作者: dull_dull 時間: 2017-9-6 20:33 標題: AMD Threadripper 的誕生，原來不在2017 product list
本帖最後由 dull_dull 於 2017-9-6 21:06 編輯
AMD Ryzen Threadripper: The Fascinating Story Behind The Processor That Beat Intel
It is very much a CPU made by enthusiasts, for enthusiasts. The story behind it hasn't really been told before as we announced and launched Threadripper this summer in a very short period of time. Before May 2017, no one knew it existed, and by August it was in people's hands (Google translate: 這是一個由愛好者為愛好者製作的CPU。 我們在今年夏天在很短的時間裡宣布並推出了Threadripper，之前的故事並沒有真正被告知。 在2017年5月之前，沒有人知道它已經存在，到八月份是在人們手中 )
Threadripper timeline:
2014 – 2015: Work began on Threadripper concept, mostly in the initial team's spare time and without a dedicated business plan
2016: The team presented the idea to Jim Anderson just before the Computex tradeshow who had joined from Intel the previous year. Loved the idea so much the launch timeline was cut to 18 months to be the same year as Ryzen and EPYC.
May 2017: AMD officially announces Threadripper, which comes mostly as a surprise to the press.
August 2017: Threadripper goes retail
好似電影情節
My boss was sharing a taxi with Jim on the way to Computex 2016 and mentioned Threadripper to him then. He loved it, gave it the green light and found a way to get it on the roadmap. He believed in the idea, especially the way it leveraged existing technology we were using for Ryzen and EPYC,
The existence of the EPYC server CPU enabled us to create Ryzen Threadripper, but Threadripper was not part of the original 2017 AMD product roadmap. That only included the Ryzen 3, 5 and 7 ranges of 4, 6 and 8-core CPUs, plus versions of EPYC ranging from eight to 32 cores.
I asked for two things – the specifications and also how quickly we could build it. I fell in love with it seeing the specifications, but the timeline was disappointing as it originally pointed at 2018 for launch. I immediately asked them to pull it in by about a year – so summer 2017.
作者: chiluns 時間: 2017-9-7 01:51
即係點呢??
作者: fon2031990 時間: 2017-9-7 03:07
其實係唔係GF良率唔好,所以係EPYC度執D出黎當民用8~16核炸?
作者: qcmadness 時間: 2017-9-7 06:40
其實係唔係GF良率唔好,所以係EPYC度執D出黎當民用8~16核炸?
fon2031990 發表於 2017-9-7 03:07
要良率好先有Threadripper
作者: 燕飛 時間: 2017-9-7 08:21
回復 4 #qcmadness
何解？
via HKEPC Reader for Android
作者: ten123456789 時間: 2017-9-7 11:03
本帖最後由 ten123456789 於 2017-9-7 11:05 編輯
回覆 5# 燕飛
體質最好果5%先拎黎做1950X(AMD PPT寫)
咁實際上都係既 輕鬆4.1GHz 一體水極限有機會上到4.3~4.4GHz
作者: usei 時間: 2017-9-7 11:15
回覆 5# 燕飛
比 1800X 多一倍核心 , 但不用降 clock
良率好至做到
epyc clock 低好多 , 本身就是 trade off 良率
作者: Stiggy930 時間: 2017-9-7 11:16
即係點呢??
chiluns 發表於 2017-9-7 01:51
大約意思係原本管理層無考慮過出Threadripper，Threadripper 係 R&D 團隊為咗自己興而做嘅 side project。前 Intel 高層，現在 AMD 嘅高層 Jim Anderson 聽到 Threadripper 嘅概念後批准Threadripper。
作者: KwokAngus 時間: 2017-9-7 11:54
尋日係forbes 睇哂足足五頁，係好正，證明amd cpu div. 依家係一間相當靈活管理嘅公司，依一點，對amd 黎講係一件好事，有限嘅r&d budget，點解造出價值最大、見到街、賣得出嘅product
作者: boysbehei 時間: 2017-9-7 12:25
基本上，在8至16核，可以行到最高速的EPYC就用來做Threadripper。
其他行唔到高速，就留返EPYC，返正都不用太高速
作者: qcmadness 時間: 2017-9-7 13:37
回復 qcmadness
何解？
via HKEPC Reader for Android
燕飛 發表於 2017-9-7 08:21
1950x係2粒1800x, 仲要高頻
良率唔高邊有咁多好die?
唔好唔記得1950x價錢都"只係"2粒1800x
無番咁上下良率不如賣貴d? (16-core i9 7960x肯定貴過US$1200)
作者: gundamdxhk 時間: 2017-9-7 17:23
同樣係16C32T既7301賣825
1950x賣999, 當然1950x高頻好多
作者: fon2031990 時間: 2017-9-8 10:31
1950x係2粒1800x, 仲要高頻
良率唔高邊有咁多好die?
唔好唔記得1950x價錢都"只係"2粒1800x
無番咁上下良 ...
qcmadness 發表於 2017-9-7 13:37
明白,感謝解答
但我唔明點解都要出一樣都係8C16T的1900X....
同1800X比只係多左PCI總線同Memory channels
作者: Stiggy930 時間: 2017-9-8 11:43
明白,感謝解答
但我唔明點解都要出一樣都係8C16T的1900X....
同1800X比只係多左PCI總線同Memory channels ...
fon2031990 發表於 2017-9-8 10:31
比人做GPU rendering／GPU machine learning。
作者: Ksec 時間: 2017-9-8 14:59
尋日係forbes 睇哂足足五頁，係好正，證明amd cpu div. 依家係一間相當靈活管理嘅公司，依一點，對amd 黎講 ...
KwokAngus 發表於 2017-9-7 11:54
雖然我覺得有 PR 成份，但係如果有七成真到基本上證明左今日嘅 AMD 同以前唔同晒。呢個無論係 Lisa Siu 和 Mark Papermaster 都有功勞。雖然好多人覺得 VEGA 唔好，但係我偏向相信佢嘅潛力未發揮出嚟，之小係 AI / GPGPU 上。
呢段文亦都解釋咗點解 Intel 會有咁多大改動，因為 Threadripper 唔係太多人知道。
下一次， Zen + VEGA 7nm 再接再厲。
作者: bam1a 時間: 2017-9-9 00:38
雖然我覺得有 PR 成份，但係如果有七成真到基本上證明左今日嘅 AMD 同以前唔同晒。呢個無論係 Lisa Siu ...
Ksec 發表於 8/9/2017 16:59
vega個問題處於efficiency未控制得好好
近排先追返1080ti都係咁啦
GPGPU咪發揮去advanced mining department到
要完全發揮,首先都係要係改進efficiency個到,但下一世代嘅N記就到,佢應該仲有條街要追,係金礦未核爆嗰陣真係要慢慢食肥自己,慢慢搞掂佢
作者: qaz11 時間: 2017-9-9 01:24
雖然我覺得有 PR 成份，但係如果有七成真到基本上證明左今日嘅 AMD 同以前唔同晒。呢個無論係 Lisa Siu ...
Ksec 發表於 2017-9-8 14:59
睇intel既反應亂晒大籠, 一係AMD 保密真係做到吓嘢, 一係真係當AMD 死既唔洗理.
作者: inone2 時間: 2017-9-9 06:42
回覆 17# qaz11
初時INTEL真係大意，所以第8代或以前應該無可能嬴RYZEN，但.......2020轉新制程就唔同講法!!!!
作者: qcmadness 時間: 2017-9-9 06:57
回覆 qaz11
初時INTEL真係大意，所以第8代或以前應該無可能嬴RYZEN，但.......2020轉新制程就唔同 ...
inone2 發表於 2017-9-9 06:42
2017都未完
2020先講