣, 差到二十度以上

[ 本帖最後由 wilsonkf 於 2019-8-18 06:07 PM 編輯 ]



引用:
原帖由 hamsterman 於 2019-8-17 10:41 PM 發表



You are comparing apples and oranges. Power supply wires are so thick compared to ICs that they don’t have EM issues. Chip design does not have  unlimited resources.To guard against potentia ...
識用啲半導體行業術語("tapeout"), 你應該係做過晶片設計. Hello, do you remember me?



引用:
原帖由 今天懷舊的水樽 於 2019-8-18 06:21 PM 發表

識用啲半導體行業術語("tapeout"), 你應該係做過晶片設計. Hello, do you remember me?
的確係呢行打雜混過。你係咪同我討論過noise source果位師兄?



[隱藏]
引用:
原帖由 wilsonkf 於 2019-8-18 06:03 PM 發表

原裝散熱唔開PBO 夠用


Intel 個散熱一樣咁廢, 9400 用電仲比 3600 多少少, 都夠用

CPU 溫度呢樣野, 同一款先有得比較
同一粒CPU die 唔同位置溫度都唔一樣, 差到二十度以上 ...
Ryzen gen3 唔係單晶片,因為總有段差所以係用墊過heatspreader,散熱能力比intel更低(intel都唔係特別好),而且如果d墊太熱恐怕會隨時間變質(intel用散熱膏都有同樣既問題,但因為溫差比較小一般耐用得多)

雖然咁既損壞修理唔難,但感覺上佢始終係粒為左減成本既同時削弱左可靠性既CPU

如果部機長開CPU超過80C無論係intel定amd接heatspreader既料都應該會在幾個月內烤壞



提示:支持鍵盤翻頁左 右
[按此隱藏 Google 建議的相符內容]